Home

CSIM : A Neural Circuit SIMulator User Manual

1. and the following commands tell the recorder r to record the field psr of synapse s as its first trace and the field Vm of neuron n as its second trace gt gt csim connect r s psr gt gt csim connect r n Vm Using the special field spikes the same syntax will be used to record the spikes of neuron n as the third trace of recorder r gt gt csim connect r n spikes The following figure summarizes which objects we have created and how they are connected gt spikes snc Del rooms Pl LifNeuron Vm psr 2 6 Setting up the input Before we are ready to run the simulation we have to define the spike train which should be emitted by the input neuron i In CSIM time varying input signals Sec 3 1 analog or spiking also called the stimulus are not considered to be properties attributes of some objects but are always explicitly specified In the case of our example we will define a spike train with randomly drawn spike times for details see Section 3 1 gt gt S spiking 1 1 spike times 0 analog data gt gt S dt NaN resolution for analog data gt gt S idx i index handle of receiving object gt gt S data sort rand 1 10 10 random spikes in the interval O to 1 sec 2 7 Running the simulation Now we are ready to run the simulation The command 11 gt gt Tsim 1 gt gt csim simulate Tsim S simulates the simpl
2. Y gs t Vin El Linject m c 1 s 1 s 1 with the following meanings of symbols e Cm membrane capacity Farad e Em reversal potential of the leak current Volts e Rm membrane resistance Ohm e N total number of channels active synaptic e g t current conductance of channel c Siemens e Ef reversal potential of channel c Volts e N total number of current supplying synapses e I t current supplied by synapse s Ampere e G total number of coductance based synapses e g t coductance supplied by synapse s Siemens e P reversal potential of synapse s Volts 35 e inject injected current Ampere At time t 0 Vm ist set to Vinit The value of Em is calculated to compensate for ionic currents such that Vm actually has a resting value of Vresting Spiking and reseting the membrane voltage If the membrane voltage Vm exceeds the threshold Viresh the CbNeuron sends a spike to all lts outgoing synapses The membrane voltage is reseted and clamped during the absolute refractory period of length Trefract tO Vreset if the flag doReset 1 This is similar to a LIF neuron see LifNeuron If the flag doReset 0 the membrane voltage is not reseted and the above equation is also applied during the absolute refractory period but the event of threshold crossing is transmitted as a spike to outgoing synapses This is usfull if one includes channels which produce a real action potential see HH K Chann
3. starts at time t 0 Simulation time can be reset to time t 0 with the reset command Sec 6 2 1 csim reset 12 input spike train x10 postsynaptic response 8 T T T T T 6F 4 T 4 J wn o dl 0 L L L L L 1 N L N N 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 membrane potential and spikes 0 T T T T T 0 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 time sec The full code of this example is contained as the file first_model m in the demo directory of the CSIM package 3 Input and Output 3 1 Input Signals When runing a network simulation via csim simulate one can specify input signals like in the command gt gt csim simulate Tsim S where S is a struct array with the following fields e S i idx array of handles of objects which should receives this signal e S i spiking binary flag 0 1 which determines if S i data should be interpreted as spike times or as an analog signal e S i dt time discretization for analog signals S i spiking 0 only e S i data signal data vector of the analog values S i spiking 0 or spike times S i spiking 1 Note that csim simulate accepts an arbirary number of such struct arrays See documentation of the simulate command Sec 6 2 2 13 3 2 Output If one simulates a network then one also wants to record the quantities of interest In our introductory example Sec 2 we were interested in
4. The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Dulnoise noise input current OuGnoise noise input conductance Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 13 CaChannel Yamada98 From Yamada Koch and Adams Multiple Channels and Calcium Dynamics Editors Koch Segev MIT Press 1998 Ca mechanism is included in the channel class not the neuron class Read writable Fields u Mol Constant step increase in the calcium concentration Ts Sec Time constant for the deactivation of the calcium concentration Ca Mol Interior calcium concentration Gbar S The maximum conductance of the channel Erev V The reversal potential Eey of the channel 30 Readonly Fields g S The coductance g t Vm of the channel 7 14 cACNeuron Uses AHP_Channel Read writable Fields STempHeight Volt Height Vthresh V If Vm exceeds Vinre
5. set o fieldname value where o is the handle of the object returned by o csim create classname and fieldname is a string identifying the field i e the name of the field 5 5 Implementaion We implemented this mechanism using four classes e csimClass the base classes of all classes in CSIM which implements the basic set and get methodes for accessable fields e csimClassInfoDB a container of csimClassInfo objects where information about all the at the matlab level available classes is stored e csimClassInfo which stores information accessible fields description about a certain class e csimFieldInfo stores information about a certain field of a given class 5 5 1 Registering classes and fields To make the set and get methodes of a class derived from csimClass work one has to register this class via csimClassInfoDB registerCsimClass and then to register each member variable which we want to be an accessable field via csimClassInfo registerField Class member variable of the types e double and double e float and float e int and int can be made accessible fields Each field has the following associated information see csimFieldInfo 18 e access READWRITE or READONLY e units e lower and upper bound e size the number of elements if the field is of type float x double or int 5 5 2 reggen It woul be a tedious work to code all the csimClassInfoDB registerCsimCla
6. A a read only field with 20 elements units Ohm and a lower and upper bound of 100 and 100 respectively A voltage scale factor readonly size 20 units 0hm range 100 100 double A Source code of reggen e Currently reggen is maintained in 1sm develop reggen e The relevant source file is 1sm develop reggen src defgen cpp e The binaries are lsm develop reggen bin reggen for Linux and 1lsm develop reggen bin reggen exe for windows e To compile reggen under Linux type cd 1sm develop reggen configure make e To compile reggen under Windows XP with MS Visual C type cd 1sm develop reggen make bat msvc 20 6 CSIM command reference The common form of all CSIM commands is csim command where command is one of the following strings create Sec 6 1 1 Create an object of a certain class set Sec 6 1 2 Set fields of an object connect Sec 6 1 3 Connect objects get Sec 6 4 1 Get field values of an object reset Sec 6 2 1 Reset the simulation to t 0 0 simulate Sec 6 2 2 Simulate the network export Sec 6 3 1 Export the network for saving it import Sec 6 3 2 Import a loaded network destroy Sec 6 3 3 Destroy delete the current netork list Sec 6 4 2 or 1s Sec 6 4 2 List various items version Sec 6 4 3 Print out a version string A detailed description of each of the commands follows 6 1 Setting up a network simu
7. InputSignals but in addition returns the cell array R which holds the output traces of all Recorder objects Sec 3 2 R csim simulate Tsim I1 12 In same as csim simulate Tsim I1 12 In but in addition returns the cell array R which holds the output traces of all Recorder objects Sec 3 2 22 e R csim simulate Tsim I1 12 In same as csim simulate Tsim 1I1 12 In but in addition returns the cell array R which holds the output traces of all Recorder objects Sec 3 2 6 3 Saving loading and deleting networks 6 3 1 export The command net csim export produces a struct array net which contains all information about the current network simu lation which is needed to set up the simulation Using save file mat net you can save the whole network simulation NOTE The struct array net returned by net csim export is only intended to be saved to disk for later use by csim import net and does not have a human readable form 6 3 2 import The command csim import net allows you to set up a network simulation from the struct array net which has previously been generated by net csim export In combination with the matlab command load this CSIM command serves the purpose of loading a simulation from disk If the current network either explicitly specified or default network does not contain any objects the net will be imported into tha
8. S The maximum conductance of the channel Erev V The reversal potential Ec of the channel Readonly Fields g S The coductance g t Vm of the channel 7 44 KChannel Korngreen02 Uses KnGate_Korngreen02 and KlGate_Korngreen02 Reference Korngreen A Sakmann B Voltage gated K channels in layer 5 neocortical pyra midal neurones from young rats subtypes and gradients J Physiol 2000 Jun 15 525 Pt 3 621 39 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 5l 7 45 LifBurstNeuron that produces bursts dependent on two additional variables analog to the implementation of dynamical synapses see Kaske amp Bertschinger Traveling wave patterns Model A nonstandard leaky integrate and fire neuron model is implemented where the membrane potential Vm of a neuron is given by dVm Tirri di Vin Vresting Rm Isyn t F Linject Inoise where Tm Cm Rm is the membrane time constant Rm is the membrane resistance syn t is the current supplied by the synapses Inject is a non specific background current and Inoise is a Gaussion random variable with zero mean and a given variance noise Vreset is modulated by u and r to produce spiking dependent on the previous spike pattern At time t 0 Vm ist set toV
