Adaptive Multi-Agent Continuous Learning System

TL;DR

This paper introduces an adaptive multi-agent system for continuous, self-supervised learning and clustering of sequential data, demonstrating effectiveness in dynamic environments through video behavior experiments.

Contribution

It presents a novel multi-agent clustering system with self-supervision and adaptability for continuous learning in dynamic environments.

Findings

Successful clustering of video behaviors in dynamic settings

Demonstrated adaptability to environmental changes

Feasibility confirmed through experiments

Abstract

We propose an adaptive multi-agent clustering recognition system that can be self-supervised driven, based on a temporal sequences continuous learning mechanism with adaptability. The system is designed to use some different functional agents to build up a connection structure to improve adaptability to cope with environmental diverse demands, by predicting the input of the agent to drive the agent to achieve the act of clustering recognition of sequences using the traditional algorithmic approach. Finally, the feasibility experiments of video behavior clustering demonstrate the feasibility of the system to cope with dynamic situations. Our work is placed here\footnote{https://github.com/qian-git/MAMMALS}.