/signal_propagation

Computational neuroscience project: sigprop, an alternative to the backpropagation.

Primary LanguagePython

SigProp analysis

Signal propagation (SigProp) is a framework used to train a model in alternative to backpropagation. This is the code used to compare ConvSpikeNN results with the ones of original paper and differences in metrics between backpropagation and SigProp.

Prerequisite

Install on local machine:

  • pip
  • python

To install the needed dependencies run the command:

pip install -r requirements.txt

Run the analysis

First analysis uses the ConvSpikeNN and could be run launching the command:

python main.py SpikeNN

The second analysis is done using VGG model. To run it the command is:

python main.py VGG

Results of analysis are collected on json files inside the directory results.