Memory Replay with Data Compression for Continual Learning

By Liyuan Wang, Xingxing Zhang, Kuo Yang, L. Yu, Chongxuan Li, Lanqing Hong, Shifeng Zhang, Zhenguo Li, Yi Zhong, Jun Zhu. 2022

141 citations recorded by the source corpus.

Continual learning needs to overcome catastrophic forgetting of the past. Memory replay of representative old training samples has been shown as an effective solution, and achieves the state-of-the-art (SOTA) performance. However, existing work is mainly built on a small memory buffer containing a few original data, which cannot fully characterize the old data distribution. In this work, we propose memory replay with data compression (MRDC) to reduce the storage cost of old training samples and thus increase their amount that can be stored in the memory buffer. Observing that the trade-off between the quality and quantity of compressed data is highly nontrivial for the efficacy of memory replay, we propose a novel method based on determinantal point processes (DPPs) to efficiently determine an appropriate compression quality for currently-arrived training samples. In this way, using a naive data compression algorithm with a properly selected quality can largely boost recent strong baselines by saving more compressed data in a limited storage space. We extensively validate this across several benchmarks of class-incremental learning and in a realistic scenario of object detection for autonomous driving.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, 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 Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

In Meta & Continual Learning

Citations in the corpus

A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, neutral): “MRDC [138] formulates experience replay with data compression as determinantal point processes (DPPs), and derives a computationally efﬁcient way for online determination of the optimal compression rate.”
Learnability and Algorithm for Continual Learning (2023, background): “As pre-training models or feature extractors are increasingly used in all types of applications, including continual learning (Wang et al., 2022c; Kim et al., 2022a; Wu et al., 2022), we also take this approach.”
The Ideal Continual Learner: An Agent That Never Forgets (2023, background): “…and more challenging case of task-agnostic or task-free continual learning, where in the training phase the learner has no access to task identities (Zeno et al., 2018; 2021; Aljundi et al., 2019a; Lee et al., 2020; Jin et al., 2021; Wang et al., 2022c; Pourcel et al., 2022; Ye & Bors, 2022).”
SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model (2023, background): “Representative strategies include regularization-based approaches [19, 1, 51, 25, 41, 7], which preserve the old model and selectively stabilize changes of parameters or predictions; replaybased approaches [49, 30, 3, 44, 40, 39], which approximate and recover the previously-learned data distributio”
A Theoretical Study on Solving Continual Learning (2022, background): “[39] Liyuan Wang, Xingxing Zhang, Kuo Yang, Longhui Yu, Chongxuan Li, Lanqing Hong, Shifeng Zhang, Zhenguo Li, Yi Zhong, and Jun Zhu.”
Continual Learning by Modeling Intra-Class Variation (2022, neutral): “Due to the memory constraint of the replay buffer, it is important to design a good sample selection strategy so that the buffer stores the most representative data to prevent forgetting (Wang et al., 2022).”
CoSCL: Cooperation of Small Continual Learners is Stronger than a Big One (2022, neutral): “Representative strategies include weight regularization [20,52,1], memory replay [36,44,46], parameter isolation [41,18] and dynamic architecture [14,50].”
FOSTER: Feature Boosting and Compression for Class-Incremental Learning (2022, background): “[32,36,38,43] allocate a memory to store exemplars of previous tasks for replay when learning tasks.”
Online Continual Learning through Mutual Information Maximization (2022, background): “…et al., 2017; de Masson d’Autume et al., 2019; Wu et al., 2019; Hou et al., 2019; Zhao et al., 2021; Korycki & Krawczyk, 2021; Sokar et al., 2021; Yan et al., 2021; Hu et al., 2021; Wang et al., 2022), which uses a memory buffer to store some previous data and replays it in training a new task.”
Orchestrate Latent Expertise: Advancing Online Continual Learning with Multi-Level Supervision and Reverse Self-Distillation (2024, methodology): “Replay-based methods [6, 9, 48, 60, 63] exploit an additional memory buffer to save a subset of old training samples, to recover previous data distributions.”
Dealing with Cross-Task Class Discrimination in Online Continual Learning (2023, background): “Replay-based approaches save some past data and replay them in new task training [8, 12, 13, 24, 37, 44, 47, 54].”
Retrospective Adversarial Replay for Continual Learning (2022, background): “However, the buffer in practice is often quite small [63], e.”
Continual Detection Transformer for Incremental Object Detection (2023, methodology): “ER methods [6, 30, 33, 34, 42, 53] work instead by memorising some of the past training data (the exemplars ), replaying them in the following phases to “re-member” the old object categories.”
ReMindRAG: Low-Cost LLM-Guided Knowledge Graph Traversal for Efficient RAG (2025, unknown): “Following previous works regarding memory replay [29, 46], which refers to the recall of valuable past memories to facilitate current tasks, we design memory replay for efficient graph traversal.”