mike2018

This repository contains the implementation of the experiments proposed in the paper Using stigmergy to incorporate the time into artificial neural networks.
If you are interested on the actual implementation of the Stigmergic Neural Networks please check out the torchsnn repository

Installation

Clone this repository

git clone https://github.com/galatolofederico/mike2018 && cd mike2018

Create a python virtualenv and activate it, make sure to use python3

virtualenv --python=/usr/bin/python3 env && source ./env/bin/activate

Install the requirements

pip install -r requirements.txt

You are ready to go!

Contents

mnist/mnist.py

Python script to train and evaluate all the architectures described in the paper. It uses the sacred framework to manage experiments configurations and results.
It uses the sacred-style to set the configuration variables

python3 mnist.py with config1=val1 config2=val2

For example

python3 mnist.py with batch_size=20 use_mongo=True

You can set the following configuration variables

Variable	Description	Default
arch	Architecture to use (possible values: 'stigmergic', 'feedforward', 'recurrent', 'lstm')	stigmergic
n_hidden	Number of hidden neurons	10
n_layers	Number of hidden layers (valid only for feedforward and lstm)	1 for feedforward and 3 for lstm
avg_window	Moving average window size for logging	100
use_mongo	Use MongoDB Observer to log the experiments	False

mnist/networks/*.py

Python implementation of all the architectures described in the paper

xor.py

Train and evaluation of the xor problem using only one stigmergic perceptron

Citing

If you want to cite us please use this BibTeX

@InProceedings{galatolo_snn,
    author="Galatolo, Federico A.
    and Cimino, Mario Giovanni C. A.
    and Vaglini, Gigliola",
    editor="Groza, Adrian
    and Prasath, Rajendra",
    title="Using Stigmergy to Incorporate the Time into Artificial Neural Networks",
    booktitle="Mining Intelligence and Knowledge Exploration",
    year="2018",
    publisher="Springer International Publishing",
    address="Cham",
    pages="248--258",
    abstract="A current research trend in neurocomputing involves the design of novel artificial neural networks incorporating the concept of time into their operating model. In this paper, a novel architecture that employs stigmergy is proposed. Computational stigmergy is used to dynamically increase (or decrease) the strength of a connection, or the activation level, of an artificial neuron when stimulated (or released). This study lays down a basic framework for the derivation of a stigmergic NN with a related training algorithm. To show its potential, some pilot experiments have been reported. The XOR problem is solved by using only one single stigmergic neuron with one input and one output. A static NN, a stigmergic NN, a recurrent NN and a long short-term memory NN have been trained to solve the MNIST digits recognition benchmark.",
    isbn="978-3-030-05918-7"
}

Contributing

This code is released under GNU/GPLv3 so feel free to fork it and submit your changes, every PR helps.
If you need help using it or for any question please reach me at federico.galatolo@ing.unipi.it or on Telegram @galatolo