Harmonizing knowledge Transfer in Neural Network with Unified Distillation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Huang, Yaomin, Yan, Zaomin, Shen, Chaomin, Fang, Faming, Zhang, Guixu
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916412545564672
author Huang, Yaomin
Yan, Zaomin
Shen, Chaomin
Fang, Faming
Zhang, Guixu
author_facet Huang, Yaomin
Yan, Zaomin
Shen, Chaomin
Fang, Faming
Zhang, Guixu
contents Knowledge distillation (KD), known for its ability to transfer knowledge from a cumbersome network (teacher) to a lightweight one (student) without altering the architecture, has been garnering increasing attention. Two primary categories emerge within KD methods: feature-based, focusing on intermediate layers' features, and logits-based, targeting the final layer's logits. This paper introduces a novel perspective by leveraging diverse knowledge sources within a unified KD framework. Specifically, we aggregate features from intermediate layers into a comprehensive representation, effectively gathering semantic information from different stages and scales. Subsequently, we predict the distribution parameters from this representation. These steps transform knowledge from the intermediate layers into corresponding distributive forms, thereby allowing for knowledge distillation through a unified distribution constraint at different stages of the network, ensuring the comprehensiveness and coherence of knowledge transfer. Numerous experiments were conducted to validate the effectiveness of the proposed method.
format Preprint
id arxiv_https___arxiv_org_abs_2409_18565
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Harmonizing knowledge Transfer in Neural Network with Unified Distillation
Huang, Yaomin
Yan, Zaomin
Shen, Chaomin
Fang, Faming
Zhang, Guixu
Computer Vision and Pattern Recognition
Knowledge distillation (KD), known for its ability to transfer knowledge from a cumbersome network (teacher) to a lightweight one (student) without altering the architecture, has been garnering increasing attention. Two primary categories emerge within KD methods: feature-based, focusing on intermediate layers' features, and logits-based, targeting the final layer's logits. This paper introduces a novel perspective by leveraging diverse knowledge sources within a unified KD framework. Specifically, we aggregate features from intermediate layers into a comprehensive representation, effectively gathering semantic information from different stages and scales. Subsequently, we predict the distribution parameters from this representation. These steps transform knowledge from the intermediate layers into corresponding distributive forms, thereby allowing for knowledge distillation through a unified distribution constraint at different stages of the network, ensuring the comprehensiveness and coherence of knowledge transfer. Numerous experiments were conducted to validate the effectiveness of the proposed method.
title Harmonizing knowledge Transfer in Neural Network with Unified Distillation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2409.18565