A modularity comparison of Long Short-Term Memory and Morphognosis neural networks

TL;DR

This paper compares the modularity performance of LSTM and Morphognosis neural networks using maze tasks, finding Morphognosis significantly outperforms LSTM in generalization to new maze configurations.

Contribution

It introduces a comparative analysis of LSTM and Morphognosis architectures on modular maze tasks, highlighting Morphognosis's superior generalization capabilities.

Findings

Morphognosis outperforms LSTM in maze generalization

Both networks perform well during training

Morphognosis demonstrates better modular task handling

Abstract

This study compares the modularity performance of two artificial neural network architectures: a Long Short-Term Memory (LSTM) recurrent network, and Morphognosis, a neural network based on a hierarchy of spatial and temporal contexts. Mazes are used to measure performance, defined as the ability to utilize independently learned mazes to solve mazes composed of them. A maze is a sequence of rooms connected by doors. The modular task is implemented as follows: at the beginning of the maze, an initial door choice forms a context that must be retained until the end of an intervening maze, where the same door must be chosen again to reach the goal. For testing, the door-association mazes and separately trained intervening mazes are presented together for the first time. While both neural networks perform well during training, the testing performance of Morphognosis is significantly better than LSTM on this modular task.