Online Continual Learning with Natural Distribution Shifts: An Empirical Study with Visual Data

By Zhipeng Cai, Ozan Sener, Vladlen Koltun. 2021

103 citations recorded by the source corpus.

Continual learning is the problem of learning and retaining knowledge through time over multiple tasks and environments. Research has primarily focused on the incremental classification setting, where new tasks/classes are added at discrete time intervals. Such an "offline" setting does not evaluate the ability of agents to learn effectively and efficiently, since an agent can perform multiple learning epochs without any time limitation when a task is added. We argue that "online" continual learning, where data is a single continuous stream without task boundaries, enables evaluating both information retention and online learning efficacy. In online continual learning, each incoming small batch of data is first used for testing and then added to the training set, making the problem truly online. Trained models are later evaluated on historical data to assess information retention. We introduce a new benchmark for online continual visual learning that exhibits large scale and natural distribution shifts. Through a large-scale analysis, we identify critical and previously unobserved phenomena of gradient-based optimization in continual learning, and propose effective strategies for improving gradient-based online continual learning with real data. The source code and dataset are available in: https://github.com/IntelLabs/continuallearning.

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, background): “OCL InstAParam [68], CBRS [75], ASER [382], InfoRS [398], DVC [143], CTN [336], NCCL [475], ILOS [158], SCR [284], CVT [441], OCM [145], ER-ACE [47], RAR [498], PoLRS [50], AOP [144], CV [159]”
The CLEAR Benchmark: Continual LEArning on Real-World Imagery (2022, methodology): “Such boundary-agnostic settings have been explored in recent works [1, 4, 23, 59], in which a non-iid data stream continuously spits out new samples without a notion of task switch.”
Continual Learning on a Diet: Learning from Sparsely Labeled Streams Under Constrained Computation (2024, background): “While some online continual learning approaches attempt to formulate the online data stream (Chaudhry et al., 2019a; Cai et al., 2021), they prioritize the formulation in batch-wise training and evaluation, and often overlook the regularization of computation and time in these algorithms.”
Learning from One Continuous Video Stream (2023, background): “Instead, they typically focus on streams of independent images [4].”
Computationally Budgeted Continual Learning: What Does Matter? (2023, methodology): “We evaluate and benchmark at scale various classical CL sampling strategies (Uniform, Class-Balanced [37], Recency-Biased [26], FIFO [11, 14], Max Loss, Uncertainity Loss [6], and KMeans [14]), CL distillation strategies (BCE [41], MSE [9], CrossEntropy [49], and Cosine [22]) and FC layer correction”
CLNeRF: Continual Learning Meets NeRF (2023, background): “However, this does not model the real-world distribution shifts introduced by the change of time [4, 16], such as the change of scene appearance ( e.g ., lighting and weather) and geometry ( e.g ., new decoration of a room).”