Understanding Continual Learning Settings with Data Distribution Drift Analysis

By Timothée Lesort, Massimo Caccia, Irina Rish. 2021

62 citations recorded by the source corpus.

Classical machine learning algorithms often assume that the data are drawn i.i.d. from a stationary probability distribution. Recently, continual learning emerged as a rapidly growing area of machine learning where this assumption is relaxed, i.e. where the data distribution is non-stationary and changes over time. This paper represents the state of data distribution by a context variable $c$. A drift in $c$ leads to a data distribution drift. A context drift may change the target distribution, the input distribution, or both. Moreover, distribution drifts might be abrupt or gradual. In continual learning, context drifts may interfere with the learning process and erase previously learned knowledge; thus, continual learning algorithms must include specialized mechanisms to deal with such drifts. In this paper, we aim to identify and categorize different types of context drifts and potential assumptions about them, to better characterize various continual-learning scenarios. Moreover, we propose to use the distribution drift framework to provide more precise definitions of several terms commonly used in the continual learning field.

Subfields: Artificial Intelligence

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

Citations in the corpus

Continual Learning via Local Module Composition (2021, background): “The rapid growth of continual learning has lead researchers to work on empirical studies [20, 58, 56], surveys [32, 42, 55, 66, 67] as well as CL-specific software [72, 22, 59].”
New Insights on Reducing Abrupt Representation Change in Online Continual Learning (2021, background): “Research on continual learning can be divided based on the sequential setting being targeted (see Zeno et al. (2018); van de Ven & Tolias (2019); Normandin et al. (2021); Lesort et al. (2021) for categorizations of the settings and De Lange et al. (2019) for a broad survey on continual learning).”
Continual Learning with Foundation Models: An Empirical Study of Latent Replay (2022, methodology): “We used those datasets to create class-incremental scenarios, i.e. a new task brings new classes (van de Ven and Tolias, 2019; Lesort et al., 2021a).”
Pretrained Language Model in Continual Learning: A Comparative Study (2022, background): “The rapid growth of continual learning has lead researchers to work on empirical studies (De Lange et al., 2019; Lesort et al., 2021b), surveys (Hadsell et al., 2020; Khetarpal et al., 2020; Lesort et al., 2021a; Mundt et al., 2020; 2021) as well as CL-speciﬁc software (Normandin et al., 2021;…”
Continual Learning in Deep Networks: an Analysis of the Last Layer (2021, methodology): “We evaluate the continual learning performance on virtual concept drifts (incremental scenario), on domain drifts (lifelong scenario) as described in [27].”
Tackling Catastrophic Forgetting and Background Shift in Continual Semantic Segmentation (2021, background): “which is not the case in continual segmentation where multiple shifts [71], [72], [90], happen in the training tasks, different from the testing tasks.”
Scaling the Number of Tasks in Continual Learning (2022, background): “These scenarios are usually built with different data distribution shifts [Lesort et al., 2021b] such as new tasks introduce new classes (class-incremental), new domains (domain-incremental), or new labelization of data Abdelsalam et al. [2021].”
Three types of incremental learning (2022, unknown): “However, this setting does not reflect well the arbitrary non-stationarity that can be observed in the real world 33–40 .”
CLEVA-Compass: A Continual Learning EValuation Assessment Compass to Promote Research Transparency and Comparability (2021, background): “…hand, several works have made important suggestions for continual learning desiderata with respect to evaluation protocols (Farquhar & Gal, 2018; Dı́az-Rodrı́guez et al., 2018; Kemker et al., 2018) and categorization of incremental training set-ups (van de Ven & Tolias, 2019; Lesort et al., 2021).”