9. The multi threaded version of CSIM is currently under development spikeOutput Flag 1 always output spikes as last cell element 0 do not output spikes obsolete The following fields are read only t The current virtual simulation time step The current time step i e step t dt 4 3 Event driven simulation Since in a typical simulation the firing rate of the neurons is about 30 Hz and the time constant of a synaptic current is typically about 3ms most of the time a synapse is idle This observation encouraged us to implement some ideas of event driven simulators like SpikeNet We decided to cut off the very far tail of the exponential decaying synaptic responses after 5 time constants and remove those synapses from the list of active synapses which are advanced every time step A synapse is re activated if it receives a further spike For large networks with lots of synapses these idea results in an average speed up of 2 to 3 5 Adding your own C model classes to CSIM This section describes how a user can add its own models to CSIM at the C level nttp www cnl salk edu arno spikenet 16 5 1 Recipe 1 Create a copy of that existing model which is closest to that you wish to implement in the directory 1sm csim src In the following we assume that the related files are called MyModel h and MyModel cpp 2 Implement your model 3 Add you model to CSIM e Linux Unix Add MyModel o to the list
10. 1 s 1 s 1 with the following meanings of symbols e Cm membrane capacity Farad e Em reversal potential of the leak current Volts e Rm membrane resistance Ohm e N total number of channels active synaptic 33 gc t current conductance of channel c Siemens e Ef reversal potential of channel c Volts e N total number of current supplying synapses e current supplied by synapse s Ampere e G total number of coductance based synapses e g t coductance supplied by synapse s Siemens e EL reversal potential of synapse s Volts Tinject injected current Ampere At time t 0 Vn ist set to Vinit The value of Em is calculated to compensate for ionic currents such that Vm actually has a resting value of Vresting Spiking and reseting the membrane voltage If the membrane voltage Vm exceeds the threshold Viresn the CbNeuronSt sends a spike to all its outgoing synapses and the membrane voltage follows a predefined spike templage during the absolute refractory period of length Tre fract if doReset 1 If the flag doReset 0 the spike template is not applied and the above equation is also applied during the absolute refractory period but the event of threshold crossing is transmitted as a spike to outgoing synapses This is usfull if one includes channels which produce a real action potential see HH_K_Channel and HH_Na_Channel but one still just wants to communicate the spikes as events in time Implem
11. 7 Have fun using CSIM 1 5 Compiling CSIM from the sources The instruction in this section are relevant in two cases a You do not have Linux x86 architecture or Windows as an operating system b You are planning to add your own models written in C to CSIM see section 5 for more information on this topic Currently we support compilation using GNU make for Linux Unix which most likly also works for Mac OS X and nmake Microsoft Visual C for Windows tested only with Windows XP The corresponding Makefiles are 1sm csim src Makefile and 1lsm csim src Makefile win Step by step instructions 1 To compile CSIM from the sources you must first setup Matlab s mex utility properly This can be achieved by issuing the command mex setup at the Matlab prompt For more information on compiling MEX files see the MEX files Guide 2 Optional steps If you are planning to add your own models written in C to CSIM and are not using Linux or Windows you have to build the reggen tool which comes with CSIM 1sm develop reggen for your platform a Go to 1sm develop reggen b Run e configure make for Linux Unix Mac OS e make bat msvc for Windows XP with MS Visual C c If this fails read 1sm develop reggen INSTALL and try to fix the build In this case you will notice that reggen is a modified version of the well known tool doxygen 3 Edit the corresponding Makefile to meet your system configuration 4 Goto csim src a
12. during the simulation The recorded traces of an Recorder with handle rec_idx can be obtained via R csim get rec_idx traces The ecxact form of R depends on the flag commonChannels see below Note that the traces returned always start at time t 0 and are recorded at an interval of dt In addition a recorder can also record spikes from spike emitting objects Via a command like csim connect rec_idx neuron_idx spikes the Recorder with handle rec_idx is set up to record the spikes form the spike emitting objects with handles neuron_idx commonChannels 0 In this case the Matlab array R is a struct array with the only field channel which is in turn a struct array with the following fields e R channel j idx handle of the object from which field the data was recorded e R channel j spiking binary flag 0 1 which determines if data should be interpreted as spike times or as an analog signal e R channel j dt time discretization for analog signals only e R channel j data signal data vector of the analog values or spike times e R channel j fieldName name of the recorded field commonChannels 1 In this case R has two fields WARNING no spikes are returned e R data A double array where R data j s is the s th recorded value of the j th field e R info A struct array where R info j idx is the handle of the object from which the field R info j fieldName is recorded Read writable
13. for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 k 1 The exponent of the gate p The state variable Readonly Fields P 1 The output P t V p t V of the gate 7 36 HChannel Stuart98 Uses HnGate_Stuart98 Reference Stuart G Spruston N Determinants of voltage attenuation in neocortical pyra midal neuron dendrites J Neurosci 1998 May 15 18 10 3501 10 46 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 37 HH K Channel Uses the gate HH_n_Gate Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 38 HH Na_ Channel Uses the gates HH h Gate and HH m Gate Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 39 HHNeuron The model is based on a CbNeuron and includes the HH_K_Channel and HH_Na_Channel for action potetial generation 47 Read writable Fields Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset Vm to after a spike doRe
14. of objects in the Makefile e Windows Add MyModel obj to the list of objects in the Makefile win 4 Compile CSIM e Linux Unix Run the command make e Windows Run the command nmake f Makefile win 5 Use your model 5 2 Details During the compilation of CSIM a tool called reggen is used to generate wrapper code which allows you to access the parameters of your model from the Matlab level This frees you from the tedious work of writing lots of code for getting and setting such fields For more details about this issue see the subsection 5 4 If you are not using Linux or Windows you have to build the reggen tool which comes with CSIM 1sm develop reggen for your platform 1 Go to 1sm develop reggen 2 Run e configure make for Linux Unix Mac OS e make bat msvc for Windows XP with MS Visual C 3 If this fails read 1sm develop reggen INSTALL and try to fix the build In this case you will notice that reggen is a modified version of the well known tool doxygen 5 3 Hints for adding user defined models e Don t forget the line DO_REGISTERING in the class declaration of you model e Use doxygen comments to document your model For Linux and Windows the precompiled reggen tool is located in 1sm develop reggen bin 17 5 4 Setting and getting field values of objects At the Matlab level CSIM allows you to set and get values of fields of objects by means of the commands v csim get o fieldname csim
15. or Parameters up the input Input and Output CDE soa aE ee ee A ee oe ee ee o a 3 2 1 Getting results of the last simulation 04 3 2 2 Getting the results of a multi stimulus simulation Hints for adding user defined models Additional Topics Adding your own C model classes to CSIM Setting and getting field values of objects 0 20 00 08 eae Implementaion 5 5 1 Registering classes and fields NID DD ot or a o o o 11 11 11 12 13 13 14 14 15 15 15 16 16 16 17 17 17 18 18 18 Se TEE A oie ee oe ee ee ee a ee ee Bs 19 6 CSIM command reference 21 6 1 Setting up a network simulation 0 2 02020 00 00 21 Gil OTRAS e ch ee a we be A ae a 21 GLI Boiro a Re A a a 21 Olio CONNECT ci ae be ON Ree a Se ea a 22 6 2 Running the network simulation e e 22 GAl TET e is s ce a Sh ee ew eS ee Bee a 22 G22 SMe Sk ek kw hea eee ee ee A G 22 6 3 Saving loading and deleting networks 0 0002 ee eee 23 Bol EXPO esse kw be ba we De we ae we ee a a ae Se 23 Go IMPOR hee a 23 A II 23 Ga Displaying FALLS TES N cos a A e A e is 23 GA Bici A AAA RA 23 Ode Mist x a e Ak ei PR a E RA Pe 24 GAS WOOO s oeoo ee Sa Oe ee Ee eB Oe Ge ERG bw 25 7 CSIM model classe reference 25 T1 AGhannel Holman o sos u cora a a ee wae ee eS 25 72 AChannel Korngreen02 sosca ca 54 be eee eb daa a eR ee ee 25 Ta AciveLalC amd o c
16. output potential of this neuron Vm0ut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 7 8 AnalogInputNeuron Read writable Fields Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron 27 Readonly Fields Vm V The current output potential of this neuron Vm0ut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 7 9 AnalogTeacher 7 10 ArmModel Read writable Fields minthetal Baseline parameter for thetal mintheta2 Baseline parameter for theta2 minU1 Baseline parameter for U1 minU2 Baseline parameter for U2 model_DT The time step of simulation Xo The X coordinate of origin Yo The Y coordinate of origin mi The mass of 1st link m2 The mass of second link 1c1 The distance of central point of link 1 1c2 The distance of central point of link 2 11 Length of first link 12 Length of second link I1 The moment of inertia of link1 12 The moment of inertia of link2 MFACT The factor by which the external noise will be proportional to the input magnitude PERT_TIME The time at which the perturbation will start DURATION The duration of perturbation inputFileNr The file number
