Exploring Flip Flop memories and beyond: training recurrent neural networks with key insights

TL;DR

This paper investigates the training and analysis of recurrent neural networks for a 3-bit Flip Flop memory task, providing detailed modeling, visualization, and a versatile codebase for various temporal processing tasks.

Contribution

It offers a comprehensive methodology for implementing and analyzing RNNs on a Flip Flop memory task, including novel visualization of memory states in reduced-dimensional space.

Findings

Networks successfully learned the Flip Flop memory task

Memory states can be stored in cube vertices in reduced space

The approach is adaptable to diverse temporal tasks

Abstract

Training neural networks to perform different tasks is relevant across various disciplines. In particular, Recurrent Neural Networks (RNNs) are of great interest in Computational Neuroscience. Open-source frameworks dedicated to Machine Learning, such as Tensorflow and Keras have produced significant changes in the development of technologies that we currently use. This work aims to make a significant contribution by comprehensively investigating and describing the implementation of a temporal processing task, specifically a 3-bit Flip Flop memory. We delve into the entire modelling process, encompassing equations, task parametrization, and software development. The obtained networks are meticulously analyzed to elucidate dynamics, aided by an array of visualization and analysis tools. Moreover, the provided code is versatile enough to facilitate the modelling of diverse tasks and systems. Furthermore, we present how memory states can be efficiently stored in the vertices of a cube in the dimensionally reduced space, supplementing previous results with a distinct approach.