/pnet

Point-Neuronal-Network simulation framework for academic use.

Primary LanguageC++

Point-Neuronal-Network Simulation Frameworks

Point neuronal network simulator framework for academic use, initiating from my bachleor thesis.12

Current supported neuronal models

  • Conductance based leaky integrate-and-fire model (LIF_G)
  • Conductance based leaky integrate-and-fire model with $1^{st}$ order smoothness in conductance (LIF_GH)
  • Current based leaky integrate-and-fire model (LIF_I)

Contents

pnet
├── doc       docs and *.ini like configuration files;
├── include   C++ header files;
├── src       C++ source files;
├── external  external dependence;
├── utils     python interface;
├── pys       python scripts;
└── test      test files;

Get Started

Installation

  1. Build external cnpy library

    mkdir build
cd build
cmake ../external/cnpy -DCMAKE_INSTALL_PREFIX=../
make
make install

  2. Build core pnet simulator

    cd ..
make -j

How to use

Runing core simulation program:

$ bin/neu_sim --prefix <path-of-config-file> -c <config filename>

Getting more config info

$ bin/neu_sim -h
All Options:
  -h [ --help ]             produce help message
  -p [ --prefix ] arg (=./) prefix of output files
  -c [ --config ] arg       config file, relative to prefix
  -v [ --verbose ]          show output

Configs:
  --network.ne arg (=1)          number of Exc. neurons
  --network.ni arg (=0)          number of Inh. neurons
  --network.simulator arg (=SSC) One of Simple, SSC, SSC_Sparse.
  --neuron.model arg (=LIF_GH)   One of LIF_I, LIF_G, LIF_GH.
  --neuron.tref arg (=2)         (ms) refractory period
  --synapse.file arg             file of synaptic strength
  --space.file arg               file of spatial location of neurons
  --driving.file arg             file of Poisson settings
  --driving.seed arg             seed to generate Poisson point process
  --time.t arg (=1000)           (ms) total simulation time
  --time.dt arg (=0.03125)       (ms) simulation time step
  --time.stp arg (=0.5)          (ms) sampling time step
  --output.poi arg (=0)          output flag of Poisson Drive
  --output.v arg (=0)            output flag of V to V.bin
  --output.i arg (=0)            output flag of I to I.bin
  --output.ge arg (=0)           output flag of GE to GE.bin
  --output.gi arg (=0)           output flag of GI to GI.bin

Python interface

Python interface for configuration of network simulation, including type of neuronal model, type of network simulator, network structure, network inputs, spatial structure of 2-D network.

>>> # simulate an all connected 100 LIF_GH neurons network
>>> from utils.pnet import *
>>> net = network(80, 20)
>>> mat = np.zeros((100,100), dtype=int)
>>> np.fill_diagonal(mat, 0)
>>> dmat = np.zeros((100,100))
>>> pm = {'model': 'LIF_GH', 'simulator' : 'SSC', 'tref': 2.0,
...:  'see' : 1e-3, 'sie' : 1e-3, 'sei' : 5e-3, 'sii' : 5e-3,
...:  'synapse_file': 'smat.npy',
...:  'con_mat' : mat,
...:  'space_file': 'dmat.npy',
...:  'delay_mat' : dmat,
...:  'pre_e' : 1.5, 'pse_e' : 5e-3,
...:  'pre_i' : 1.5, 'pse_i' : 5e-3,
...:  'poisson_file': 'PoissonSetting.csv', 'poisson_seed': 3,
...:  'T': 1e3, 'dt': 0.03125, 'stp': 0.5,
...:  'v_flag': True}
>>> # add new simulation configs
>>> net.add(**pm)
>>> # show network configs
>>> net.show()
========================================
Neuron Population:
        ne        ni        ke        ki
        80        20      99.0      99.0
----------------------------------------
Synapses:
       see       sie       sei       sii
  1.00e-03  1.00e-03  5.00e-03  5.00e-03
----------------------------------------
FFWD Poisson:
       pre       pse       pri       psi
  1.50e+00  5.00e-03  0.00e+00  0.00e+00
----------------------------------------
Afferent Connection:
   99   99   99   99   99   99   99   99
========================================
>>> # update configuration files
>>> net.updatefiles()
>>> # run simulation
>>> net.run('verbose')
(number of connections in sparse-mat 9900)
>> Initialization :     0.007 s
>> Done!
>> Simulation :         0.341 s
Total inter-neuronal interaction : 112563
Mean firing rate : 11.37 Hz

Reference

Footnotes

  1. Shelley, M. J., & Tao, L. (2001). Efficient and accurate time-stepping schemes for integrate-and-fire neuronal networks. Journal of Computational Neuroscience, 11(2), 111-119. ↩

  2. Rangan, A. V., & Cai, D. (2007). Fast numerical methods for simulating large-scale integrate-and-fire neuronal networks. Journal of computational neuroscience, 22(1), 81-100. ↩