Measuring Catastrophic Forgetting in Neural Networks

By Ronald Kemker, Angelina Abitino, Marc McClure, Christopher Kanan. 2017

889 citations recorded by the source corpus.

Deep neural networks are used in many state-of-the-art systems for machine perception. Once a network is trained to do a specific task, e.g., bird classification, it cannot easily be trained to do new tasks, e.g., incrementally learning to recognize additional bird species or learning an entirely different task such as flower recognition. When new tasks are added, typical deep neural networks are prone to catastrophically forgetting previous tasks. Networks that are capable of assimilating new information incrementally, much like how humans form new memories over time, will be more efficient than re-training the model from scratch each time a new task needs to be learned. There have been multiple attempts to develop schemes that mitigate catastrophic forgetting, but these methods have not been directly compared, the tests used to evaluate them vary considerably, and these methods have only been evaluated on small-scale problems (e.g., MNIST). In this paper, we introduce new metrics and benchmarks for directly comparing five different mechanisms designed to mitigate catastrophic forgetting in neural networks: regularization, ensembling, rehearsal, dual-memory, and sparse-coding. Our experiments on real-world images and sounds show that the mechanism(s) that are critical for optimal performance vary based on the incremental training paradigm and type of data being used, but they all demonstrate that the catastrophic forgetting problem is not yet solved.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Optimization & Training

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Optimization & Training

Citations in the corpus

The Natural Language Decathlon: Multitask Learning as Question Answering (2018, background): “Much effort has gone into mitigating catastrophic forgetting (McCloskey and Cohen, 1989; Ratcliff, 1990; Kemker et al., 2017) by penalizing the norm of parameters when training on a new task (Kirkpatrick et al.”
Webly Supervised Joint Embedding for Cross-Modal Image-Text Retrieval (2018, methodology): “We utilize a smaller learning rate in Stage II, as the network achieves competitive performance after Stage I and tuning the embedding network with a high learning rate from weakly-annotated data may lead to catastrophic forgetting [25].”
RuleBERT: Teaching Soft Rules to Pre-Trained Language Models (2021, background): “, it started to unlearn what it had learned at pre-fine-tuning (Kirkpatrick et al., 2017; Kemker et al., 2018; Biesialska et al., 2020).”
Re-evaluating Continual Learning Scenarios: A Categorization and Case for Strong Baselines (2018, background): “) and use one of them in one task [17, 18, 19, 20].”
Continuous Learning in Single-Incremental-Task Scenarios (2018, methodology): “Here we propose, in line with [12] and [33], a three-way fuzzy categorization of the most common CL strategies: ar X iv :1 80 6.”
Lifelong Learning of Spatiotemporal Representations With Dual-Memory Recurrent Self-Organization (2018, background): “…to address supervised learning on image datasets of very limited complexity such as MNIST (LeCun et al., 1998) and CIFAR-10 (Krizhevsky, 2009) while not scaling up to incremental learning tasks with larger-scale datasets of natural images and videos (Kemker et al., 2018; Parisi et al., 2018a).”
FearNet: Brain-Inspired Model for Incremental Learning (2017, background): “FEL requires a large number of units to work well (Kemker et al., 2018).”
Memory Efficient Experience Replay for Streaming Learning (2018, methodology): “For high-dimensional datasets, they all typically perform worse than ofﬂine DNNs.”
FeTrIL: Feature Translation for Exemplar-Free Class-Incremental Learning (2022, background): “The main problem faced by class-incremental learning (CIL) methods is catastrophic forgetting [14, 25], the tendency of neural nets to underﬁt past classes when ingesting new data.”
CODA-Prompt: COntinual Decomposed Attention-Based Prompting for Rehearsal-Free Continual Learning (2022, background): “Rehearsal with stored data [3–5, 7, 9, 10, 18, 19, 24, 29, 42, 51–53, 64, 68] or samples from a generative model [27, 28,48,56,60] is highly effective when storing training data or training/saving a generative model is possible.”
One-Pass Learning via Bridging Orthogonal Gradient Descent and Recursive Least-Squares (2022, neutral): “forget” past information while learning new data [9].”
Balancing Stability and Plasticity through Advanced Null Space in Continual Learning (2022, background): “The problem, referred to as catastrophic forgetting, is a key challenge in continual learning [16,21,30,34,38,48].”
Few-Shot and Continual Learning with Attentive Independent Mechanisms (2021, background): “Current deep learning methods are inefficient in remembering old concepts after being fed with new concepts, also widely know as catastrophic forgetting [34, 23].”
Always Be Dreaming: A New Approach for Data-Free Class-Incremental Learning (2021, methodology): “Experience replay with stored data [2, 3, 7, 8, 13, 17, 22, 27, 38, 44, 45, 46, 57] or samples from a generative model [25, 26, 43, 49, 54] is highly effective when storing training data or training/saving a generative model is possible, unlike the DFCIL setting.”
Continual Learning with Fully Probabilistic Models (2021, background): “Systematic comparisons between different approaches to avoid CF are performed in, e.g., [5, 6].”
Replay in Deep Learning: Current Approaches and Missing Biological Elements (2021, neutral): “Similar to partial replay methods, generative replay methods can generate veridical inputs (Shin et al., 2017; Kemker & Kanan, 2018; Parisi, Tani, Weber, and Wermter, 2018; He, Wang, Shan, & Chen, 2018; Wu et al., 2018; Ostapenko et al., 2019; Abati et al., 2020; Liu, Su, Liu, Schiele, & Sun, 2020;…”
Self-Supervised Training Enhances Online Continual Learning (2021, background): “We study the common continual learning paradigm in which pre-training precedes continual learning [28, 29, 55, 10, 71, 32, 33, 34, 68, 4].”
Continual Semantic Segmentation via Repulsion-Attraction of Sparse and Disentangled Latent Representations (2021, background): “Deep learning models are prone to catastrophic forgetting [20, 30, 48], i.”
Online Continual Learning in Image Classification: An Empirical Survey (2021, background): “Another two empirical studies on the performance of CL include [29, 30], but only a small number of CL methods are compared.”