Remembering for the Right Reasons: Explanations Reduce Catastrophic Forgetting

Abstract

The goal of continual learning (CL) is to learn a sequence of tasks without suffering from the phenomenon of catastrophic forgetting. Previous work has shown that leveraging memory in the form of a replay buffer can reduce performance degradation on prior tasks. We hypothesize that forgetting can be further reduced when the model is encouraged to remember the evidence for previously made decisions. As a first step towards exploring this hypothesis, we propose a simple novel training paradigm, called Remembering for the Right Reasons (RRR), that additionally stores visual model explanations for each example in the buffer and ensures the model has “the right reasons” for its predictions by encouraging its explanations to remain consistent with those used to make decisions at training time. Without this constraint, there is a drift in explanations and increase in forgetting as conventional continual learning algorithms learn new tasks. We demonstrate how RRR can be easily added to any memory or regularization-based approach and results in reduced forgetting, and more importantly, improved model explanations. We have evaluated our approach in the standard and few-shot settings and observed a consistent improvement across various CL approaches using different architectures and techniques to generate model explanations and demonstrated our approach showing a promising connection between explainability and continual learning. Our code is available at \url{https://github.com/SaynaEbrahimi/Remembering-for-the-Right-Reasons}

Memory Replay For Continual Learning With Spiking Neural Networks

“…In this way, we could choose the smallest number of samples that contain the most influential information about each task. Several works in the literature propose methods for mitigating catastrophic forgetting in artificial neural networks that exploit explainable AI [32,33,34], and they manage to achieve comparable and, in some cases, higher performances than other state-of-the-art approaches.…”

Sparse Coding in a Dual Memory System for Lifelong Learning

“…The base method, Experience Replay (ER) (Riemer et al 2018) interleaves the training of the current task with the memory samples to train the model on the approximate joint distribution of tasks. Several studies focus on the different aspects of rehearsal: memory sample selection (Isele and Cosgun 2018), sample retrieval from memory (Aljundi et al 2019a), and what information to extract and replay (Ebrahimi et al 2020;Bhat, Zonooz, and Arani 2022).…”

Memory Population in Continual Learning via Outlier Elimination

“…Memory-based methods address catastrophic forgetting by incorporating data from previous tasks into the training process for the current task [14,7]. These approaches can use raw samples [38,10] CIFAR-100 CORe50 The dotted lines represent the Reservoir method.…”