17. reset Vm to after a spike Vinit V The initial condition for Vm at time t 0 Trefract sec The length of the absolute refractory period Inoise A The standard deviation of the noise to be added each integration time constant Tinject A A constant current to be injected into the LIF neuron h internal parameter h deinactivation of Ca channels tau_p tau_p time constant for recovery from inactivation taum taum time constant for inactivation Vh Vh constant for onset of Ca channel activation gh gh constant for relative impact of Ca current to leak current 0 5 2 type Type e g inhibitory or excitatory of the neuron 49 Readonly Fields Vm V The membrane voltage Vm Isyn synaptic input current nIncoming Number of incoming synapses n0utgoing Number of outgoing synapses 7 42 Izhi Neuron 9 see Izhikevich E Simple model of Spiking Neurons Model A canonical neuron model is implemented where the membrane potential Vm of a neuron is given by Rm is the membrane resistance fsyn t is the current supplied by the synapses Linject is a non specific background current and Inoise is a Gaussion random variable with zero mean and a given variance noise At time t 0 Vm ist set to Ving If Vm exceeds the threshold voltage Vihresh it is reset to Vreset and hold there for the length Tre fraci of the absolute refractory period Implementation The simple Euler method is used for numerical
18. source srcIdx object e g a synapse to the destination object dstIdx e g the postsynaptic neuron where dstIdx and srcldx are indices handles to objects dstIdx and srcIdx can also be vectors In that case for each i 1 length srcIdx a signal flow dstIdx i srcldx i is set up csim connect dstldx srIdx vialdx for each i 1 length srcIdx a signal flow of the form dstIdx i viaIdx i srcIdx i is set up where dstIdx srcldx and vialdx are vectors of the same length of indices handles to objects csim connect recIdx objIdx fieldName connects the object with handle objldx to the recorder with handle recIdx The recorder recIdx will then record a trace of the values of the field fieldName of the objects specified by the vector of hadles objIdx Running the network simulation reset csim reset resets the simulation time t back to t 0 0 6 2 2 simulate csim simulate Tsim InputSignals runs the network simulation for Tsim sec onds starting at the time where the last simulate command stopped with input signals InputSignals Sec 3 1 csim simulate Tsim I1 12 In same as above but with input signals 11 to In There is no special meaning in the ordering of the input signals It is equvivalent to concatenate I1 to In into one struct array and pass this array as the single input signal R csim simulate Tsim InputSignals same as csim simulate Tsim
19. the absolute refractory period of length Tre fract if doReset 1 If the flag doReset 0 the spike template is not applied and the above equation is also applied during the absolute refractory period but the event of threshold crossing is transmitted as a spike to outgoing synapses This is usfull if one includes channels which produce a real action potential see HH K Channel and HH Na Channel but one still just wants to communicate the spikes as events in time Implementation The exponential Euler method is used for numerical integration Read writable Fields ge S exc and inh conductances noise gi S exc and inh conductances noise ge0 S exc and inh mean conductances noise gi0 S exc and inh mean conductances noise tau_e S time constants and std for exc and inh conductances noise noise tau i S time constants and std for exc and inh conductances sig_e S time constants and std for exc and inh conductances noise ars A aes sig_i S time constants and std for exc and inh conductances noise Ee V Reversal potential for exc and inh currents noise Ei V Reversal potential for exc and inh currents noise STempHeight Volt Height Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset V to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory
20. the membrane potential of the leaky integrate and fire neuron CSIM can record any field of any object by means of Recorder Sec 7 58 objects The following code fragment shows how to set up a Recorder to record the membrane potential Vm of a LifNeuron Sec 7 46 object with handle n gt gt rec csim create Recorder gt gt csim connect rec n Vm Note that one Recorder object can record an arbitrary number of fields from arbitrary objects 3 2 1 Getting results of the last simulation After a network simulation via a command like gt gt csim simulate Tsim InputSignal the recorded data of the the last simulation Recorder rec can be obtained by the command gt gt R csim get rec traces The exact structure of R for the recorder rec depends on the value of the field commonChannels of the Recorder Sec 7 58 object commonChannels O In this case default R is a struct array with the only field channel which is in turn a struct array with a similar structure as an input signal Sec 3 1 That is e R channel j idx handle of the object from which field the data was recorded e R channel j spiking binary flag 0 1 which determines if data should be inter preted as spike times or as an analog signal e R channel j dt time discretization for analog signals only e R channel j data signal data vector of the analog values or spike times Note that the data always
21. version string e v csim version returns the version string 7 CSIM model classe reference 7 1 AChannel_Hoffman97 Uses AmGate_Hoffman97 and AhGate_Hoffman97 Reference Hoffman D Magee J C Colbert C M and Johnston D Potassium channel regulation of signal propagation in dendrites of hippocampal pyramidal neurons Nature 387 869 875 1997 Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 2 AChannel Korngreen02 Uses AnGate_Korngreen02 and AlGate_Korngreen02 Reference Korngreen A Sakmann B Voltage gated K channels in layer 5 neocortical pyra midal neurones from young rats subtypes and gradients J Physiol 2000 Jun 15 525 Pt 3 621 39 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 25 7 3 ActiveCaChannel Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 4 ActiveChannel The Model Generic active channel with arbitray number of ion gates g t Vm is computed as g t Vm 9 Pit Vin where g is th
22. voltage Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron myIndex ID of external input is equal index of array beReal Flag if external input is used or normal input Readonly Fields Vm V The current output potential of this neuron Vm0ut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 7 33 GaussianAnalogFilter Read writable Fields m_std_dev Standard deviation of the filter mask nKernelLength The length of the kernel 45 Readonly Fields nChannels Number of Input channels to filter 7 34 GVD _cT_Gate Read writable Fields Vh Volt Inflection point of the sigmoidal function p V Vc Volt Slope of the sigmoidal function p V Ts Voltage independent time constant 7 nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 k 1 The exponent of the gate p The state variable Readonly Fields P 1 The output P t V p t V of the gate 7 35 GVD Gate Read writable Fields Vh Volt Inflection point of the sigmoidal function p V Vc Volt Determines the slope of the sigmoidal function p V Ts Scale of the time constant T V Te Additional exponential factor of the time constant T V nummethod flag Numerical method
23. CSIM A Neural Circuit SIM ulator Version 1 1 User Manual 2002 The IGI LSM Group www lsm tugraz at July 10 2006 This document is part of CSIM Release 1 1 Copyright 2002 The IGI LSM group CSIM is free software you can redistribute it and or modify it under the terms of the GNU General Public License as published by the Free Software Foundation either version 2 or at your option any later version CSIM is distributed in the hope that it will be useful but WITHOUT ANY WARRANTY without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License for more details To get a copy of the GNU General Public License point your browser to http www gnu org copyleft gpl html The IGI LSM group Institute for Theoretical Computer Science Graz University of Technology Inffeldgasse 16 b A 8010 Graz AUSTRIA 1smCigi tu graz ac at www 1lsm tugraz at Contents 1 Preliminaries 1 1 Ls 1 3 1 4 1 5 Features of the current version Getting and Installing CSIM Compil ing CSIM from the sources 2 A short Tutorial sl 2 2 2 3 2 4 2 5 2 6 2T 2 8 3 1 3 2 4 1 How the network is simulated 4 2 Global fields 4 3 Event driven simulation 5 1 5 2 5 3 5 4 5 5 Creating Objects Handles Object Setting Making connections Recording Setting Running the simulation Plotting the recorded traces Input Signals Fields
24. Channel Hoffman97 39 Read writable Fields STempHeight Volt Height Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset V to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 24 dNACOUNeuron Uses AChannel_Hoffman97 Read writable Fields ge S exc and inh conductances noise gi S exc and inh conductances noise ge0 S exc and inh me
25. Fields commonChannels Flag 1 output all channels in one matrix WARNING no spikes are returned yet 0 output each recorded field as seperate channel dt sec The timestep at which an recording should be done no meaning if recording spikes Tprealloc sec Provide your best guess how long the network will be simulated in simulation time enabled Flag 0 recorder disabled 1 recoder enabled 57 7 59 SICChannel Maciokas02 Uses SICmGate_Maciokas02 and SIChGate_Maciokas02 Reference Maciokas J Goodman P Kenyon J Accurate Dynamical Model of Interneu ronal GABAergic Channel Physiologies Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 60 SigmoidalNeuron Read writable Fields thresh W Itot logsig beta Itot thresh beta 1 Itot logsig beta Itot thresh taum time constant A max 1 the output of a sig neuron is scaled to 0 W_max I_inject W external current injected into neuron Vm_init V initial membrane voltage Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron Readonly Fields Vm V The current output potential of this neuron Vm0ut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming N
26. Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses n0utgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 16 CaGate Yamada98 From Yamada Koch and Adams Multiple Channels and Calcium Dynamics Editors Koch Segev MIT Press 1998 Can only be connected to CaChannel_Yamada98 class where the Ca mechanism is included 32 Read writable Fields Ts Scale of the time constant T Ca nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 k 1 The exponent of the gate p The state variable Readonly Fields P 1 The output P t V p t V of the gate 7 17 CALChannel Destexhe98 Uses CALmGate_Destexhe98 and CALhGate_Destexhe98 Reference Destexhe A Contreras D Steriade M Mechanisms underlying the synchronizing action of corticothalamic feedback through inhibition of thalamic relay cells J Neurophysiol 1998 Feb 79 2 999 1016 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 18 CbNeuronSt The Model The membrane voltage Vm is governed by Vin Vp Em Ns Gs Cm dt R 5 gc t Vin Ena 5 Is t 5 gs t Vm BE Lingect m e
27. ON simulations are performed quite fast We also implemented some ideas from event driven simulators like SpikeNet which result on an average speedup of 3 assum ing an average firing rate of the neurons of 20 Hz and short synaptic time constants compared to a standard fixed time step simulation scheme Runs on Windows and Unix Linux CSIM is developed on Linux but also runs under Windows XP we have no more experience with Windows 98 yet but it should also run there and should in principle run on any platform for which Matlab is available External Interface There is an external interface which allows CSIM to communicate with external programs In this way one can for example control the miniature robot Khepera with CSIM This feature is not available in the Windows version 1 4 Getting and Installing CSIM CSIM is distributed under the GNU General Public License and can be downloaded from http www igi tugraz at csim To install CSIM perform the following steps 1 Donwload CSIM from www igi tugraz at csim 2 Unzip the file csim VER zip where VER stands for the version you have downloaded This will create a subdirectory 1sm and 1sm csim 3 Start Matlab and change into the directory 1sm 4 Run the Matlab script install m 5 Add the path 1sm to the Matlab search path e g e addpath home jack 1sm or e addpath C Work Neuroscience lsm 6 Change into the directory 1sm csim demos and play around with them
28. OO SiemoidalNenroii s ane sa Goia a A Be aia au a 58 7 61 SpikinginputNeuron 225 584 5448 bE aa aaa aa wae a ee a 58 T02 SpikingTeacher o c aoe doa kee ee eed ae ae E a 59 TOs StaticAnalog BAYTARRE e eoo os os os e a e eu E e E a E ee ae a 59 7 64 StaticAnalogSynaps os s acasa mp e aae ha a a ee ee a ED 59 7 65 StaticSpikingCbSynapse o edia ni Cenaa aaia 59 7 06 StaticSpikingSynapse s so eos aoe w oe ee E a e a 60 TO SAESP VIADEO 2 esc A a HEP eGR eee A a 60 TOS Traubs HHR Channel es oc a a A A oiL a 61 7 69 Traubs HH Na Channel e s a sos bee doa wa ra a ee a a 61 TADO TraubsHHANEUrON ee ce e e aa ee a ee ee ee ee 61 Tol TriangularAndlog FIRST es cu edea aoa e a a ed ee E e Ea 63 1 72 UserAnalog FUREE o a arra eh ak aaa ea bane ha A pA aA Ree aa 63 1 Preliminaries 1 1 What is CSIM CSIM is a tool for simulating heterogeneous networks composed of different model neurons and synapses This simulator is written in C with an MEX interface to Matlab It is intended to simulate networks containing up to 10 000 neurons and up to the order of a few millions of synapses the actual size depends of course on the amount of RAM available on your machine 1 2 About this Manual This manual is intended to describe how to use CSIM from the Matlab users point of view It does not try to explain or give an introduction to the type of models which can be simulated with CSIM Regarding neural modeling we refer the reader to Da
29. STDP tauneg Timeconstant of exponential decay of negative learning window for STDP dw STDPgap No learning is performed if Delta tpost tpre lt ST DPgap activeSTDP Set to 1 to activate STDP learning No learning is performed if set to 0 useFroemkeDanSTDP activate extended rule by Froemke and Dan default 1 Wex The maximal minimal weight of the synapse Aneg Defines the peak of the negative exponential learning window Apos Defines the peak of the positive exponential learning window mupos Extended multiplicative positive update dw Wer W P09S x Apos x exp Delta taupos Set to 0 for basic update See Guetig Aharonov Rotter and Sompolinsky 2003 Learning input correlations through non linear asymmetric Hebbian plasticity Journal of Neuroscience 23 pp 3697 3714 muneg Extended multiplicative negative update dw W x Aneg x exp Delta tauneg Set to 0 for basic update tau sec The synaptic time constant 7 W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay 60 Readonly Fields psr The psr postsynaptic response is the result of whatever computation is going on in a synapse steps2cutoff 7 68 Traubs_HH_K_Channel Used in the TraubsHHNeuron neuron model Uses the gate Traubs_HH_n_Gate gKd 0 03 S cm2 EK 90 mV Read writable Fields Gbar S The maximum conductance of the channel Erev V T
30. Vm is governed by Ne Ns Gs dt Rm Vm Vm Hm c s ezz Cm 5 Ielt Vm Egon gt 104 gs t VEIA 1 Vi c 1 s 1 s 1 with the following meanings of symbols e Cm membrane capacity Farad e Em reversal potential of the leak current Volts e Rm membrane resistance Ohm Ne total number of channels active synaptic e g t current conductance of channel c Siemens e ES C reversal potential of channel c Volts e N total number of current supplying synapses I t current supplied by synapse s Ampere e G total number of coductance based synapses e g t coductance supplied by synapse s Siemens e EL reversal potential of synapse s Volts e g t excitatory coductance given Ornstein Uhlenbeck process noise vise reversal potential for excitatory Ornstein Uhlenbeck process noise inh gu t inhibitory coductance given Ornstein Uhlenbeck process noise e Fink reversal potential for inhibitory Ornstein Uhlenbeck process noise 37 EXC nose inh Gnoise t Vin L e inject injected current Ampere At time t 0 Vm ist set to Vinit The value of Em is calculated to compensate for ionic currents such that Vm actually has a resting value of Vresting Spiking and reseting the membrane voltage If the membrane voltage Vm exceeds the threshold Viresn the CbNeuronSt sends a spike to all its outgoing synapses and the membrane voltage follows a predefined spike templage during
31. an conductances noise gio S exc and inh mean conductances noise tau_e S time constants and std for exc and inh conductances noise tau_i S time constants and std for exc and inh conductances noise sig_e S time constants and std for exc and inh conductances noise sig i S time constants and std for exc and inh conductances noise Ee V Reversal potential for exc and inh currents noise Ei V Reversal potential for exc and inh currents noise 40 STempHeight Volt Height Vthresh V If Vm exceeds Vinresn a Spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter V should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the lea
32. and inh conductances noise tau_i S time constants and std for exc and inh conductances sig_e S time constants and std for exc and inh conductances noise noise sig i S time constants and std for exc and inh conductances noise Ee V Reversal potential for exc and inh currents noise Ei V Reversal potential for exc and inh currents noise STempHeight Volt Height Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset V to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Dulnoise noise input current DuGnoise noise input conductance
33. annel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses n0utgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 12 bNACOUNeuron Uses AHP_Channel Read writable Fields ge S exc and inh conductances noise gi S exc and inh conductances noise ge0 S exc and inh mean conductances noise gi0 S exc and inh mean conductances noise tau_e S time constants and std for exc and inh conductances noise 29 tau_i S time constants and std for exc and inh conductances noise sig_e S time constants and std for exc and inh conductances noise sig_i S time constants and std for exc and inh conductances noise Ee V Reversal potential for exc and inh currents noise Ei V Reversal potential for exc and inh currents noise STempHeight Volt Height Vthresh V If Vin exceeds Vinresn a Spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V
34. brane voltage Vm 7 47 linear classification Read writable Fields nClasses Number of Classes range low Lower bound of algorithms range range high Upper bound of algorithms range 53 Readonly Fields nInputRows Number of rows for input vectors 7 48 LinearNeuron Read writable Fields Iinject W external current injected into neuron Vinit V initial membrane voltage Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron Readonly Fields Vm V The current output potential of this neuron Vm0ut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 7 49 LinearPreprocessor Readonly Fields nInputRows Number of rows for input vectors nOutputRows Number of rows for output vectors 7 50 linear_regression Read writable Fields range_low Lower bound of algorithms range range high Upper bound of algorithms range Readonly Fields nInputRows Number of rows for input vectors 7 51 MChannel Mainen96 Uses MnGate_Mainen96 Reference Mainen ZF Sejnowski TJ Influence of dendritic structure on firing pattern in model neocortical neurons Nature 1996 Jul 25 382 6589 363 6 54 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of t
35. ction 4 2 During each time step a certain method advance is called for each object part of the simulation There is a fixed schedule which depends a on the class of the object and b on the time of creation Here is the order of the classes 1 Neuron 2 Synapse 3 Recorder This means that all objects of classes derived from a generic Neuron are advanced first followed by objects of classes derived from a generic Synapse and so on Note that for example a more detailed neuron model like the CbNeuron Sec 7 19 has to advance all its child objects like ion channels After a csim simulate command returns the network is kept in its current state at this time t A further csim simulate command continues the simulation at this time t Use csim reset to reset the simulation to time t 0 Note that recorded traces you get by the appropriate get command Sec 6 4 1 always start at time t 0 15 4 2 Global fields There are a few fields which determine the global i e not object specific behavior of CSIM which can be modified via the set command Sec 6 1 2 gt gt csim set lt fieldname gt lt value gt Read writeable fields dt The integration time step used by all objects randSeed The seed value for the random number generator verboseLevel How much stuff should be written to the console window no effect yet nThreads Number of threads to use for simulation no effect yet
