Neural Weight Search for Scalable Task Incremental Learning

Jian Jiang; Oya Celiktutan

Neural Weight Search for Scalable Task Incremental Learning

Jian Jiang, Oya Celiktutan

Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

123 Downloads (Pure)

Abstract

Task incremental learning aims to enable a system to maintain its performance on previously learned tasks while learning new tasks, solving the problem of catastrophic forgetting. One promising approach is to build an indi- vidual network or sub-network for future tasks. However, this leads to an ever-growing memory due to saving extra weights for new tasks and how to address this issue has re- mained an open problem in task incremental learning. In this paper, we introduce a novel Neural Weight Search technique that designs a fixed search space where the optimal combinations of frozen weights can be searched to build new models for novel tasks in an end-to-end manner, result- ing in a scalable and controllable memory growth. Exten- sive experiments on two benchmarks, i.e., Split-CIFAR-100 and CUB-to-Sketches, show our method achieves state-of- the-art performance with respect to both average inference accuracy and total memory cost.

Original language	English
Title of host publication	IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
Publisher	IEEE
Publication status	Accepted/In press - 2023

Access to Document

NWS_author_accepted_versionAccepted author manuscript, 0.98 MB

Cite this

@inbook{e3273e47c69643d2b8f38dcc269472da,

title = "Neural Weight Search for Scalable Task Incremental Learning",

abstract = "Task incremental learning aims to enable a system to maintain its performance on previously learned tasks while learning new tasks, solving the problem of catastrophic forgetting. One promising approach is to build an indi- vidual network or sub-network for future tasks. However, this leads to an ever-growing memory due to saving extra weights for new tasks and how to address this issue has re- mained an open problem in task incremental learning. In this paper, we introduce a novel Neural Weight Search technique that designs a fixed search space where the optimal combinations of frozen weights can be searched to build new models for novel tasks in an end-to-end manner, result- ing in a scalable and controllable memory growth. Exten- sive experiments on two benchmarks, i.e., Split-CIFAR-100 and CUB-to-Sketches, show our method achieves state-of- the-art performance with respect to both average inference accuracy and total memory cost.",

author = "Jian Jiang and Oya Celiktutan",

year = "2023",

language = "English",

booktitle = "IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)",

publisher = "IEEE",

}

TY - CHAP

T1 - Neural Weight Search for Scalable Task Incremental Learning

AU - Jiang, Jian

AU - Celiktutan, Oya

PY - 2023

Y1 - 2023

N2 - Task incremental learning aims to enable a system to maintain its performance on previously learned tasks while learning new tasks, solving the problem of catastrophic forgetting. One promising approach is to build an indi- vidual network or sub-network for future tasks. However, this leads to an ever-growing memory due to saving extra weights for new tasks and how to address this issue has re- mained an open problem in task incremental learning. In this paper, we introduce a novel Neural Weight Search technique that designs a fixed search space where the optimal combinations of frozen weights can be searched to build new models for novel tasks in an end-to-end manner, result- ing in a scalable and controllable memory growth. Exten- sive experiments on two benchmarks, i.e., Split-CIFAR-100 and CUB-to-Sketches, show our method achieves state-of- the-art performance with respect to both average inference accuracy and total memory cost.

AB - Task incremental learning aims to enable a system to maintain its performance on previously learned tasks while learning new tasks, solving the problem of catastrophic forgetting. One promising approach is to build an indi- vidual network or sub-network for future tasks. However, this leads to an ever-growing memory due to saving extra weights for new tasks and how to address this issue has re- mained an open problem in task incremental learning. In this paper, we introduce a novel Neural Weight Search technique that designs a fixed search space where the optimal combinations of frozen weights can be searched to build new models for novel tasks in an end-to-end manner, result- ing in a scalable and controllable memory growth. Exten- sive experiments on two benchmarks, i.e., Split-CIFAR-100 and CUB-to-Sketches, show our method achieves state-of- the-art performance with respect to both average inference accuracy and total memory cost.

M3 - Conference paper

BT - IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

PB - IEEE

ER -

Neural Weight Search for Scalable Task Incremental Learning

Abstract

Access to Document

Fingerprint

Cite this