Three scenarios for continual learning
read the original abstract
Standard artificial neural networks suffer from the well-known issue of catastrophic forgetting, making continual or lifelong learning difficult for machine learning. In recent years, numerous methods have been proposed for continual learning, but due to differences in evaluation protocols it is difficult to directly compare their performance. To enable more structured comparisons, we describe three continual learning scenarios based on whether at test time task identity is provided and--in case it is not--whether it must be inferred. Any sequence of well-defined tasks can be performed according to each scenario. Using the split and permuted MNIST task protocols, for each scenario we carry out an extensive comparison of recently proposed continual learning methods. We demonstrate substantial differences between the three scenarios in terms of difficulty and in terms of how efficient different methods are. In particular, when task identity must be inferred (i.e., class incremental learning), we find that regularization-based approaches (e.g., elastic weight consolidation) fail and that replaying representations of previous experiences seems required for solving this scenario.
This paper has not been read by Pith yet.
Forward citations
Cited by 18 Pith papers
-
PrimeKG-CL: A Continual Graph Learning Benchmark on Evolving Biomedical Knowledge Graphs
PrimeKG-CL supplies the first continual graph learning benchmark using authentic temporal snapshots from nine biomedical databases, showing strong interactions between embedding decoders and learning strategies plus l...
-
MIST: Reliable Streaming Decision Trees for Online Class-Incremental Learning via McDiarmid Bound
MIST fixes unreliable splits in streaming decision trees for class-incremental learning by replacing Hoeffding-style bounds with a K-independent McDiarmid radius on Gini, plus Bayesian parent-to-child inheritance and ...
-
MIST: Reliable Streaming Decision Trees for Online Class-Incremental Learning via McDiarmid Bound
MIST fixes unreliable splits in streaming decision trees for class-incremental learning by using a K-independent McDiarmid bound on Gini impurity, Bayesian moment projection for knowledge transfer, and KLL quantile sk...
-
Characterizing and Correcting Effective Target Shift in Online Learning
Online kernel regression equals offline regression with shifted targets; correcting the targets lets online learning match offline performance and outperform true targets in continual image classification.
-
Continual Learning for fMRI-Based Brain Disorder Diagnosis via Functional Connectivity Matrices Generative Replay
A structure-aware VAE generates realistic FC matrices for replay, combined with multi-level knowledge distillation and hierarchical contextual bandit sampling, to enable continual fMRI-based brain disorder diagnosis a...
-
SLE-FNO: Single-Layer Extensions for Task-Agnostic Continual Learning in Fourier Neural Operators
SLE-FNO achieves zero forgetting and strong plasticity-stability balance in continual learning for FNO surrogate models of pulsatile blood flow by adding minimal single-layer extensions across four out-of-distribution tasks.
-
Learning to Discover at Test Time
TTT-Discover applies test-time RL to set new state-of-the-art results on math inequalities, GPU kernels, algorithm contests, and single-cell denoising using an open model and public code.
-
Exemplar-Free Continual Learning for State Space Models
Inf-SSM constrains the infinite-horizon evolution of SSMs via Grassmannian geometry and an efficient O(n^2) Sylvester solver to enable exemplar-free continual learning with reduced forgetting.
-
The Curse of Recursion: Training on Generated Data Makes Models Forget
Use of model-generated content in training causes irreversible loss of distribution tails, termed model collapse, in VAEs, GMMs, and LLMs.
-
NORACL: Neurogenesis for Oracle-free Resource-Adaptive Continual Learning
NORACL dynamically grows network capacity via neurogenesis-inspired signals to achieve oracle-level continual learning performance without pre-specifying architecture size.
-
Cortex-Inspired Continual Learning: Unsupervised Instantiation and Recovery of Functional Task Networks
FTN achieves near-zero forgetting on continual learning benchmarks by isolating task subnetworks via self-organizing binary masks generated through gradient descent, smoothing, and k-winner-take-all.
-
Fine-Tuning Regimes Define Distinct Continual Learning Problems
The relative rankings of continual learning methods are not preserved across different fine-tuning regimes defined by trainable parameter depth.
-
MePo: Meta Post-Refinement for Rehearsal-Free General Continual Learning
MePo refines pretrained backbones via meta-learning on constructed pseudo tasks and initializes a meta covariance matrix to enable robust second-order alignment, yielding 12-15% gains on CIFAR-100, ImageNet-R and CUB-...
-
Measuring the stability and plasticity of recommender systems
A new offline protocol to profile recommender algorithms by stability in retaining past patterns and plasticity in adapting to changes upon retraining, with preliminary results on the GoodReads dataset.
-
A Parameter-free Adaptive Resonance Theory-based Topological Clustering Algorithm Capable of Continual Learning
An ART-based topological clustering algorithm estimates vigilance and edge deletion thresholds automatically using determinantal point process and edge age for parameter-free continual learning.
-
RECALL: Rehearsal-free Continual Learning for Object Classification
RECALL achieves rehearsal-free continual learning for object classification by logit recall before new training, regression regularization, Mahalanobis loss on known categories, and new heads per sequence, outperformi...
-
HEDP: A Hybrid Energy-Distance Prompt-based Framework for Domain Incremental Learning
HEDP uses energy regularization inspired by Helmholtz free energy plus hybrid energy-distance weighting in prompts to improve domain selection and achieve a 2.57% accuracy gain on benchmarks like CORe50 while mitigati...
-
Incremental learning for audio classification with Hebbian Deep Neural Networks
A kernel plasticity approach in Hebbian DNNs for incremental sound classification achieves 76.3% accuracy over five steps on ESC-50, outperforming the 68.7% baseline without plasticity.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.