36. e delay sec The synaptic transmission delay Readonly Fields u The time varying state variable u for facilitation r The time varying state variable u for depression psr The psr postsynaptic response is the result of whatever computation is going on in a synapse steps2cutoff 7 28 ExpSpikeFilter Read writable Fields m_taul Time constant Readonly Fields nChannels Number of Input channels to filter m_dt Time step length nValuesPerChannel Number of values stored for each channel in last Values 43 7 29 ExtInputNeuron Read writable Fields Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron myIndex ID of external input is equal index of array beReal Flag if external input is used or normal input Readonly Fields Vm V The current output potential of this neuron VmOut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming Number of incoming synapses n0utgoing Number of outgoing synapses 7 30 ExtOutLifNeuron Read writable Fields myIndex ID of external input is equal index of array beReal Flag if external input is used or normal input Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vthresh V If Vm exceeds Vinresn a Spike is emmited Vresting V The resting membrane voltage Vreset V The voltage to reset Vm to after a sp
37. e Lo a Ee a we OO Ee a A a 26 TA Active Channel 2 ca a hae he he eee Pe a ee 26 To AHP Chomnel o ee he ee a ee a ee 26 TO AlphasSpikePilter 2 2 2 24 0 kb eb eee ee ee De ee 27 7 7 AnalogFeedbackNeuron 1 0 ee ee es 27 T8 An aloglapuiNeuron o ee re ewe oe ee ee ee 27 TO AnalogTeacher oe coe Sew Po eo a ee ee A 28 TW ArmModel 265025 he Gh tae kee te eka eee ee bee he Ea 28 Poll BRAIN o cos see A ee ee OM ae eee ee ee 29 T12 PNACOUNGUTOM 2 665504 be de he eee eae bee Bee hee eee 29 7 14 CaGhannel YamadadS s co e eee ee a ee ee 30 FLA CACNEITON se i aci ba Ree ee eee ee Pa Ae RS 31 T19 CARN os nee A a ae we a ee ae E a 31 7 16 CaGate YamadaB8 oo se ceata ioe keg a ee a a e a 32 tly 7 18 7 19 7 20 tel Ges 7 23 7 24 7 25 1 26 Gar 7 28 1 20 7 30 7 31 7 32 7 33 7 34 7 39 7 36 Gar 7 38 7 39 7 40 7 41 7 42 7 43 7 44 7 45 7 46 7 47 7 48 7 49 7 00 7 51 CALChannel DestexheOS o coo a Re ee 33 LIN SUE xd a we eee a a A oe 33 ASO OURO los oe eed o A ta ew a ei 35 ea AI 37 A OMNES DIMER INET ue eles Oe Bel howe Ge ag ak e wed See Fe 39 DiscretizationPreprocessor e sosse e 39 CANN QUEDO So ey ee e Rad ee RA ee ee De Vd 39 ONACUUNeWION eo s ee ka a s dogok Eaa aa E d ae we eee as 40 DynaraicSpikingCbSynapse oca a a oe ee ee oe ke ee E 41 DynaraicSpikingSynapse s a o sacs e cati eud ukora aa a 42 Dinamo tp DADES e sa a eh E e ety A E we Rae a 42 ExpopikeFilter o edcka dias AR eee AA eS 43
38. e maximal conductance and P t Vm 0 1 is fraction of i type gates currently open Note that no specific gates are modeled Instead arbitrary ion gate objects can be connected to the channel Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 5 AHP_Channel Reference Fuhrmann G Markram H Tsodyks M Spike frequency adaptation and neocortical rhythms J Neurophysiol 2002 Aug 88 2 761 70 Read writable Fields Gbar S The maximum conductance of the channel u Constant step increase in n for each spike Ts Time constant for the deactivation of the current Erev V The reversal potential Ec of the channel 26 Readonly Fields n Fraction of the open conductance g S The coductance g t Vm of the channel 7 6 AlphaSpikeFilter Read writable Fields m_taul Time constant 1 m_tau2 Time constant 2 Readonly Fields nChannels Number of Input channels to filter m_dt Time step length nValuesPerChannel Number of values stored for each channel in last Values 7 7 AnalogFeedbackNeuron Read writable Fields feedback Feedback Mode 0 external input 1 feedback Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron Readonly Fields Vm V The current
39. e model for 1 sec starting at time t 0 with the stimulus S 2 8 Plotting the recorded traces The traces recorded by the recorder can be obtained by the command gt gt t csim get r traces which yields the output t channel 1x3 struct A closer look at e g the first channel reveals that t channel is a struct array with a similar structure as we have seen above for the input signal gt gt t channel 1 ans idx 1 fieldName psr data 1x2000 double spiking 0 dt 5 0000e 04 The output indicates that t channel 1 data holds the psr trace t channel 1 fieldName with a resolution of 0 5ms t channel 1 dt See the Recorder class documentation Sec 7 58 for details abput the structure of the trace output Similarly t channel 2 holds the the Vm trace and t channel 3 holds the output spike times Hence the following command will plot the psr trace gt gt plot t channel 1 dt t channel 1 dt Tsim t channel 1 data and the command gt gt stem t channel 3 data ones size t channel 3 data will draw the output spike train Using another set of plot commands one can easily create the following figure which shows the input the postsynaptic response and the output of the neuron the voltage and the spikes 3In general the simulation will be continued at the time where the last simulate command Sec 6 2 2 stopped However the first simulate command as in the case of the example
40. eae Go es Sb eee ee a PD E ewe od eee Pee 44 PAL NERO cea ke ee ae p E a ee Sa ee a 44 sd do le see ee Rod ek hk ee PS ee GO a ia 45 PxtOutsigmoidalNeuron s s a o 2 kk a sa RE eR we eR es 45 ansiada nao Fiter io ee es hoe ee gon Oe ee Be eR a 45 isv De eae fe ee eek ei a eed Sema ea eh ee Re we oe 46 E eed Ge koh oe Be ee ew Ra i 46 in in a A 46 HHR Channel A i as Ae Oe a ee Bs 47 HH Na Channel ec cos ee Cw ek eae pE a ee Sa d e Eas 47 ANCORA ey e e a ok eas Pe ee ne e E A eed a Pe e a 47 HVACAChannel_Brown93 aoaaa aaa a 48 o sce de che i i a A ee ee Be ee Bo a e EAA 49 Toni NWO sn oe ee ee ea ed eee ee eR OR RR a RG a ioa 50 KCAC pannel MamenOG lt ooo sos fog kk eh ge ee we a ee 51 KChannel Kornsreen02 0c kk ke RE RR ee es 51 dsd A ee Pak Ae Oe a e Bs 52 LIINEUT N ee ee ee a Ke eae ee EE ee Sa ee Cae 53 lnearelassihegtioni a Oe ee a RR A Ea 53 LiR ar Nero a ce caa eo hehe he ee a a dee 54 LinearPreprocessor ee 54 linear regressi cocaina ee ke a 54 MChannel Malnendl o era ee ee eee a N 54 ToS OK hannel Was Se ook oh ae au a ee EE Re Ea 55 7 00 Mean Std Preprocessor 6 is oe ee 55 1 01 Mexkecorder 2 55 be ek bea ee ee a ee EE A a G 55 7 55 NPChannel McCormick02 o o coue a a a a e ee a i 56 T56 PCAPreprot ss t lt co esni we k e ee RR p ARA RR e 56 Tol Readout so s ae ae e a a ie a Pe be hee a a a a 56 TOS TOOONIET o a e ia Pe OP ee ee a O A 57 7 59 SlOChannel MaciokasO2 oo oa c sacara coir ee 58 T
41. ec 6 1 1 of the given class gt gt i csim create SpikingInputNeuron gt gt s csim create DynamicSpikingSynapse gt gt n csim create LifNeuron 2 2 Object Handles The values i s and n returned by these create commands Sec 6 1 1 are the indices or handles to the created objects The handles are the only means to further access or modify existing objects 2 3 Setting Fields or Parameters Each of the created objects has several fields which usually correspond to a parameter of the model which is implemented by the class of the object To display for example all fields of the LifNeuron Sec 7 46 object we can use the command gt gt csim get n This will yield the following output O LifNeuron Cm 3e 08 F Rm 1e 06 Ohm Vresting 0 06 V Vreset 0 06 V Vinit 0 06 V Trefract 0 003 sec Inoise O A Iinject 0 A Vthresh 0 045 V Vm 0 V type 0 nSpikes 0 nIncoming 0 nOutgoing 0 Some of the fields are parameters of the model Cm Rm which can be modified denoted by the others are internal state variables Vm in this case which can not be changed denoted by the and yet other provide auxiliary information nSpikes nIncoming See the Class Reference Sec 7 for details about fields of a particular object Suppose we want to change the absolut refractory period of the LIF neuron n to be of l
42. el and HH _Na Channel but one still just wants to communicate the spikes as events in time Implementation The exponential Euler method is used for numerical integration Read writable Fields Vthresh V If Vm exceeds Vinresn a Spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter V should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Iinject A Constant current to be injected into the CB neuron 36 Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 20 CbStOuNeuron The Model The membrane voltage
43. ength 2ms and add a noisy current of 50nA This can be done with the command gt gt csim set n Trefract 0 002 Inoise 50e 9 For the dynamic synapse s we choose parameters such that it will show a depressing behaviour gt gt csim set s W 10 U 0 1 D 1 F 0 05 gt gt csim set s u0 0 1 r0 1 2 4 Making connections So far we have generated three independent objects and set their fields to the desired values Now we have to connect the objects to implement our simple model We have to connect the input i to the synapse s and the synapse to the neuron n gt gt csim connect n s 4 synapse to neuron gt gt csim connect s i input to synapse Note the the connect command Sec 6 1 3 uses the convention that the signal destination is the first argument Alternatively one can use the three argument form of the command gt gt csim connect n i s input to neuron via synapse 10 2 5 Recording The model is now fully implemented In addition we want to record traces of some quantities of interest Suppose we want record the membrane potential and the spikes of neuron n as well as the postsynaptic respones the field psr of the synapse To do so we have to create a Recorder object Sec 7 58 and tell this object which fields from which objects to record The recorder is created with the command gt gt r csim create Recorder
