Biologically-Inspired Continual Learning of Human Motion Sequences

TL;DR

This paper introduces a biologically-inspired generative model for continual learning of human motion sequences, achieving high accuracy in sequential task learning and generation, inspired by brain mechanisms.

Contribution

It develops a novel BI-CTVAE model that enhances continual learning of temporal sequences by incorporating a mixture-of-Gaussians latent space, advancing brain-inspired replay methods.

Findings

Final classification accuracy of 78% after sequential learning.

63% improvement over no-replay baseline.

Only 5.4% below state-of-the-art offline models.

Abstract

This work proposes a model for continual learning on tasks involving temporal sequences, specifically, human motions. It improves on a recently proposed brain-inspired replay model (BI-R) by building a biologically-inspired conditional temporal variational autoencoder (BI-CTVAE), which instantiates a latent mixture-of-Gaussians for class representation. We investigate a novel continual-learning-to-generate (CL2Gen) scenario where the model generates motion sequences of different classes. The generative accuracy of the model is tested over a set of tasks. The final classification accuracy of BI-CTVAE on a human motion dataset after sequentially learning all action classes is 78%, which is 63% higher than using no-replay, and only 5.4% lower than a state-of-the-art offline trained GRU model.