A computational model of systems memory consolidation and reconsolidation (Helfer & Shultz 2019)

 Download zip file 
Help downloading and running models
A neural-network framework for modeling systems memory consolidation and reconsolidation.
1 . Helfer P, Shultz TR (2019) A computational model of systems memory consolidation and reconsolidation. Hippocampus [PubMed]
Citations  Citation Browser
Model Information (Click on a link to find other models with that property)
Model Type: Connectionist Network; Synapse;
Brain Region(s)/Organism: Hippocampus; Neocortex;
Cell Type(s): Abstract integrate-and-fire neuron;
Gap Junctions:
Receptor(s): AMPA;
Transmitter(s): Glutamate;
Simulation Environment: C or C++ program;
Model Concept(s): Memory; Synaptic Plasticity;
Implementer(s): Helfer, Peter [peter.helfer at mail.mcgill.ca];
Search NeuronDB for information about:  AMPA; Glutamate;
NS (Neural Simulator) - a neural-network framework

Helfer, P. & Shultz, T.R. (2019)
A Computational Model of Systems Memory Consolidation and Reconsolidation
Submitted for publication

Preprint: https://arxiv.org/ftp/arxiv/papers/1703/1703.01357.pdf

This implementation by Peter Helfer to whom questions should be addressed.

This directory tree contains three subdirectories:

ns :     software for running the Neural Simulator
lib:     a C++ library of utility functions used by ns
include: C++ include file used by lib and ns

The 'ns' directory contains:

- Source code for 'ns', a C++ program that implements a neural network
  designed to model memory consolidation and reconsolidation; ns --help shows
  available options 

- props, a subdirectory containing property files that define different
  test cases for ns

- multi_ns, a python script that repeatedly invokes ns to run all the test cases
  defined in the props directory, or - if specified - one or more selected test
  cases, then plots the results;

- mat.cc, a C++ program that applies any of several matrix operations to sets
  of matrices given as text input files. Used by multi_ns; mat --help shows
  available options

- ns.col, specification of colors to use when plotting results

- plot, a wrapper for gnuplot; plot --help shows available options

- nsplot, a wrapper for plot; nsplot --help shows available options


To build and run the software, you will need a GNU/Linux system with the GNU
Make system, a C++ compiler, the Python interpreter and GNUPlot available.

On a Debian Linux system, the required utilities can be installed like this:

$ sudo apt-get install build-essential
$ sudo apt-get install gnuplot

Building ns

cd into the lib directory and type 'make'.
cd into the ns directory and type 'make'.

Running ns, an example invocation:

$ ./multi_ns -p ns_01 -v 10

   This will execute 10 simulations as specified in the properties file(s)
   matching props/ns_01*.props and display the results (mean recall scores
   with standard-deviation ribbons) in a GnuPlot window. The three plots
   indicate recall performance with intact network (intact_score), with the
   HPC region inactivated during each recall test (hpc_score), and with the
   ACC region inactivated during each recall test (acc_score).

The command "./multi_ns --help" will show other available options.

Output files

Unless you specify a directory using the -d option, multi_ns will create an
output directory using the current date and time as name,
e.g. out/2019_09_28__19_51_26. A subdirectory of the output directory is
then created for each test case, e.g. 'ns_01_react' for the example
above. Each execution of ns produces a raw output file with a name like
0.raw, 1.raw, etc. From these, multi_ns extracts columns of output data into
*.out files and further into *.hpc and *.acc files. These are then processed
to calulate means (avg.out), standard deviations (stdevs.out), and standard
errors (sterr.out). Finally, the results are combined in stats.out, which is
used as input to the 'nsplot' script.

To plot the results of a previous invocation of multi_ns, use the -d option
to specify an existing output directory, for example:

$ ./multi_ns -d out/2019_09_12__20_50_06/ -p ns_01

Properties file format

A .props configuration file specifies parameters for a simulation.

Comments: Any text following a # character is a comment.

include: file_name    # read lines from the named file.

name: value           # specify a value for a property 

Most poperties are self-explanatory and/or described in the paper.

Properties that specify times for events like reactivation, PSI infusion or HPC lesion
accept values of the form dd[::hh], so that 2::6 means 2 days and 6 hours, and 13 means
13 days.

Some of the event properties accept multiple values. These are specified as
a list of values within curly braces. For example,

accFreezeTimes:         { 30::00   31::00 }

means "Freeze (inactivate) the ACC region at 30 days zero hours, and reactivate it
at 31 days zero hours".

The timeStepChanges property allows the simulation time step to be modified
during the simulation, e.g. to use a smaller step size during the first few
days after reactivation, when things change more rapidly (see ns_01_react.props
for an example). Each time step change is specified as a time and a step size.
For example,

timeStepChanges::       { 0::00 1 1::00 24 30::00 1 31::00 24 }

means "At zero days zero hours, set the time step to 1 hour, then at 1 day zero hours, 
set the time step to 24 hours, etc."

Property values specified in the .props file may be overridden by name=value
pairs given as command line arguments given to multi_ns. For example, in 

$ ./multi_ns -p ns_01 10  maxInhibition=15

the value 15 overrides whatever value is specified for maxInhibition in the .props


'columns' is a utility that selects columns from a text file and prints them
          in a specified order. For example, the command
          './columns -f myfile.text t P A_I PP' will extract the columns with
          headers t, P, A_I and PP from myfile.txt.

'mat'     is a utility that applies operations to matrices given as input files.
	  All input files must contain the same number of rows and columns. Type
	  './mat --help' to see supported operations and options.