Few-Shot Self Reminder to Overcome Catastrophic Forgetting
This addresses the problem of catastrophic forgetting for deep learning vision systems in continual learning settings, representing an incremental improvement with a simple approach.
The paper tackles catastrophic forgetting in deep neural networks during continual learning by introducing Few-shot Self Reminder (FSR), a method that uses logit matching on few-shot samples from old tasks, achieving superior knowledge retention and outperforming previous methods on benchmarks.
Deep neural networks are known to suffer the catastrophic forgetting problem, where they tend to forget the knowledge from the previous tasks when sequentially learning new tasks. Such failure hinders the application of deep learning based vision system in continual learning settings. In this work, we present a simple yet surprisingly effective way of preventing catastrophic forgetting. Our method, called Few-shot Self Reminder (FSR), regularizes the neural net from changing its learned behaviour by performing logit matching on selected samples kept in episodic memory from the old tasks. Surprisingly, this simplistic approach only requires to retrain a small amount of data in order to outperform previous methods in knowledge retention. We demonstrate the superiority of our method to the previous ones in two different continual learning settings on popular benchmarks, as well as a new continual learning problem where tasks are designed to be more dissimilar.