Exploring Feature-based Knowledge Distillation for Recommender System: A Frequency Perspective

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhu, Zhangchi, Zhang, Wei
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917890748317696
author Zhu, Zhangchi
Zhang, Wei
author_facet Zhu, Zhangchi
Zhang, Wei
contents In this paper, we analyze the feature-based knowledge distillation for recommendation from the frequency perspective. By defining knowledge as different frequency components of the features, we theoretically demonstrate that regular feature-based knowledge distillation is equivalent to equally minimizing losses on all knowledge and further analyze how this equal loss weight allocation method leads to important knowledge being overlooked. In light of this, we propose to emphasize important knowledge by redistributing knowledge weights. Furthermore, we propose FreqD, a lightweight knowledge reweighting method, to avoid the computational cost of calculating losses on each knowledge. Extensive experiments demonstrate that FreqD consistently and significantly outperforms state-of-the-art knowledge distillation methods for recommender systems. Our code is available at https://github.com/woriazzc/KDs.
format Preprint
id arxiv_https___arxiv_org_abs_2411_10676
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Exploring Feature-based Knowledge Distillation for Recommender System: A Frequency Perspective
Zhu, Zhangchi
Zhang, Wei
Information Retrieval
Artificial Intelligence
Machine Learning
In this paper, we analyze the feature-based knowledge distillation for recommendation from the frequency perspective. By defining knowledge as different frequency components of the features, we theoretically demonstrate that regular feature-based knowledge distillation is equivalent to equally minimizing losses on all knowledge and further analyze how this equal loss weight allocation method leads to important knowledge being overlooked. In light of this, we propose to emphasize important knowledge by redistributing knowledge weights. Furthermore, we propose FreqD, a lightweight knowledge reweighting method, to avoid the computational cost of calculating losses on each knowledge. Extensive experiments demonstrate that FreqD consistently and significantly outperforms state-of-the-art knowledge distillation methods for recommender systems. Our code is available at https://github.com/woriazzc/KDs.
title Exploring Feature-based Knowledge Distillation for Recommender System: A Frequency Perspective
topic Information Retrieval
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2411.10676