Online Continual Learning under Extreme Memory Constraints

By Enrico Fini, Stéphane Lathuilière, Enver Sangineto, Moin Nabi, Elisa Ricci. 2020

74 citations recorded by the source corpus.

Continual Learning (CL) aims to develop agents emulating the human ability to sequentially learn new tasks while being able to retain knowledge obtained from past experiences. In this paper, we introduce the novel problem of Memory-Constrained Online Continual Learning (MC-OCL) which imposes strict constraints on the memory overhead that a possible algorithm can use to avoid catastrophic forgetting. As most, if not all, previous CL methods violate these constraints, we propose an algorithmic solution to MC-OCL: Batch-level Distillation (BLD), a regularization-based CL approach, which effectively balances stability and plasticity in order to learn from data streams, while preserving the ability to solve old tasks through distillation. Our extensive experimental evaluation, conducted on three publicly available benchmarks, empirically demonstrates that our approach successfully addresses the MC-OCL problem and achieves comparable accuracy to prior distillation methods requiring higher memory overhead.

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

A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, methodology): “In addition, there are some OCL methods without using experience replay, such as by distilling knowledge from the current batch of training samples (BLD [117]), and using”
A soft nearest-neighbor framework for continual semi-supervised learning (2022, neutral): “The first category comprises regularization-based methods [13,14,20,22,24,25,30,45,54], which address the problem of catastrophic forgetting by introducing appropriate regularization terms in the objective function or identifying a set of parameters that are most relevant for certain tasks.”
Uncertainty-aware Contrastive Distillation for Incremental Semantic Segmentation (2022, background): “Existing works in the literature can be grouped into three main categories: rehearsal-based methods [16], [30], [31], [32], [33], parameter isolation approaches [34], [35] and distillation-based techniques [14], [36], [37].”
Self-Supervised Models are Continual Learners (2021, background): “[7, 12, 20], while a few works employ feature distillation [10, 16] which is viable even without supervision.”
Incremental Learning in Semantic Segmentation from Image Labels (2021, background): “Even if in image classification it has been exhaustively studied [1, 11, 20, 22, 24, 30, 38, 41, 54, 57, 68], in semantic segmentation it is still in its early stages [8, 9, 21, 33, 43, 45–47].”
Continual Learning on Noisy Data Streams via Self-Purified Replay (2021, background): “There have been three main branches to train a model from continual data streams: regularization [51, 19, 38, 3], expansion [71, 89, 43], and replay [52, 9, 10, 69, 34].”
Recent Advances of Continual Learning in Computer Vision: An Overview (2021, neutral): “In object detection, Perez et al. [81] proposed OpeNended Centre nEt (ONCE) which adapts CentreNet [83] detector towards the continual learning scenario where new tasks are registered with the help of meta learning.”
Layerwise Optimization by Gradient Decomposition for Continual Learning (2021, positive): “We also achieve the state-of-the-art results which are the same MUC [11].”
Rainbow Memory: Continual Learning with a Memory of Diverse Samples (2021, methodology): “Many previous works have been evaluated either of online [13, 1, 23] or ofﬂine [48, 39, 5, 4] setup, while GDumb [38] reports on both of setups.”
Continual Adaptation of Visual Representations via Domain Randomization and Meta-learning (2020, background): “…replay and overcome catastrophic forgetting by storing samples from old tasks/distributions and regularly feeding them again to the model [23, 36, 5, 7, 49, 21, 45]; (iii) regularization meth-ods, that propose ways of constraining the tasks’ objectives in order to avoid forgetting [35, 29, 64, 17].”
Formalizing the Generalization-Forgetting Trade-off in Continual Learning (2021, background): “For example, common solutions to the catastrophic forgetting issue include (1) representation-driven approaches [49, 25], (2) regularization approaches [27, 2, 34, 16, 48, 48, 24, 37, 10, 43], and (3) memory/experience replay [31, 32, 11, 12, 17].”