Evolutionary Generalized Zero-Shot Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Dubing, Jiang, Chenyi, Zhang, Haofeng
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914791888519168
author Chen, Dubing
Jiang, Chenyi
Zhang, Haofeng
author_facet Chen, Dubing
Jiang, Chenyi
Zhang, Haofeng
contents Attribute-based Zero-Shot Learning (ZSL) has revolutionized the ability of models to recognize new classes not seen during training. However, with the advancement of large-scale models, the expectations have risen. Beyond merely achieving zero-shot generalization, there is a growing demand for universal models that can continually evolve in expert domains using unlabeled data. To address this, we introduce a scaled-down instantiation of this challenge: Evolutionary Generalized Zero-Shot Learning (EGZSL). This setting allows a low-performing zero-shot model to adapt to the test data stream and evolve online. We elaborate on three challenges of this special task, \ie, catastrophic forgetting, initial prediction bias, and evolutionary data class bias. Moreover, we propose targeted solutions for each challenge, resulting in a generic method capable of continuous evolution from a given initial IGZSL model. Experiments on three popular GZSL benchmark datasets demonstrate that our model can learn from the test data stream while other baselines fail. Codes are available at \url{https://github.com/cdb342/EGZSL}.
format Preprint
id arxiv_https___arxiv_org_abs_2211_13174
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Evolutionary Generalized Zero-Shot Learning
Chen, Dubing
Jiang, Chenyi
Zhang, Haofeng
Computer Vision and Pattern Recognition
Machine Learning
Neural and Evolutionary Computing
Attribute-based Zero-Shot Learning (ZSL) has revolutionized the ability of models to recognize new classes not seen during training. However, with the advancement of large-scale models, the expectations have risen. Beyond merely achieving zero-shot generalization, there is a growing demand for universal models that can continually evolve in expert domains using unlabeled data. To address this, we introduce a scaled-down instantiation of this challenge: Evolutionary Generalized Zero-Shot Learning (EGZSL). This setting allows a low-performing zero-shot model to adapt to the test data stream and evolve online. We elaborate on three challenges of this special task, \ie, catastrophic forgetting, initial prediction bias, and evolutionary data class bias. Moreover, we propose targeted solutions for each challenge, resulting in a generic method capable of continuous evolution from a given initial IGZSL model. Experiments on three popular GZSL benchmark datasets demonstrate that our model can learn from the test data stream while other baselines fail. Codes are available at \url{https://github.com/cdb342/EGZSL}.
title Evolutionary Generalized Zero-Shot Learning
topic Computer Vision and Pattern Recognition
Machine Learning
Neural and Evolutionary Computing
url https://arxiv.org/abs/2211.13174