Code for the network of rate units used in the article 'Dynamics of feature categorization' by Daniel Martí and John Rinzel in Neural Computation, January 2013, vol. 25, No. 1, pp 1-45. The repository contains tools to stimulate the network with different types of spatio-temporal input patterns.

Just get the code and compile the code with either make or scons in the ring_network folder:

git clone https://github.com/djmarti/ring_network
cd ring_network
make  # or scons

This will create a file called simulate_trajectory.

The code depends on the GNU Scientific Library. To install the library in a Debian-like box, run

apt-get install libgsl0ldbl libgsl0ldbl-dev

To install it on a Mac, run

brew install gsl

You can also install it by hand, downloading the source from the official page.

The main executable is simulate_trajectory. It simulates the time evolution of the activity of the ring network, and stores the resulting activity in several datafiles, one per dynamical variable (firing rate and, if present, adaptation, facilitation, or depression). These activities are stored both in text format and in bitmap format.

A typical run of the executable would be

./simulate_trajectory -c config_pulse_sequences -x i_single_pulse

where the file config_pulse_sequences is a configuration file specifying the network and synaptic parameters (see below for details), and i_single_pulse is an example of input descriptor, a file that describes the input pattern fed into the network (see also below). You can find examples of configuration files under the configurations folder, and examples of input descriptors in the example_input_patterns folder.

A standard config file looks like the following:

# Output files
# ------------
extinputs_file = "inputs"   # Default filename for of input descriptor
output_rate_file = "rates"  # Prefix for the firing rate filename
output_adaptation_file = "adaptation"      # and so on
output_facilitation_file = "facilitation"
normalize = 0               # Normalize rates to the maximum rate achieved during the simulation?

# Simulation parameters
# ---------------------
dt = 1e-2  # Time step for Euler method
T = 10.0   # Total simulated time (time units)

# Parameters sigmoid
# ------------------
N = 512         # Number of neurons
N_layers = 1    # Number of layers
max_rate = 1.0  # Maximum rate
beta = 3.0      # Temperature parameter
sigm_thr = 1.0  # Bias / soft threshold of sigmoid

# Negative feedback parameters
# ----------------------------
J_A = 0.0        # Adaptation amplitude (0: turned off)
tau_A = 10       # Adaptation timescale
facilitation = 0 # Facilitation amplitude (0: turned off)
depression = 0   # Depression amplitude (0: turned off)


# Connectivity parameters
# -----------------------
# Recurrent connections
J_rec_E = 3.5    # Coefficient excitatory profile
J_rec_I = 3.5    # Coefficient inhibitory profile
m_rec_E = 100    # Concentration parameter exc. profile
m_rec_I = 1      # Concentration parameter inh. profile

# Feedforward connections (irrelevant if N_layers = 1)
J_ff_E = 5.0     # Coefficient excitatory profile
J_ff_I = 0.2     # Coefficient inhibitory profile
m_ff_E = 10      # Concentration parameter exc. profile
m_ff_I = 1       # Concentration parameter inh. profile

The example should be self-explanatory.

This type of file describes how the network is stimulated. A self-descriptive example of an input descriptor is the following:

#
# Configuration of external inputs
#
# Each line specifies the characteristics of a particular pulse of stimulation,
# and is structured in 5 different fields separated by spaces. The order of the
# fields is:
#     1. time onset of the pulse,
#     2. duration of the pulse
#     3. angle (in degrees) where the pulse is applied,
#     4. intensity of the pulse,
#     5. concentration of the pulse (in a von Mises distribution)
#
# time onset    duration      theta     Imax   concentration
# -----------------------------------------------------------
1.0              1.0        -45.0      1.0       10.0
4.0              0.5         10.5      0.2       10.0

Writing these files by hand become cumbersome when there are more than a few pulses. The directory utils/ contains a python script called generator_inputs.py that generates input descriptions for random input patterns. One simple example of such random pattern would be "consecutive random pulses of a given duration, concentrated around a location, with a given concentration parameter". Other, more complex patterns are also possible. Feel free to modify the script to your needs.

This program 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 of the License, or (at your option) any later version.

This program 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.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.