RATT: Recurrent Attention to Transient Tasks for Continual Image Captioning
This addresses the problem of continual learning for recurrent models in image captioning, which is incremental as it adapts existing approaches to a new domain.
The paper tackles catastrophic forgetting in continual learning for LSTM-based image captioning by proposing RATT, an attention-based method that accommodates transient vocabularies, and shows it can sequentially learn five tasks without forgetting on new benchmarks using MS-COCO and Flickr30 datasets.
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.