44. entation The exponential Euler method is used for numerical integration Read writable Fields STempHeight Volt Height Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm 34 Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 19 CbNeuron The Model The membrane voltage Vm is governed by Vin Vp Em 2 N Gs Ona 7 Y gelt Vm ES Y Is t
45. for input of Xdest Ydest Thetal Theta2 filename is T x mat where x is this number Readonly Fields ti Angle theta 1 t2 Angle theta 2 wi Angular velocity 1 updated every time step w2 Angular velocity 2 updated every time step ul Torque 1 updated every time step u2 Torque 2 updated every time step nIncoming Number of inputs nOutputChannels Number of output channels 28 7 11 bNACNeuron Uses AHP_Channel Read writable Fields STempHeight Volt Height Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage ch
46. he channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 52 MChannel_Wang98 From Wang H S Pan Z Shi W Brown B S Wymore R S Cohen I S Dixon J E and McKinnon D 1998 KCNQ2 and KCNQ3Potassium Channel Subunits Molecular Correlates of the M Channel Science 282 1890 1893 Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 53 Mean_Std_Preprocessor Every row x i of the input vector is transformed into x x m std where mi and std_i are the mean and standard deviation of the i th row Readonly Fields nInputRows Number of rows for input vectors nOutputRows Number of rows for output vectors 7 54 MexRecorder Read writable Fields commonChannels Flag 1 output all channels in one matrix WARNING no spikes are returned yet 0 output each recorded field as seperate channel dt sec The timestep at which an recording should be done no meaning if recording spikes Tprealloc sec Provide your best guess how long the network will be simulated in simulation time enabled Flag 0 recorder disabled 1 recoder enabled 55 7 55 NPChannel McCormick02 Uses NPmGate_McCormick92 Reference McCormick DA Huguenard JR A model of the electrophysiological
47. he reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 69 Traubs_HH_Na_Channel Used in the TraubsHHNeuron neuron model Uses the gates Traubs_HH_h_Gate and Traubs_HH_m_Gate gNa 0 1 S cm2 ENa 50 mV Read writable Fields Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 70 TraubsHHNeuron This model is a modified version of Traub s model of Na K currents for action potentials in hippocampal pyramidal cells which was modified from HH equations to fit the kinetics of those neurons the rate constants correspond to a temperature of 36 deg C The model is based on a CbNeuron and includes the Traubs_HH_K_Channel and Traubs HH_Na_Channel for action potetial generation INa gNa m 3 h v ENa IKd gKd x n 4 x v EK 61 w m inf m taum bh hinf h tau h n ninf n taun v2 v Vtr alpha m 0 32 x 13 v2 exp 13 v2 4 1 beta m 0 28 x v2 40 exp v2 40 5 1 tau_m 1 alpha_m beta_m m_inf alpha_m alpha_m beta_m alpha_h 0 128 x exp 17 v2 18 beta_h 4 1 exp 40 v2 5 tau_h 1 alpha_h beta_h h inf alpha_h alpha h beta_h alpha_n 0 032 x 15 v2 exp 15 v2 5 1 beta n 0 5 exp 10 v2 40 taun 1 alpha_n beta_n n_inf alpha_n alpha_
48. ike Vinit V The initial condition for Vm at time t 0 Trefract sec The length of the absolute refractory period Inoise A The standard deviation of the noise to be added each integration time constant Tinject A A constant current to be injected into the LIF neuron type Type e g inhibitory or excitatory of the neuron Readonly Fields Isyn synaptic input current nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses Vm V The membrane voltage Vm 44 7 31 ExtOutLinearNeuron Read writable Fields Iinject W external current injected into neuron Vinit V initial membrane voltage Vresting V The resting membrane voltage Inoise 4 The noise of analog neurons type Type e g inhibitory or excitatory of the neuron myIndex ID of external input is equal index of array beReal Flag if external input is used or normal input Readonly Fields Vm V The current output potential of this neuron Vm0ut The vlaue wich will actualle be propagated to the outgoing synapses nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 7 32 ExtOutSigmoidalNeuron Read writable Fields thresh W Itot logsig beta Itot thresh beta 1 Itot logsig beta Itot thresh taum time constant A max 1 the output of a sig neuron is scaled to 0 W_max I_inject W external current injected into neuron Vm_ init V initial membrane
49. init If Vm exceeds the threshold voltageVinresn it is reset to Vrese and hold there for the length T refract of the absolute refractory period Implementation The exponential Euler method is used for numerical integration Read writable Fields Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vthresh V If Vm exceeds Vinresn a Spike is emmited Vresting V The resting membrane voltage Vreset V The default voltage to reset Vm to after a spike VBthresh V spike pattern dependent treshold Vm to after a spike Vinit V The initial condition for Vm at time t 0 Trefract sec The length of the absolute refractory period Inoise A The standard deviation of the noise to be added each integration time constant Tinject A A constant current to be injected into the LIF neuron uB internal parameter uB recovery from spike facilitation with time constant FB rB internal parameter rB recovery from spike depression with time constant DB FB FB time constant for recovery from facilitation of spike generation DB DB time constant for recovery from depression of spike generation UB UB constant for onset of facilitation of spike generation type Type e g inhibitory or excitatory of the neuron Readonly Fields Vm V The membrane voltage Vm Isyn synaptic input current nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 52 7 46 LifNeu
50. integration Read writable Fields Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vthresh V If Vm exceeds Vinresn a spike is emmited Vresting V The resting membrane voltage Vreset V The voltage to reset Vm to after a spike Vinit V The initial condition for Vm at time t 0 Trefract sec The length of the absolute refractory period Inoise A The standard deviation of the noise to be added each integration time constant Tinject A A constant current to be injected into the LIF neuron a A constant 0 02 01 describing the coupling of variable u to Vm b A constant controlling sensitivity og u c A constant controlling reset of Vm 1000 Vreset d A constant controlling reset of u type Type e g inhibitory or excitatory of the neuron 50 Readonly Fields Vm V The membrane voltage Vm Vb The membrane voltage in millivolt Vint The membrane voltage in millivolt u internal variable ub The internal variable u adapted to millivolt and millisecond Isyn synaptic input current nIncoming Number of incoming synapses n0utgoing Number of outgoing synapses 7 43 KCAChannel Mainen96 Uses KCAnGate_ Mainen96 Reference Mainen ZF Sejnowski TJ Influence of dendritic structure on firing pattern in model neocortical neurons Nature 1996 Jul 25 382 6589 363 6 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar
51. iving objects of the object specified by the index handle idx This is currently only implemented for Synapses and Neurons R csim get recorder_idx traces returns the traces of the fields recorded by the Recorder Sec 7 58 object with handle recorder_idx during the last simulation See the Recorder documentation Sec 7 58 for details about the format of R list csim list and csim list classes print a list of all available classes csim list classes fields prints a list of all available classes with infor mation about the fields of each class Read writeable fields are marked by a between the field name and its description wheres a denotes a read only field csim list objects prints a list of all created objects csim list objects fields prints a list of all created objects and the values of its fields Read writeable fields are marked by a between the field name and its value wheres a denotes a read only field csim list networks prints a list of all networks csim list networks fields prints a list of all networks and the values of its fields Read writeable fields are marked by a between the field name and its value wheres a denotes a read only field Not that instead of list one can use 1s as a shortcut 24 6 4 3 version e csim version prints a
52. kage channel Vm V The membrane voltage Dulnoise noise input current DuGnoise noise input conductance Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 25 DynamicSpikingCbSynapse The conductance g t of the synapse is increased by W r u when a presynaptic spike hits the synapse and decays exponentially time constant T otherwise u and r model the current state of facilitation and depression Read writable Fields E Equilibrium potential given by the Nernst equation U The use parameter of the dynamic synapse D sec The time constant of the depression of the dynamic synapse F sec The time constant of the facilitation of the dynamic synapse u0 Value of the time varying facilitation state variable u for the first spike ro Value of the time varying depression state variable r for the first spike tau sec The synaptic time constant T W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay 41 Readonly Fields u The time varying state variable u for facilitation r The time varying state variable u for depression psr The psr postsynaptic response is the result of whatever computation is going on in a synapse steps2cutoff 7 26 DynamicSpikingSynapse The time varying s
53. l Readonly Fields nChannels Number of Input channels to filter 7 72 UserAnalogFilter Read writable Fields m_weights User defined weights nKernelLength The length of the kernel Readonly Fields nChannels Number of Input channels to filter References Dayan and Abbott 2001 Dayan P and Abbott L 2001 Theoretical Neuroscience Computational and Mathematical Modeling of Neural Systems MIT Press See also http people brandeis edu abbott book Gerstner and Kistler 2002 Gerstner W and Kistler W 2002 Spiking Neuron Models Cambridge University Press See also http diwww epf1 ch gerstner BUCH html 63
