--- title: Class-Continual Learning Experiments url: https://www.emergentmind.com/topics/class-continual-learning-experiments type: topic --- # Class-Continual Learning Experiments Class-Continual Learning Experiments Class-continual learning (class-incremental learning, class-IL) investigates the sequential adaptation of models to streams of classification tasks where new classes are introduced incrementally. The objective is to enable models to learn newly arriving classes while maintaining performance on all previously encountered classes, *without* access to the full data history. Catastrophic forgetting, where performance on earlier classes degrades upon learning new ones, is the central challenge. Research in this field develops methods, benchmarks, and evaluation metrics for addressing this challenge under increasingly realistic constraints, with a significant body of work emerging on arXiv and related venues. ## 1. Experimental Protocols and Benchmarking Paradigms Most class-IL research adopts a sequential task protocol where, at each step $t$, a model receives data from a disjoint set of classes $C_t$, with $C_t \cap C_{t'} = \emptyset$ for $t \neq t'$. After each step, the learner must classify among all $\bigcup_{s\leq t} C_s$ observed classes, operating without explicit task identifiers at inference time. **Representative datasets and splits:** - **Image domain:** CIFAR-10/100 (e.g., 5$\times$2 or 10$\times$10 splits) [2310.03898, 2311.14905, 2210.05398, 2112.01713], TinyImageNet (10$\times$20) [2504.07240, 2211.12292], miniImageNet (20$\times$5) [2201.00766], Caltech-256 [2504.07240]. - **Video domain:** UCF101, Kinetics, ActivityNet—each split into 10 or 20 class increments [2201.09381]. - **Specialized domains:** Medical imaging (PathMNIST, CheXpert, etc.) with cross-specialty splits [2311.04301]. **Protocols vary along:** - *Strictness of constraints:* Whether a memory buffer is allowed (replay), whether any exemplars or features from old tasks can be stored, constant model capacity, presence/absence of pretraining [2310.03898, 2211.12292, 2303.07811]. - *Task structure realism:* RealCL evaluates in non-uniform, purely random class-incremental streams [2404.07729]; OTFCL (Online Task-Free CL) removes precise task boundaries [2404.13576]. - *Class repetition:* Class-Incremental with Repetition (CIR) allows class reappearance in later tasks and the use of unlabeled external data [2508.12692]. **Evaluation:** At the end of each task or task sequence, models are evaluated on unioned test sets, with no information about task identity. ## 2. Algorithmic Paradigms Techniques for class-continual learning fall into several main paradigms, with numerous methodological innovations in recent literature: **A. Memory-based Replay** Models maintain a buffer (of data, features, or prototypes) to replay past examples during new task learning. Strategies include: - **Experience Replay (ER):** Reservoir sampling of raw images or features [2210.05398, 2112.01713, 2308.03810], with variants for class balancing (E-BRS) [2308.03810] and selective replay (C-CMR). - **Prototype-based Replay:** Label-free selection of cluster centers in latent space, with replay buffer storing only prototypes/support, and cluster preservation via MMD [2504.07240]. - **Generative Replay:** Synthetic data for old classes are generated by VAEs or GANs, with training guided by time-aware regularization of loss terms based on class age [2310.03898]. - **Dark Experience Replay (DER/DER++):** Storing not just samples but also their historical logits for distillation [2201.00766]. X-DER [2201.00766] further adds memory revision and pre-allocation for future heads. **B. Regularization-based Methods** Introduce penalties to prevent significant drift in parameters critical for previous tasks: - **EWC, MAS, SI:** Fisher or sensitivity-based penalties on parameter changes [2310.03898, 2311.04301]. - **Projected Functional Regularization (PFR):** For ViTs, aligns backbone features across task shifts through a chain of learnable projectors [2211.12292]. **C. Gradient and Subspace Methods** Orthogonalize updates with respect to important subspaces: - **Task and Class Gradient Projection:** GPM projects gradients away from task subspaces; CGP extends this to class-specific bases, merging similar classes, and adding supervised contrastive loss for plasticity [2311.14905]. **D. Feature-Space and Contrastive Methods** Leverage embedding alignment and contrastive learning: - **Feature Propagation & Contrastive Alignment:** Replay and alignment of feature extractors via contrastive losses and embedding propagation [2112.01713]. - **Diversified Feature Augmentation:** MOCA creates intra-class diversity via perturbations, both model-agnostic and model-aware, to mitigate collapse [2210.05398]. - **Specialized Video CIL:** Temporal consistency loss prepares representations for under-sampled and downsampled video frames [2201.09381]. **E. Exemplar-Free and Interpretability-Preserving Approaches** - **ICICLE:** Exemplar-free, interpretable class-incremental learning with interpretability regularization, prototype proximity-based initialization, and logit bias compensation [2303.07811]. - **Gated Class Attention:** Parameter-masking mechanisms in ViTs for class-IL without exemplars, plus drift compensation [2211.12292]. **F. Exemplar-Free/Fully Unsupervised Methods** - **I²CANSAY:** Non-exemplar, task-free CL with memoryless online adaptation using inter-class analogical pseudo-features and per-class significance reweighting [2404.13576]. **G. Knowledge Distillation and External Data Use** - **Multi-level knowledge distillation:** Models maintain EMA snapshots for feature/logit distillation on unlabeled/external data, with dynamic SSL regularization and class-repetition [2508.12692]. - **Discriminative distillation:** Additional