Meta Continual Learning

By Risto Vuorio, Dong-Yeon Cho, Daejoong Kim, Jiwon Kim. 2018

28 citations recorded by the source corpus.

Using neural networks in practical settings would benefit from the ability of the networks to learn new tasks throughout their lifetimes without forgetting the previous tasks. This ability is limited in the current deep neural networks by a problem called catastrophic forgetting, where training on new tasks tends to severely degrade performance on previous tasks. One way to lessen the impact of the forgetting problem is to constrain parameters that are important to previous tasks to stay close to the optimal parameters. Recently, multiple competitive approaches for computing the importance of the parameters with respect to the previous tasks have been presented. In this paper, we propose a learning to optimize algorithm for mitigating catastrophic forgetting. Instead of trying to formulate a new constraint function ourselves, we propose to train another neural network to predict parameter update steps that respect the importance of parameters to the previous tasks. In the proposed meta-training scheme, the update predictor is trained to minimize loss on a combination of current and past tasks. We show experimentally that the proposed approach works in the continual learning setting.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Citations in the corpus

Continual Learning with Adaptive Weights (CLAW) (2019, background): “…approach to balance adaptability with catastrophic forgetting: a level of stability is kept via protecting parameters that greatly influence the prediction against radical changes, while allowing the rest of the parameters to change without restriction (Li & Hoiem, 2016; Vuorio et al., 2018).”
Orthogonal Gradient Descent for Continual Learning (2019, methodology): “Meta learning algorithms use a data-driven inductive bias to enhance learning new tasks (Jerfel et al., 2019; He and Jaeger, 2018; Vuorio et al., 2018; Al-Shedivat et al., 2017; Riemer et al., 2018).”
Online Fast Adaptation and Knowledge Accumulation: a New Approach to Continual Learning (2020, background): “In Vuorio et al. (2018), a gradient-based meta-continual learning is proposed.”
Defining Benchmarks for Continual Few-Shot Learning (2020, methodology): “For instance, the method proposed by Vuorio et al. (2018), an extention of MAML able to act as a loss function in the inner loop of the algorithm, has been tested exclusively on variants of MNIST.”
Meta-Learning in Neural Networks: A Survey (2020, background): “Some early stepsweremade towards defining meta-learning ready continual benchmarks in [97], [171],”
Continual Representation Learning for Biometric Identification (2020, background): “…LFL and LWF as competitors because they are efﬁcient enough for large-scale training on 5.8M images and 86K classes, while those generative model [53, 68, 69], meta-learning [19, 21, 24, 61] and dynamic-network [5,43,74] based methods are not salable or efﬁcient to train on such large benchmarks.”
Automatic Recall Machines: Internal Replay, Continual Learning and the Brain (2020, methodology): “Parameter-level methods include EWC [Kirkpatrick et al., 2017], SI [Zenke et al., 2017], GEM [LopezPaz and Ranzato, 2017], A-GEM [Chaudhry et al., 2018], Meta Continual Learning [Vuorio et al., 2018], IMM [Lee et al., 2017], Memory Aware Synapses [Aljundi et al., 2018].”
Network Transplanting (2018, background): “Especially, continual learning [16, 6, 19, 27, 23] transfers knowledge from previous tasks to guide the new task.”
Meta-learnt priors slow down catastrophic forgetting in neural networks (2019, background): “One of the first approaches that exploited a meta-learning setting to address catastrophic forgetting is Risto et al. (2018) (Vuorio et al., 2018), who proposed to learn an optimizer robust to catastrophic forgetting using a meta-dataset and pairs of adjacent tasks in a sequence.”
Understanding Continual Learning Settings with Data Distribution Drift Analysis (2021, background): “Meta-Learning: In recent years, meta-learning (Hospedales et al., 2020) has been used in many continual-learning approaches (Riemer et al., 2019; Vuorio et al., 2018; Rajasegaran et al., 2020; Javed and White, 2019; Liu et al., 2020) to solve their perilous sequence of tasks.”
Edge-Labeling Graph Neural Network for Few-Shot Learning (2019, background): “…meta-learning [1] has been recently arisen in various areas including especially task-generalization problems such as few-shot learning [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], learn-to-learn [16, 17, 18], non-stationary reinforcement learning[19, 20, 21], and continual learning [22, 23].”
Adversarial Meta-Learning (2018, unknown): “A recent work [25] presented a meta-training scheme for mitigating catastrophic forgetting by training another neural network to predict parameter update steps with respect to importance of parameters to previous tasks.”