Graph-Based Continual Learning

By Binh Tang, David S. Matteson. 2020

42 citations recorded by the source corpus.

Despite significant advances, continual learning models still suffer from catastrophic forgetting when exposed to incrementally available data from non-stationary distributions. Rehearsal approaches alleviate the problem by maintaining and replaying a small episodic memory of previous samples, often implemented as an array of independent memory slots. In this work, we propose to augment such an array with a learnable random graph that captures pairwise similarities between its samples, and use it not only to learn new tasks but also to guard against forgetting. Empirical results on several benchmark datasets show that our model consistently outperforms recently proposed baselines for task-free continual learning.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Graph Learning, 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 Graph Learning

In Meta & Continual Learning

Citations in the corpus

Flattening Sharpness for Dynamic Gradient Projection Memory Benefits Continual Learning (2021, background): “As noted by [4] and [38], EWC performs poorly without multiple passes over the datasets, and GEM is not very effective under the single-head variants.”
Continual Learning on Noisy Data Streams via Self-Purified Replay (2021, background): “Recently, [77] adopts graphs to represent relational structures between samples, and [25] employs the meta-loss for learning per-parameter learning rates along with model parameters.”
New Insights on Reducing Abrupt Representation Change in Online Continual Learning (2021, background): “Zhao et al. (2019) proposes to correct last layer weights after a group of classes is learned via adjusting the weights norm. Wu et al. (2019) suggests to deploy extra additional parameters in order to linearly correct the “bias” in the shared output layer.”
How to Reuse and Compose Knowledge for a Lifetime of Tasks: A Survey on Continual Learning and Functional Composition (2022, methodology): “A drastically different technique uses graph learning to discover pairwise similarities between memory and current samples, penalizing forgetting the edges between samples instead of the predictions in order to maintain correlations between samples while permitting significant changes to the network”
Graph Few-shot Class-incremental Learning (2021, methodology): “The first family includes replay-based methods [22, 30, 32, 35], which maintain a subset of previous samples, and train models together with samples in the new session.”
DDGR: Continual Learning with Deep Diffusion-based Generative Replay (2023, background): “In addition, graph-based replay (Tang & Matteson, 2021) and gradient projection memory (Saha et al., 2021) have also been proven to be efficient.”