pub2012memorynets

This is the simulation code which I used to generate data that Richard Naud plotted in Figure 17.11 of the spiking Hopfield network in Gerstner, W., Kistler, W.M., Naud, R., and Paninski, L. (2014). Neuronal dynamics: from single neurons to networks and models of cognition (Cambridge: Cambridge University Press).

Files

It includes

• This file README.md
• The simulation files sim_hopfield.cpp and sim_hopfield2.cpp
• A Makefile to build them
• A run.sh file which illstrates the usage of the simulation
• A simple script to visualize output plot_activity.gnu
• The patterns I used in the book in the directory ./stim
• Some additional files history.txt and params_sim2.txt with some parameter values that worked for me to make the original figure in the book

Compile

The version I am publishing here has been adapted to compile with Auryn v0.8, but you find older revisions of the code in the commit history.

Either way you will need Auryn (fzenke.net/auryn) installed for this to run. You can build the simulation code with make after having updated the Makefile you point to your Auryn library.

Example

Here is an example

fzenke@grumpy:~/archive/2012-book-wulfram/spiking_hopfield/src$ make && mpirun -n 2 ./sim_hopfield2 --load ../stim/wulfp3.pat --chi 0.3 --gamma 0.2  --wei1 0.15 --wei2 0.1 --simtime 30
make: Nothing to be done for `all'.
load from matrix ./stim/wulfp3.pat.
chi from matrix 0.3.
gamma factor 0.2.
wei1 factor 0.15.
wei2 factor 0.1.
simtime 30.
load from matrix ./stim/wulfp3.pat.
chi from matrix 0.3.
gamma factor 0.2.
wei1 factor 0.15.
wei2 factor 0.1.
simtime 30.
( 0) Setting up neuron groups ...
( 0) Setting up external connections ...
( 0) Setting up EE connections ...
( 0) Setting up EI connections ...
( 0) Setting up IE connections ...
(!!) on rank 1: I1E (Connection.cpp): Some elements of pattern 0 are larger than the underlying NeuronGroups!
(!!) on rank 0: I1E (Connection.cpp): Some elements of pattern 0 are larger than the underlying NeuronGroups!
( 0) Setting up II connections ...
( 0) Setting up stimulus ...
( 0) Setting up monitors ...
( 0) Simulating ...
[=========================] 100%      t=30.0s  f=10.9 Hz in NeuronGroup   
( 0) Freeing ...

which will produce a bunch of output files. With `plot_activity.gnu you see how this indeed switches between different patterns.

fzenke@grumpy:~/archive/2012-book-wulfram/spiking_hopfield/src$ gnuplot 

	G N U P L O T
	Version 5.0 patchlevel 1    last modified 2015-06-07 

	Copyright (C) 1986-1993, 1998, 2004, 2007-2015
	Thomas Williams, Colin Kelley and many others

	gnuplot home:     http://www.gnuplot.info
	faq, bugs, etc:   type "help FAQ"
	immediate help:   type "help"  (plot window: hit 'h')

Terminal type set to 'wxt'
gnuplot> load 'plot_activity.gnu'

The output should look somewhat like this.

You can see that there is a pattern switch happening in those 30s. Firing rates become "less crazy" with paramter tuning and as you load more patterns (there are files with different numbers of random patterns in ./stim) which unfortunately always requires retuning.

Tuning parameters

run.sh gives you an idea of how to run the simulation code.

Moreover, ./sim_hopfield2 --help will show you which paramters are exposed via CLI commands.

And history.txt as well ass params_sim2.txt includes some parameter with which the code worked. However, There were some changes to Auryn over the years.One included a bug which made the connecting probability slightly smaller that was fixed. By using a newer version of Auryn where these problems are fixed will thus also necessitate to retuned the critical synaptic strength parameters. However, the ones in history.txt and params_sim2.txt might be good starting points.

fzenke, Aug 02, 2016