Wakening past concepts without past data: Class-incremental learning from online placebos

Not forgetting old class knowledge is a key challenge for class-incremental learning (CIL) when the model continuously adapts to new classes. A common technique to address this is knowledge distillation (KD), which penalizes prediction inconsistencies between old and new models. Such prediction is m...

Full description

Saved in:
Bibliographic Details
Main Authors: LIU, Yaoyao, LI, Yingying, SCHIELE, Bernt, SUN, Qianru
Format: text
Language:English
Published: Institutional Knowledge at Singapore Management University 2024
Subjects:
Online Access:https://ink.library.smu.edu.sg/sis_research/9207
https://ink.library.smu.edu.sg/context/sis_research/article/10212/viewcontent/Liu_Wakening_Past_Concepts_Without_Past_Data_Class_Incremental_Learning_From_Online_WACV_2024_paper.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Singapore Management University
Language: English
id sg-smu-ink.sis_research-10212
record_format dspace
spelling sg-smu-ink.sis_research-102122024-08-13T05:03:52Z Wakening past concepts without past data: Class-incremental learning from online placebos LIU, Yaoyao LI, Yingying SCHIELE, Bernt SUN, Qianru Not forgetting old class knowledge is a key challenge for class-incremental learning (CIL) when the model continuously adapts to new classes. A common technique to address this is knowledge distillation (KD), which penalizes prediction inconsistencies between old and new models. Such prediction is made with almost new class data, as old class data is extremely scarce due to the strict memory limitation in CIL. In this paper, we take a deep dive into KD losses and find that "using new class data for KD"not only hinders the model adaption (for learning new classes) but also results in low efficiency for preserving old class knowledge. We address this by "using the placebos of old classes for KD", where the placebos are chosen from a free image stream, such as Google Images, in an automatical and economical fashion. To this end, we train an online placebo selection policy to quickly evaluate the quality of streaming images (good or bad placebos) and use only good ones for one-time feed-forward computation of KD. We formulate the policy training process as an online Markov Decision Process (MDP), and introduce an online learning algorithm to solve this MDP problem without causing much computation costs. In experiments, we show that our method 1) is surprisingly effective even when there is no class overlap between placebos and original old class data, 2) does not require any additional supervision or memory budget, and 3) significantly outperforms a number of top-performing CIL methods, in particular when using lower memory budgets for old class exemplars, e.g., five exemplars per class. https://github.com/yaoyao-liu/online-placebos 2024-01-01T08:00:00Z text application/pdf https://ink.library.smu.edu.sg/sis_research/9207 info:doi/10.1109/WACV57701.2024.00222 https://ink.library.smu.edu.sg/context/sis_research/article/10212/viewcontent/Liu_Wakening_Past_Concepts_Without_Past_Data_Class_Incremental_Learning_From_Online_WACV_2024_paper.pdf http://creativecommons.org/licenses/by-nc-nd/4.0/ Research Collection School Of Computing and Information Systems eng Institutional Knowledge at Singapore Management University And algorithm Data class Deep dives Formulation Incremental learning Learning architectures Machine learning architecture Machine-learning Markov Decision Processes Model adaption Databases and Information Systems Theory and Algorithms
institution Singapore Management University
building SMU Libraries
continent Asia
country Singapore
Singapore
content_provider SMU Libraries
collection InK@SMU
language English
topic And algorithm
Data class
Deep dives
Formulation
Incremental learning
Learning architectures
Machine learning architecture
Machine-learning
Markov Decision Processes
Model adaption
Databases and Information Systems
Theory and Algorithms
spellingShingle And algorithm
Data class
Deep dives
Formulation
Incremental learning
Learning architectures
Machine learning architecture
Machine-learning
Markov Decision Processes
Model adaption
Databases and Information Systems
Theory and Algorithms
LIU, Yaoyao
LI, Yingying
SCHIELE, Bernt
SUN, Qianru
Wakening past concepts without past data: Class-incremental learning from online placebos
description Not forgetting old class knowledge is a key challenge for class-incremental learning (CIL) when the model continuously adapts to new classes. A common technique to address this is knowledge distillation (KD), which penalizes prediction inconsistencies between old and new models. Such prediction is made with almost new class data, as old class data is extremely scarce due to the strict memory limitation in CIL. In this paper, we take a deep dive into KD losses and find that "using new class data for KD"not only hinders the model adaption (for learning new classes) but also results in low efficiency for preserving old class knowledge. We address this by "using the placebos of old classes for KD", where the placebos are chosen from a free image stream, such as Google Images, in an automatical and economical fashion. To this end, we train an online placebo selection policy to quickly evaluate the quality of streaming images (good or bad placebos) and use only good ones for one-time feed-forward computation of KD. We formulate the policy training process as an online Markov Decision Process (MDP), and introduce an online learning algorithm to solve this MDP problem without causing much computation costs. In experiments, we show that our method 1) is surprisingly effective even when there is no class overlap between placebos and original old class data, 2) does not require any additional supervision or memory budget, and 3) significantly outperforms a number of top-performing CIL methods, in particular when using lower memory budgets for old class exemplars, e.g., five exemplars per class. https://github.com/yaoyao-liu/online-placebos
format text
author LIU, Yaoyao
LI, Yingying
SCHIELE, Bernt
SUN, Qianru
author_facet LIU, Yaoyao
LI, Yingying
SCHIELE, Bernt
SUN, Qianru
author_sort LIU, Yaoyao
title Wakening past concepts without past data: Class-incremental learning from online placebos
title_short Wakening past concepts without past data: Class-incremental learning from online placebos
title_full Wakening past concepts without past data: Class-incremental learning from online placebos
title_fullStr Wakening past concepts without past data: Class-incremental learning from online placebos
title_full_unstemmed Wakening past concepts without past data: Class-incremental learning from online placebos
title_sort wakening past concepts without past data: class-incremental learning from online placebos
publisher Institutional Knowledge at Singapore Management University
publishDate 2024
url https://ink.library.smu.edu.sg/sis_research/9207
https://ink.library.smu.edu.sg/context/sis_research/article/10212/viewcontent/Liu_Wakening_Past_Concepts_Without_Past_Data_Class_Incremental_Learning_From_Online_WACV_2024_paper.pdf
_version_ 1814047791279767552