Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

On the Soft-Subnetwork for Few-shot Class Incremental Learning

About

Inspired by Regularized Lottery Ticket Hypothesis (RLTH), which hypothesizes that there exist smooth (non-binary) subnetworks within a dense network that achieve the competitive performance of the dense network, we propose a few-shot class incremental learning (FSCIL) method referred to as \emph{Soft-SubNetworks (SoftNet)}. Our objective is to learn a sequence of sessions incrementally, where each session only includes a few training instances per class while preserving the knowledge of the previously learned ones. SoftNet jointly learns the model weights and adaptive non-binary soft masks at a base training session in which each mask consists of the major and minor subnetwork; the former aims to minimize catastrophic forgetting during training, and the latter aims to avoid overfitting to a few samples in each new training session. We provide comprehensive empirical validations demonstrating that our SoftNet effectively tackles the few-shot incremental learning problem by surpassing the performance of state-of-the-art baselines over benchmark datasets.

Haeyong Kang, Jaehong Yoon, Sultan Rizky Hikmawan Madjid, Sung Ju Hwang, Chang D. Yoo• 2022

Related benchmarks

TaskDatasetResultRank
Few-Shot Class-Incremental LearningminiImageNet (test)
Accuracy (Session 1)75.08
173
Few-Shot Class-Incremental LearningCIFAR100 (test)
Session 4 Top-1 Acc66.14
140
Few-Shot Class-Incremental LearningCUB200 (test)
Accuracy (Session 1)74.58
92
Few-Shot Class-Incremental LearningCIFAR100
Accuracy (S0)79.88
89
Few-Shot Class-Incremental LearningCUB-200
Session 1 Accuracy74.58
85
Few-Shot Class-Incremental LearningCUB200 (incremental sessions)
Session 0 Accuracy78.07
37
Semantic segmentationMed JASCL-Disjoint Session 0: TS
Dice Score82
28
Semantic segmentationMed JASCL-Disjoint Session 1: AMOS
Dice Score30.5
28
Semantic segmentationMed JASCL-Disjoint Session 2: BCV
Dice Score14.6
28
Continual SegmentationMed JASCL Disjoint
Total Drop (%)82.2
28
Showing 10 of 27 rows

Other info

Follow for update