RATT: Recurrent Attention to Transient Tasks for Continual Image Captioning

TL;DR

This paper introduces RATT, a novel recurrent attention method for continual image captioning that effectively mitigates forgetting across multiple tasks using LSTM models.

Contribution

It proposes RATT, an attention-based approach tailored for recurrent models in continual learning, addressing transient vocabularies and task overlap.

Findings

RATT successfully learns five captioning tasks without forgetting.

The approach outperforms existing continual learning methods on new benchmarks.

It adapts weight regularization and knowledge distillation for recurrent models.

Abstract

Research on continual learning has led to a variety of approaches to mitigating catastrophic forgetting in feed-forward classification networks. Until now surprisingly little attention has been focused on continual learning of recurrent models applied to problems like image captioning. In this paper we take a systematic look at continual learning of LSTM-based models for image captioning. We propose an attention-based approach that explicitly accommodates the transient nature of vocabularies in continual image captioning tasks -- i.e. that task vocabularies are not disjoint. We call our method Recurrent Attention to Transient Tasks (RATT), and also show how to adapt continual learning approaches based on weight egularization and knowledge distillation to recurrent continual learning problems. We apply our approaches to incremental image captioning problem on two new continual learning benchmarks we define using the MS-COCO and Flickr30 datasets. Our results demonstrate that RATT is able to sequentially learn five captioning tasks while incurring no forgetting of previously learned ones.