54. lation 6 1 1 create e idx csim create class creates an object of the class specified by the string class and returns a uint32 index or handle for the created object This index handle is the only means to access the created object e idx csim create class N creates N objects of the class specified by the string class and returns a uint32 vector of indices handles for the created objects The handles returned by idx csim create are used to identify the objects later in set get and connect calls See the Class Reference Sec 7 for a full description of all available classes 6 1 2 set e csim set field value sets the network field Sec 4 2 specified by the string field to value e csim set idx field value sets the field specified by the string field of the object with index handle idx to value idx can also be a vector of indices handles of objects of the same class If idx is a vector and value is a scalar then the fields of all objects are set to value If idx and value are vectors of the same size then the field field of all objects idx i is set to value i 21 6 1 3 6 2 6 2 1 csim set idx f1 vi f2 f2 fn vn sets the fields specified by the strings f1 f2 fn of the object with index handle idx to the values v1 v2 vn idx and v1 v2 vn can also be vectors connect csim connect dstIdx srcIdx sets up a signal flow from the
55. n beta_n HH parameters gNa 0 1 S cm2 gkd 0 03 S cm2 ENa 50 mV EK 90 mV Vtr 63 mV adjusts threshold to around 50 mV Read writable Fields Vthresh V If Vm exceeds Vinresn a spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 62 7 71 TriangularAnalogFilter Read writable Fields nKernelLength The length of the kerne
56. nd run e make for Linux Unix Mac OS e nmake f Makefile win for Windows XP The compiled MEX files csim mexglx and csim d11 which already come with CSIM should run on any Linux or Windows operationg system respectively For Linux and Windows the precompiled reggen tool is located in 1sm develop reggen bin We do not know whether this instructions work on systems other than Windows XP with MS Visual C and Linux with gcc 3 4 or higher The External Interfaces API section of the Mathworks web site and the Tech Note 1601 contain valuable information about compiler requirements for compiling C MEX files 2 A short Tutorial In this section we will introduce CSIM by means of a simple example We will use CSIM to simulate a model where a leaky integrate and fire Sec 7 46 neuron LIF neuron is driven by a Poisson spike train which is transmitted by a dynamic synapse Sec 7 26 input spike train encon output spike train Hl s I I dynamic synapse 2 1 Creating Objects Each element entity of the simple model we will implement will be simulated by a correspond ing object in CSIM The following table shows the correspondence between the elements of the model and the class of the object used to simulate the element element of model class of CSIM object input spike train SpikingInputNeuron dynamic synapse DynamicSpikingSynapse LIF neuron LifNeuron Hence for the simulation to run one must create objects S
57. ndritic backpropagating spike the synapse sees the postsynaptic spike delayed by back_delay 42 tauspost Used for extended rule by Froemke and Dan See Froemke and Dan 2002 Spike timing dependent synaptic modification induced by natural spike trains Nature 416 3 2002 tauspre Used for extended rule by Froemke and Dan taupos Timeconstant of exponential decay of positive learning window for STDP tauneg Timeconstant of exponential decay of negative learning window for STDP dw STDPgap No learning is performed if Delta tpost tpre lt ST DPgap activeSTDP Set to 1 to activate STDP learning No learning is performed if set to 0 useFroemkeDanSTDP activate extended rule by Froemke and Dan default 1 Wex The maximal minimal weight of the synapse Aneg Defines the peak of the negative exponential learning window Apos Defines the peak of the positive exponential learning window mupos Extended multiplicative positive update dw Wer W P S x Anos x exp Delta taupos Set to 0 for basic update See Guetig Aharonov Rotter and Sompolinsky 2003 Learning input correlations through non linear asymmetric Hebbian plasticity Journal of Neuroscience 23 pp 3697 3714 muneg Extended multiplicative negative update dw W P0S x Aneg x exp Delta tauneg Set to 0 for basic update tau sec The synaptic time constant T W The weight scaling factor strenght maximal amplitude of the synaps
58. period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 38 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Dulnoise noise input current DuGnoise noise input conductance Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses n0utgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 21 CountSpikeFilter Read writable Fields time_window Length of time window Readonly Fields nChannels Number of Input channels to filter m_dt Time step length nValuesPerChannel Number of values stored for each channel in last Values 7 22 DiscretizationPreprocessor Readonly Fields nInputRows Number of rows for input vectors nOutputRows Number of rows for output vectors 7 23 dNACNeuron Uses A
59. properties of thalamocortical relay neurons J Neurophysiol 1992 Oct 68 4 1384 400 See also Alexander D Protopapas Michael Vanier and James M Bower Chapter 12 Simulating Large Networks of Neurons in Methods in Neuronal Modeling From Ions to Networks 2nd edition Christof Koch and Idan Segev eds MIT Press Cambridge Mass 1998 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel Readonly Fields g S The coductance g t Vm of the channel 7 56 PCAPreprocessor Readonly Fields nInputRows Number of rows for input vectors nOutputRows Number of rows for output vectors 7 57 Readout This class implements readouts within the simulation Readouts take the responses of the liquid and calculate a new function with these inputs Readouts can be trained to approximate some target function Readouts can be connected to physical models or feedback neurons for closed loop simulations Therefore the interface of readouts is similar to that of synapses The output of a readout can be recorded Read writable Fields enabled Flag 0 readout disabled 1 readout enabled offset Offset of output values range Range of output values Readonly Fields output Output variable of a readout 56 7 58 Recorder This class allows you to record traces of any fields of any object
60. ron Model A standard leaky integrate and fire neuron model is implemented where the membrane po tential Vm of a neuron is given by dVm Tm di Vin Vresting Rm 7 Isyn t Linject E Inoise where Tm Cm Rm is the membrane time constant Rm is the membrane resistance syn t is the current supplied by the synapses Jin ject is a non specific background current and Inoise is a Gaussion random variable with zero mean and a given variance noise At time t 0 Vm ist set to Vini If Vm exceeds the threshold voltage Vihresh it is reset to Vreset and hold there for the length Tre fract of the absolute refractory period Implementation The exponential Euler method is used for numerical integration Read writable Fields Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vthresh V If Vm exceeds Vinresn a spike is emmited Vresting V The resting membrane voltage Vreset V The voltage to reset Vm to after a spike Vinit V The initial condition for Vm at time t 0 Trefract sec The length of the absolute refractory period Inoise A The standard deviation of the noise to be added each integration time constant Tinject A A constant current to be injected into the LIF neuron type Type e g inhibitory or excitatory of the neuron Readonly Fields Isyn synaptic input current nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses Vm V The mem
61. set flag Flag which determines wheter V should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 40 HVACAChannel_Brown93 Uses HVACAuGate_Brown93 and HVACAvGate_Brown93 Reference Brown AM Schwindt PC Crill WE Voltage dependence and activation kinetics of pharmacologically defined components of the high threshold calcium current in rat neocortical neurons J Neurophysiol 1993 Oct 70 4 1530 43 See also Seamans JK Durstewitz D Christie BR Stevens CF Sejnow
62. ski TJ Dopamine D1 D5 receptor modulation of excitatory synaptic inputs to layer V prefrontal cortex neurons Proc Natl Acad Sci U S A 2001 Jan 2 98 1 301 6 Read writable Fields Ts Scale of the time constants of the gating variables T V Gbar S The maximum conductance of the channel Erev V The reversal potential Epey of the channel 48 Readonly Fields g S The coductance g t Vm of the channel 7 41 IfbNeuron see Izhikevich E Which modell to use for cortical spiking neurons Model A leaky integrate and fire or burst neuron model is implemented where the membrane poten tial Vm of a neuron is given by AV Tri di Vm Vresting Rm Isyn t F Linject F Inoise where Tm Cm Rm is the membrane time constant Rm is the membrane resistance syn t is the current supplied by the synapses linject is a non specific background current and Inoise is a Gaussion random variable with zero mean and a given variance noise At time t 0 Vm ist set to Ving If Vm exceeds the threshold voltage Vihresh it is reset to Vreset and hold there for the length Tre fract of the absolute refractory period Implementation The exponential Euler method is used for numerical integration Read writable Fields Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vthresh V If Vm exceeds Vinresn a spike is emmited Vresting V The resting membrane voltage Vreset V The voltage to
63. sn a spike is emmited Vreset V The voltage to reset Vm to after a spike doReset flag Flag which determines wheter Vm should be reseted after a spike Trefract sec Length of the absolute refractory period nummethod flag Numerical method for the solution of the differential equation Exp Euler 0 Crank Nicolson 1 type Type e g inhibitory or excitatory of the neuron Cm F The membrane capacity Cm Rm Ohm The membrane resistance Rm Vresting V The resting membrane voltage Vinit V Initial condition forV at time t 0 VmScale V Defines the difference between Vresting and the Vthresh for the calculation of the iongate tables and the ionbuffer Erev Inoise W Variance of the noise to be added each integration time constant Tinject A Constant current to be injected into the CB neuron Readonly Fields Em V The reversal potential of the leakage channel Vm V The membrane voltage Isyn synaptic input current Gsyn synaptic input conductance nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses nBuffers Number of ion buffers nChannels Number of channels 7 15 cACOUNeuron Uses AHP_Channel 31 Read writable Fields ge S exc and inh conductances noise gi S exc and inh conductances noise ge0 S exc and inh mean conductances noise gi0 S exc and inh mean conductances noise tau e S time constants and std for exc