loss for similar (confused) class pairs [2108.05187]. ## 3. Key Metrics and Their Role Several metrics have evolved to expose both average and worst-case behavior of continual learners: - **Average Accuracy (AAC / ACC):** Mean accuracy across all classes seen so far, after each task and finally [2310.03898, 2311.14905]. - **Backward Transfer (BWT):** Measures forgetting. $BWT=(1/(T−1))\sum_{i=1}^{T−1}(A_{T,i}–A_{i,i})$ [2311.14905]. - **Forgetting (FGT / AF):** Drop in per-task accuracy since the task was first learned [2112.01713, 2309.00462]. - **Minimal Incremental Class Accuracy (MICA):** Worst-class accuracy after each task; critical for industrial and safety-sensitive applications [2404.06972]. - **Weighted Aggregate MICA (WAMICA):** A stability–fairness scalar penalizing large swings in classwise minima [2404.06972]. - **Rescaled Accuracy (RAA) and Forgetting (RAF):** Normalize for task difficulty, correcting misleading impressions due to growing class set size [2309.00462]. - **Supervised/unsupervised class-based metrics:** For methods without access to labels during replay, cluster alignment, and proxy-based accuracy are used [2504.07240]. A key insight is that average accuracy can systematically overestimate true performance, masking the presence of unlearned classes or especially poor worst-case class performance [2404.06972]. Adoption of MICA, WAMICA, and rescaled metrics addresses this oversight and provides a more rigorous assessment. ## 4. Controlled Studies, Ablations, and Model Insights Ablation and sensitivity analysis play a crucial role in evaluating the contributions of individual components: - **Time-aware regularization ablations:** Show improved sample quality and accuracy only if both α, β decay dynamically by class age [2310.03898]. - **Gradient basis refinement and contrastive regularization:** Each adds significant accuracy and reduces forgetting (CGP) [2311.14905]. - **Replay buffer size and class-balance sensitivity:** Adaptive replay and entropy balancing maintain performance even with smaller buffers and imbalanced classes [2308.03810]. - **Discriminative and feature-space distillation ablations:** Demonstrate that targeted separation of confused class pairs further shrinks error rates vs. generic distillation alone [2108.05187]. - **Unsupervised prototype and pseudo-feature selection:** Eliminating label dependence from the buffer or replay yields only modest accuracy loss, and can surpass standard replay-based baselines in some domains [2504.07240, 2404.13576]. - **Interpretability regularization and prototype initialization strategies:** Simultaneous deployment reduces concept drift and enhances both accuracy and explanatory consistency in CUB-200 class splits [2303.07811]. A consistent theme is that strategies combining plasticity (for new-class acquisition) with adaptive regularization, memory diversity, and explicit intra- or inter-class separation yield the most robust, stable performance across realistic incremental learning streams. ## 5. Challenges, Trends, and Open Directions Major open challenges and trends in class-continual learning experiments include: - **Exemplar-free and privacy-preserving methods:** Driven by privacy and resource constraints, research increasingly targets exemplar-free (or fully label-free) replay and representation methods, with mechanisms such as pseudo-feature generation, generative replay, and compressed feature buffering [2310.03898, 2303.07811, 2504.07240, 2404.13576]. - **Realism in experimental design:** More complex scenarios, such as RealCL (random/imbalanced class streams) [2404.07729], CIR (class-incremental with repetition and large external unlabeled data) [2508.12692], and concept drift adaptation (reactive subspace buffers) [2104.11861], expose the limitations of methods tuned for idealized, task-structured benchmarks. - **Evaluation best practices:** There is a continuing shift away from reporting only average accuracy, with newer work rigorously tracking worst-case accuracy, fairness across classes, and proper normalization for expanding class sets [2404.06972, 2309.00462]. - **Modular, plug-and-play components:** Model-agnostic regularization (e.g., MOCA), orthogonalization (CGP, GPM), and interpretability-aware layers (ICICLE, Gated Class-Attention) are being adopted to dissociate stability–plasticity tradeoffs from dataset specifics [2210.05398, 2311.14905, 2211.12292]. ## 6. Domain-Specific and Multi-Domain Class-Continual Learning Recent research evaluates class-IL in domains where class definitions, data distributions, and dynamics differ substantially from standard vision tasks: - **Medical imaging CL:** Scenario-based evaluation reveals that hybrid approaches—memory-based replay with regularization—reduce forgetting and perform robustly in inter-hospital, cross-specialty, and intra-specialty shifts [2311.04301]. - **Video CL:** Temporal frame-subsampling, memory constraints, and action-class imbalances create unique continual learning bottlenecks, addressed by temporal consistency losses and frame-level replay [2201.09381]. - **Fine-grained and high-class-count domains:** Datasets such as CUB-200, Stanford Cars, and Caltech-256 are used with interpretable and prototype-based methods, requiring new regularization strategies for both performance and explanation stability [2303.07811, 2504.07240]. This diversification of domain and protocol emphasizes the need for flexible, general-purpose class-IL methodology and rigorous, task-structure-agnostic evaluation. --- These converging advances collectively define the state of class-continual learning experiments as of 2026: an active field characterized by methodological breadth, increasingly realistic benchmarks, and a robust, metric-driven evaluation culture. Source: https://www.emergentmind.com/topics/class-continual-learning-experiments