Lifelong Learning from Event-based Data

TL;DR

This paper explores lifelong learning from event camera data, proposing a combined approach with feature extraction and habituation to reduce forgetting during incremental learning in dynamic environments.

Contribution

It introduces a novel habituation-based method and demonstrates how combining techniques can mitigate catastrophic forgetting in event-based data learning.

Findings

Habituation-based method effectively reduces forgetting.

Combining multiple techniques improves incremental learning performance.

Model successfully learns from event camera data without catastrophic forgetting.

Abstract

Lifelong learning is a long-standing aim for artificial agents that act in dynamic environments, in which an agent needs to accumulate knowledge incrementally without forgetting previously learned representations. We investigate methods for learning from data produced by event cameras and compare techniques to mitigate forgetting while learning incrementally. We propose a model that is composed of both, feature extraction and continuous learning. Furthermore, we introduce a habituation-based method to mitigate forgetting. Our experimental results show that the combination of different techniques can help to avoid catastrophic forgetting while learning incrementally from the features provided by the extraction module.