Constrained Few-shot Class-incremental Learning

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

204 citations recorded by the source corpus.

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, mini-ImageNet, 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.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, 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 Radiology, Nuclear Medicine and Imaging

In Meta & Continual Learning

Citations in the corpus

A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, neutral): “…extraconstraintsfrommargin-basedrepresentations[258],etc. Thesecondistoobtainanadaptiveclassiﬁerfromasequence of novel classes, such as by evolving the classiﬁer weights with a graph attention network [259], performing hyperdimen-sional computing [260], sampling stochastic classiﬁers [261], etc.”
First Session Adaptation: A Strong Replay-Free Baseline for Class-Incremental Learning (2023, methodology): “Some class incremental learning methods leverage large initial training session or start from a pre-trained network, e.g., [35, 9, 36] and show improved over-all continual learning performance without questioning the role of such pre-training steps and whether the learned representations have in…”
Towards Label-Efficient Incremental Learning: A Survey (2023, background): “Constrained-FSCIL [Hersche et al., 2022] is a recent technique that operates on classifier embeddings (prototypes) instead.”
A Survey on Few-Shot Class-Incremental Learning (2023, background): “According to Hersche et al. (2022), the input images are mapped to quasi-orthogonal prototypes from the perspective of hyperdimensional computing.”
Warping the Space: Weight Space Rotation for Class-Incremental Few-Shot Learning (2023, methodology): “Additionally, we further compare our method with more recent works (Zhou et al., 2022b; Chi et al., 2022; Hersche et al., 2022; Liu et al., 2022) on mini ImageNet dataset.”
Neural Collapse Inspired Feature-Classifier Alignment for Few-Shot Class Incremental Learning (2023, methodology): “Following Hersche et al. (2022), we only keep a memory M ( t ) of the mean intermediate feature h c for each old class c .”
Pre-trained Vision and Language Transformers are Few-Shot Incremental Learners (2024, methodology): “Prototype-based methods [6, 19, 39] aim to align prototypes with classifier weights to enhance the classification performance.”