64. ss and csim ClassInfo registerField function calls by hand So we decided to generate it automatically from the source code We use a tool named reggen a quick and dirty hack based on doxygen to gather the infomation from the source code and its doxygen style documentation reggen writes all relevant function calls to register fields an classes to registerclass i Default behaviour of c reggen e public class member variables are registered as READWRITE fields protected class member variables are registered as READONLY e private class member variables are not registered e no units are assumed i e a dimensionless field lower and upper bound are set to Inf and Inf respectively e an warning is issued if no size information is given for arrays the brief doxygen comment is used as descriptio of the field the brief doxygen comment of a class is used as its description Specifying information for use by reggen If you want to register a class you just put the macro DO_REGISTERING somewhere in the class definition For each of the member variables where you want to change the default behaviour of reggen you have to put a doxygen brief comment where you specify the relevant information Example 1 Make the float variable S a read writable field with units Volt and a lower and upper bound of 1 and 1 respectively A voltage scale factor readwrite units Volt range 1 1 float S 19 Example 2 Make the double variable
65. starts at time t 0 e R channel j fieldName name of the recorded field commonChannels 1 In this case R has two fields e R data A double array where R data j s is the s th recorded value of the j th field recorded Note that the data always starts at time t 0 Remark The number of values which are recorded depends on the field dt of the Recorder object Sec 7 58 e R info A struct array where R info j idx is the handle of the object from which the field R info j fieldName is recorded 14 3 2 2 Getting the results of a multi stimulus simulation Reminder The command gt gt R csim get rec traces returns only the results of the last simulation In the case of a multi stimulus simulation with stimulus array S of length n this command returns only the result of the simulation with stimulus S n To get the results of all n simulations one has to use gt gt R csim simulate Tsim S In that case R contains the results of all m stimulations R r s channel contains the recorded data of the r th recorder during the s th simulation 1 lt s lt n R r s channel is of the format as described in Section 3 2 1 4 Additional Topics 4 1 How the network is simulated CSIM employs a fixed time step simulation scheme for the integration of the differential equations involved represented by objects Currently all objects use the exponential Euler method with the same integration time step see Se
66. t network If there are already objects a new network will be created and and net will be imported to that NOTE The struct array net returned by net csim export does not contain information about the current state of the simulation at time t Hence it is only possible to start simulation based on such an arry at time t 0 0 6 3 3 destroy csim destroy all the networks 6 4 Displaying Getting information 6 4 1 get e csim get diplays the values of the global fields Sec 4 2 e v csim get field return the value of the global field Sec 4 2 specified by the string field 23 6 4 2 csim get idx displays the values of all fields of the object specified by the array index handle idx v csim get idx field returns a double array v where v i contains the values of the field field of the object with index handle idx i o csim get idx struct return a struct array describing the object specified by the index handle idx 1 o className String containing the name of class of object l o spiking Double value which is 1 if the object is a spike emitting object 0 otherwise o fields Cell array of strings containing the field names of the object and all fields with its values incomming outgoing csim get idx connections returns the in dices handles of the incomming i e input delivering and outgoing i e output rece
67. tate x t of the synapse is increased by W r u when a presynaptic spike hits the synapse and decays exponentially time constant 7 otherwise u and r model the current state of facilitation and depression Read writable Fields U The use parameter of the dynamic synapse D sec The time constant of the depression of the dynamic synapse F sec The time constant of the facilitation of the dynamic synapse u0 Value of the time varying facilitation state variable u for the first spike rO Value of the time varying depression state variable r for the first spike tau sec The synaptic time constant 7 W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay Readonly Fields u The time varying state variable u for facilitation r The time varying state variable u for depression psr The psr postsynaptic response is the result of whatever computation is going on in a synapse steps2cutoff 7 27 DynamicStdpSynapse See DynamicSpikingSynapse and StdpSynapse for details Read writable Fields U The use parameter of the dynamic synapse D sec The time constant of the depression of the dynamic synapse F sec The time constant of the facilitation of the dynamic synapse ud Value of the time varying facilitation state variable u for the first spike rO Value of the time varying depression state variable r for the first spike back_delay sec Delay of de
68. tau sec The synaptic time constant 7 W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay Readonly Fields psr The psr postsynaptic response is the result of whatever computation is going on in a synapse steps2cutoff 59 7 66 StaticSpikingSynapse We call a synapse a static synapse if the amplitude of each postsynaptic response is equal Here we implement a synaptic response x t of the form x t W exp t 7 for each spike which hits the synapse at time t with an amplitude of W and a decay time constant of 7 The responses of all the spikes are added up linearly Read writable Fields tau sec The synaptic time constant 7 W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay Readonly Fields psr The psr postsynaptic response is the result of whatever computation is going on in a synapse steps2cutoff 7 67 StaticStdpSynapse Read writable Fields back_delay sec Delay of dendritic backpropagating spike the synapse sees the postsynaptic spike delayed by back_delay tauspost Used for extended rule by Froemke and Dan See Froemke and Dan 2002 Spike timing dependent synaptic modification induced by natural spike trains Nature 416 3 2002 tauspre Used for extended rule by Froemke and Dan taupos Timeconstant of exponential decay of positive learning window for
69. umber of incoming synapses n0utgoing Number of outgoing synapses 7 61 SpikingInputNeuron Read writable Fields type Type e g inhibitory or excitatory of the neuron Readonly Fields nIncoming Number of incoming synapses nOutgoing Number of outgoing synapses 58 7 62 SpikingTeacher 7 63 StaticAnalogCbSynapse Read writable Fields E Equilibrium potential given by the Nernst equation Inoise The noise of our analog synapses W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay Readonly Fields psr The psr postsynaptic response is the result of whatever computation is going on in a synapse 7 64 StaticAnalogSynapse Read writable Fields Inoise The noise of our analog synapses W The weight scaling factor strenght maximal amplitude of the synapse delay sec The synaptic transmission delay Readonly Fields psr The psr postsynaptic response is the result of whatever computation is going on in a synapse 7 65 StaticSpikingCbSynapse We call a synapse a static synapse if the amplitude of each evoked conductance change is equal Here we implement a conductance g t of the form g t W exp t 7 for each spike which hits the synapse at time t with an amplitude of W and a decay time constant of 7 The responses of all the spikes are added up linearly Read writable Fields E Equilibrium potential given by the Nernst equation
70. yan and Abbott 2001 and Gerstner and Kistler 2002 Furthermore Matlab programming knowledge is assumed This manual is also available in HTML format 1 3 Features of the current version Different levels of modeling Different neuron models leaky integrate and fire neurons compartmental based neurons sigmoidal neurons Different synapse models static synapses and a certain model of dynamic synapses are available for spiking as well as for sigmoidal neurons Spike time dependent synaptic plasticity is also implemented Easy to use Matlab interface Since CSIM is incorporated into Matlab it is not necessary to learn any other script laguage to set up the simulation This is all done with Matlab scripts Furthermore the results of a simulation are directly returned as Matlab arrays and hence any plotting and analysis tools available in Matlab can easily be applied Object oriented design We adopted an object oriented design for CSIM which is similar to the approaches taken in GENESIS and NEURON That is there are objects e g a LifNeuron object implements the standard leaky integrate and fire model which are interconnected by means of some signal channels The creation of objects the connec tion of objects and the setting of parameters of the objects is controlled at the level of Matlab scipts whereas the actual simulation is done in the fast C core Fast C core Since CSIM is implemented in C and is not as general as GENESIS or NEUR

CSIM : A Neural Circuit SIMulator User Manual

Contents

Download Pdf Manuals

Related Search

Related Contents