Learning Fast, Learning Slow: A General Continual Learning Method based on Complementary Learning System

By Elahe Arani, Fahad Sarfraz, Bahram Zonooz. 2022

177 citations recorded by the source corpus.

Humans excel at continually learning from an ever-changing environment whereas it remains a challenge for deep neural networks which exhibit catastrophic forgetting. The complementary learning system (CLS) theory suggests that the interplay between rapid instance-based learning and slow structured learning in the brain is crucial for accumulating and retaining knowledge. Here, we propose CLS-ER, a novel dual memory experience replay (ER) method which maintains short-term and long-term semantic memories that interact with the episodic memory. Our method employs an effective replay mechanism whereby new knowledge is acquired while aligning the decision boundaries with the semantic memories. CLS-ER does not utilize the task boundaries or make any assumption about the distribution of the data which makes it versatile and suited for"general continual learning". Our approach achieves state-of-the-art performance on standard benchmarks as well as more realistic general continual learning settings.

Subfields: Artificial Intelligence, Cognitive Neuroscience, Experimental and Cognitive Psychology, Meta & Continual Learning

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 Cognitive Neuroscience

In Experimental and Cognitive Psychology

In Meta & Continual Learning

Citations in the corpus

A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, methodology): “Feature Space LUCIR [164], PODNet [100], TPCIL [403], PCL [170], AANets [270], DER [468], DDE [171], GeoDL [387], PASS [510], Co2L [57], AFC [199], SP-CIL [456], ELI [192], CwD [380], CSCCT [22], FOSTER [424], FASP [300], CLS-ER [18]”
Task-Aware Information Routing from Common Representation Space in Lifelong Learning (2023, background): “Function Distance Regularization (FDR; Benjamin et al. (2018)), Dark Experience Replay (DER++; Buzzega et al. (2020)) and CLS-ER (Arani et al., 2022) leverage soft targets in addition to ground truth labels to enforce consistency regularization across previous and current model predictions.”
On the Effectiveness of Lipschitz-Driven Rehearsal in Continual Learning (2022, positive): “To assess our proposal, we perform a suite of experiments through the Mammoth framework [12, 17, 9, 6, 27, 13, 14, 50], an open-source codebase introduced in [15] for testing CL algorithms.”
Online Continual Learning with Contrastive Vision Transformer (2022, methodology): “We compare CVT with state-of-the-art and well-established Online CL baselines, including 11 rehearsal-based methods (ER [63], GEM [49], AGEM [14], GSS [4], FDR [8], HAL [13], ASERμ [67], ERT [10], RM [6], SCR [52], and CLS [5]), 2 methods leveraging Knowledge Distillation (iCaRL [62] and DER++ [9]).”
DualPrompt: Complementary Prompting for Rehearsal-free Continual Learning (2022, background): “We argue such a design could be sub-optimal from the perspective of theory of Complementary Learning Systems (CLS) [39,23], an intuition that many recent advanced CL methods are based on [8,45,4,2].”
Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (2024, methodology): “…algorithms have been proposed in continual learning (CL) area, aiming to enhance the incremental learning ability of DNNs on streaming tasks (Lopez-Paz & Ranzato, 2017; Serra et al., 2018; Gupta et al., 2020; Guo et al., 2020; Arani et al., 2022; Wang et al., 2023; Chrysakis & Moens, 2023).”
Task Agnostic Representation Consolidation: a Self-supervised based Continual Learning Approach (2022, methodology): “Samples in the replay buffer are used as model inputs for rehearsal and/or for constraint optimization of the new task (Delange et al., 2021; Arani et al., 2022).”
Continual Learning, Fast and Slow (2021, methodology): “Such studies try to emulate the hippocampus and neocortex by introducing additional learning components [43], [45] or memory units [46], [47].”
Experts Collaboration Learning for Continual Multi-Modal Reasoning (2023, background): “continual learning methods [49], [50], [66], [67], [68], [69].”
Consistency is the key to further mitigating catastrophic forgetting in continual learning (2022, background): “…by the complementary-learning system (CLS) theory (Kumaran et al., 2016) in the brain, several methods include fast and slow learning networks to consolidate the knowledge better: CLS-ER (Arani et al., 2022) used dual memory experience replay to maintain short-term and long-term semantic memories.”