Task Agnostic Continual Learning via Meta Learning

By Xu He, Jakub Sygnowski, Alexandre Galashov, Andrei A. Rusu, Yee Whye Teh, Razvan Pascanu. 2019

103 citations recorded by the source corpus.

While neural networks are powerful function approximators, they suffer from catastrophic forgetting when the data distribution is not stationary. One particular formalism that studies learning under non-stationary distribution is provided by continual learning, where the non-stationarity is imposed by a sequence of distinct tasks. Most methods in this space assume, however, the knowledge of task boundaries, and focus on alleviating catastrophic forgetting. In this work, we depart from this view and move the focus towards faster remembering -- i.e measuring how quickly the network recovers performance rather than measuring the network's performance without any adaptation. We argue that in many settings this can be more effective and that it opens the door to combining meta-learning and continual learning techniques, leveraging their complementary advantages. We propose a framework specific for the scenario where no information about task boundaries or task identity is given. It relies on a separation of concerns into what task is being solved and how the task should be solved. This framework is implemented by differentiating task specific parameters from task agnostic parameters, where the latter are optimized in a continual meta learning fashion, without access to multiple tasks at the same time. We showcase this framework in a supervised learning scenario and discuss the implication of the proposed formalism.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, 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 Meta & Continual Learning

Citations in the corpus

Improving Task-free Continual Learning by Distributionally Robust Memory Evolution (2022, background): “Task-free continual learning (He et al., 2019; Zeno et al., 2019; Aljundi et al., 2019b; Chrysakis & Moens, 2020) instead focuses on the more general case where the data distribution could change arbitrarily without explicit task splits.”
Continual Learning via Local Module Composition (2021, background): “Unfortunately, in many realistic CL scenarios task identities are unavailable at test time [23, 35, 14].”
Continual semi-supervised learning through contrastive interpolation consistency (2021, background): “Free Learning does not provide task identities either at inference or at training time [18, 19, 12].”
Optimizing Reusable Knowledge for Continual Learning via Metalearning (2021, methodology): “It has been used as a training strategy in sequence learning [43, 44] and continual learning scenarios [45, 46, 47, 48].”
New Insights on Reducing Abrupt Representation Change in Online Continual Learning (2021, background): “The single-head assumption is further studied in task-agnostic continual learning settings (He et al., 2019; Caccia et al., 2020; Ostapenko et al., 2021; Von Oswald et al., 2021) in which the task-boundary assumption, amongst others, is also relaxed.”
Posterior Meta-Replay for Continual Learning (2021, background): “Although auxiliary systems can be used to infer task identity [19, 75], here we exploit predictive uncertainty, assuming task identity can be inferred from the input alone.”
Embracing Change: Continual Learning in Deep Neural Networks. (2020, background): “As pointed out in [83], this perspective can be insightful for continual learning.”
Gradient-based Editing of Memory Examples for Online Task-free Continual Learning (2020, background): “Due to its broader applicability to realworld data-streams, a number of algorithms have been adapted to the task-free setup [2, 14, 15, 22, 44].”
Understanding the Role of Training Regimes in Continual Learning (2020, methodology): “Continual Learning as an emerging field in Artificial Intelligence is connected to many other areas such as Meta Learning [4, 35, 24, 62], Few Shot Learning [75, 20], Multi-task and Transfer Learning [24, 35] and the closely related problem of exploring task boundary or task detection [60, 2, 25, 36”
Orthogonal Gradient Descent for Continual Learning (2019, background): “This makes it applicable when the task shift does not arrive as a distinct event, but rather a gradual shift (He et al., 2019).”
Continual learning with hypernetworks (2019, methodology): “Interestingly, our task-conditioned hypernetworks would ﬁt very well with methods that rely on fast remembering (a recently proposed approach which appeared in parallel to our paper, He et al., 2019).”
Memory-Efficient Semi-Supervised Continual Learning: The World is its Own Replay Buffer (2021, neutral): “Approaches include regularization penalties [50, 51, 52, 53, 54], meta learning [55], model compression [56, 57, 58], or knowledge distillation [5, 7, 8, 9].”
Task-Agnostic Continual Learning Using Online Variational Bayes With Fixed-Point Updates (2020, methodology): “Indeed, some of thus extensions were considered in He et al. (2019) using the diagonal version of FOO-VB (published in our preliminary preprint Zeno, Golan, Hoffer, & Soudry, 2019).”
Continuous Meta-Learning without Tasks (2019, methodology): “Moreover, a switching sinusoid problem is a popular benchmark in continuous learning (He et al., 2019; Javed & White, 2019).”
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.”