Natural Statistics of Network Activations and Implications for Knowledge Distillation
In a matter that is analog to the study of natural image statistics, we study the natural statistics of the deep neural network activations at various layers. As we show, these statistics, similar to image statistics, follow a power law. We also show, both analytically and empirically, that with dep...
Gespeichert in:
Veröffentlicht in: | arXiv.org 2021-06 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | arXiv.org |
container_volume | |
creator | Rotman, Michael Wolf, Lior |
description | In a matter that is analog to the study of natural image statistics, we study the natural statistics of the deep neural network activations at various layers. As we show, these statistics, similar to image statistics, follow a power law. We also show, both analytically and empirically, that with depth the exponent of this power law increases at a linear rate. As a direct implication of our discoveries, we present a method for performing Knowledge Distillation (KD). While classical KD methods consider the logits of the teacher network, more recent methods obtain a leap in performance by considering the activation maps. This, however, uses metrics that are suitable for comparing images. We propose to employ two additional loss terms that are based on the spectral properties of the intermediate activation maps. The proposed method obtains state of the art results on multiple image recognition KD benchmarks. |
format | Article |
fullrecord | <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2536123415</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2536123415</sourcerecordid><originalsourceid>FETCH-proquest_journals_25361234153</originalsourceid><addsrcrecordid>eNqNy98KgjAYBfARBEn5Dh90LejmrNvoD0XgTXUtQ2fM1mbbZ75-Fj5AVwfO75wJCShjSbROKZ2R0PsmjmOarSjnLCC3XGDnhIYLClQeVenB1pBL7K17wKZE9R7AGg_CVHB6tlqVY1FbB2djey2ru4Td9631zxZkWgvtZTjmnCwP--v2GLXOvjrpsWhs58xABeUsSyhLE87-W30AG5dBdg</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2536123415</pqid></control><display><type>article</type><title>Natural Statistics of Network Activations and Implications for Knowledge Distillation</title><source>Free E- Journals</source><creator>Rotman, Michael ; Wolf, Lior</creator><creatorcontrib>Rotman, Michael ; Wolf, Lior</creatorcontrib><description>In a matter that is analog to the study of natural image statistics, we study the natural statistics of the deep neural network activations at various layers. As we show, these statistics, similar to image statistics, follow a power law. We also show, both analytically and empirically, that with depth the exponent of this power law increases at a linear rate. As a direct implication of our discoveries, we present a method for performing Knowledge Distillation (KD). While classical KD methods consider the logits of the teacher network, more recent methods obtain a leap in performance by considering the activation maps. This, however, uses metrics that are suitable for comparing images. We propose to employ two additional loss terms that are based on the spectral properties of the intermediate activation maps. The proposed method obtains state of the art results on multiple image recognition KD benchmarks.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Artificial neural networks ; Distillation ; Object recognition ; Power law ; Statistics</subject><ispartof>arXiv.org, 2021-06</ispartof><rights>2021. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>780,784</link.rule.ids></links><search><creatorcontrib>Rotman, Michael</creatorcontrib><creatorcontrib>Wolf, Lior</creatorcontrib><title>Natural Statistics of Network Activations and Implications for Knowledge Distillation</title><title>arXiv.org</title><description>In a matter that is analog to the study of natural image statistics, we study the natural statistics of the deep neural network activations at various layers. As we show, these statistics, similar to image statistics, follow a power law. We also show, both analytically and empirically, that with depth the exponent of this power law increases at a linear rate. As a direct implication of our discoveries, we present a method for performing Knowledge Distillation (KD). While classical KD methods consider the logits of the teacher network, more recent methods obtain a leap in performance by considering the activation maps. This, however, uses metrics that are suitable for comparing images. We propose to employ two additional loss terms that are based on the spectral properties of the intermediate activation maps. The proposed method obtains state of the art results on multiple image recognition KD benchmarks.</description><subject>Artificial neural networks</subject><subject>Distillation</subject><subject>Object recognition</subject><subject>Power law</subject><subject>Statistics</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNy98KgjAYBfARBEn5Dh90LejmrNvoD0XgTXUtQ2fM1mbbZ75-Fj5AVwfO75wJCShjSbROKZ2R0PsmjmOarSjnLCC3XGDnhIYLClQeVenB1pBL7K17wKZE9R7AGg_CVHB6tlqVY1FbB2djey2ru4Td9631zxZkWgvtZTjmnCwP--v2GLXOvjrpsWhs58xABeUsSyhLE87-W30AG5dBdg</recordid><startdate>20210601</startdate><enddate>20210601</enddate><creator>Rotman, Michael</creator><creator>Wolf, Lior</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20210601</creationdate><title>Natural Statistics of Network Activations and Implications for Knowledge Distillation</title><author>Rotman, Michael ; Wolf, Lior</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_25361234153</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Artificial neural networks</topic><topic>Distillation</topic><topic>Object recognition</topic><topic>Power law</topic><topic>Statistics</topic><toplevel>online_resources</toplevel><creatorcontrib>Rotman, Michael</creatorcontrib><creatorcontrib>Wolf, Lior</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni)</collection><collection>ProQuest Central</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database (Proquest) (PQ_SDU_P3)</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Rotman, Michael</au><au>Wolf, Lior</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Natural Statistics of Network Activations and Implications for Knowledge Distillation</atitle><jtitle>arXiv.org</jtitle><date>2021-06-01</date><risdate>2021</risdate><eissn>2331-8422</eissn><abstract>In a matter that is analog to the study of natural image statistics, we study the natural statistics of the deep neural network activations at various layers. As we show, these statistics, similar to image statistics, follow a power law. We also show, both analytically and empirically, that with depth the exponent of this power law increases at a linear rate. As a direct implication of our discoveries, we present a method for performing Knowledge Distillation (KD). While classical KD methods consider the logits of the teacher network, more recent methods obtain a leap in performance by considering the activation maps. This, however, uses metrics that are suitable for comparing images. We propose to employ two additional loss terms that are based on the spectral properties of the intermediate activation maps. The proposed method obtains state of the art results on multiple image recognition KD benchmarks.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | EISSN: 2331-8422 |
ispartof | arXiv.org, 2021-06 |
issn | 2331-8422 |
language | eng |
recordid | cdi_proquest_journals_2536123415 |
source | Free E- Journals |
subjects | Artificial neural networks Distillation Object recognition Power law Statistics |
title | Natural Statistics of Network Activations and Implications for Knowledge Distillation |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-01T08%3A36%3A37IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Natural%20Statistics%20of%20Network%20Activations%20and%20Implications%20for%20Knowledge%20Distillation&rft.jtitle=arXiv.org&rft.au=Rotman,%20Michael&rft.date=2021-06-01&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2536123415%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2536123415&rft_id=info:pmid/&rfr_iscdi=true |