Prediction error-driven memory consolidation for continual learning. On the case of adaptive greenhouse models

TL;DR

This paper introduces an adaptive continual learning architecture that uses prediction-error driven memory consolidation to mitigate catastrophic forgetting, demonstrated through greenhouse modeling in horticulture.

Contribution

It proposes a novel memory consolidation method based on prediction error, applied to online learning in greenhouse models for horticulture industry.

Findings

Effective mitigation of catastrophic forgetting in greenhouse models

Successful transfer of models from research to production environments

Demonstrated adaptability in real-world horticultural settings

Abstract

This work presents an adaptive architecture that performs online learning and faces catastrophic forgetting issues by means of episodic memories and prediction-error driven memory consolidation. In line with evidences from the cognitive science and neuroscience, memories are retained depending on their congruency with the prior knowledge stored in the system. This is estimated in terms of prediction error resulting from a generative model. Moreover, this AI system is transferred onto an innovative application in the horticulture industry: the learning and transfer of greenhouse models. This work presents a model trained on data recorded from research facilities and transferred to a production greenhouse.