Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Constrained Few-shot Class-incremental Learning

About

Continually learning new classes from fresh data without forgetting previous knowledge of old classes is a very challenging research problem. Moreover, it is imperative that such learning must respect certain memory and computational constraints such as (i) training samples are limited to only a few per class, (ii) the computational cost of learning a novel class remains constant, and (iii) the memory footprint of the model grows at most linearly with the number of classes observed. To meet the above constraints, we propose C-FSCIL, which is architecturally composed of a frozen meta-learned feature extractor, a trainable fixed-size fully connected layer, and a rewritable dynamically growing memory that stores as many vectors as the number of encountered classes. C-FSCIL provides three update modes that offer a trade-off between accuracy and compute-memory cost of learning novel classes. C-FSCIL exploits hyperdimensional embedding that allows to continually express many more classes than the fixed dimensions in the vector space, with minimal interference. The quality of class vector representations is further improved by aligning them quasi-orthogonally to each other by means of novel loss functions. Experiments on the CIFAR100, miniImageNet, and Omniglot datasets show that C-FSCIL outperforms the baselines with remarkable accuracy and compression. It also scales up to the largest problem size ever tried in this few-shot setting by learning 423 novel classes on top of 1200 base classes with less than 1.6% accuracy drop. Our code is available at https://github.com/IBM/constrained-FSCIL.

Michael Hersche, Geethan Karunaratne, Giovanni Cherubini, Luca Benini, Abu Sebastian, Abbas Rahimi• 2022

Related benchmarks

TaskDatasetResultRank
Few-Shot Class-Incremental LearningminiImageNet (test)
Accuracy (Session 1)76.45
173
Few-Shot Class-Incremental LearningCIFAR100 (test)
Session 4 Top-1 Acc59.84
122
Few-Shot Class-Incremental LearningCIFAR100
Accuracy (S0)77.47
67
Few-Shot Class-Incremental LearningMiniImagenet
Avg Accuracy61.61
31
Few-Shot Class-Incremental LearningminiImageNet 60 base classes 5-way 5-shot (incremental)
Session 0 Accuracy76.4
20
Few-Shot Class-Incremental LearningCIFAR-100 60 base classes, 5-way 5-shot, Session 0
Accuracy77.47
20
Few-Shot Class-Incremental LearningCIFAR-100 60 base classes 5-way 5-shot Session 1
Accuracy72.4
20
Few-Shot Class-Incremental LearningCIFAR-100 60 base classes 5-way 5-shot Session 2
Accuracy67.47
20
Few-Shot Class-Incremental LearningCIFAR-100 60 base classes 5-way 5-shot Session 3
Accuracy63.25
20
Few-Shot Class-Incremental LearningCIFAR-100 60 base classes 5-way 5-shot Session 4
Accuracy59.84
20
Showing 10 of 17 rows

Other info

Code

Follow for update