A Novel Stochastic Gradient Descent Algorithm for Learning Principal Subspaces
Many machine learning problems encode their data as a matrix with a possibly very large number of rows and columns. In several applications like neuroscience, image compression or deep reinforcement learning, the principal subspace of such a matrix provides a useful, low-dimensional representation o...
Gespeichert in:
Veröffentlicht in: | arXiv.org 2022-12 |
---|---|
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 | Charline Le Lan Greaves, Joshua Farebrother, Jesse Rowland, Mark Pedregosa, Fabian Agarwal, Rishabh Bellemare, Marc G |
description | Many machine learning problems encode their data as a matrix with a possibly very large number of rows and columns. In several applications like neuroscience, image compression or deep reinforcement learning, the principal subspace of such a matrix provides a useful, low-dimensional representation of individual data. Here, we are interested in determining the \(d\)-dimensional principal subspace of a given matrix from sample entries, i.e. from small random submatrices. Although a number of sample-based methods exist for this problem (e.g. Oja's rule \citep{oja1982simplified}), these assume access to full columns of the matrix or particular matrix structure such as symmetry and cannot be combined as-is with neural networks \citep{baldi1989neural}. In this paper, we derive an algorithm that learns a principal subspace from sample entries, can be applied when the approximate subspace is represented by a neural network, and hence can be scaled to datasets with an effectively infinite number of rows and columns. Our method consists in defining a loss function whose minimizer is the desired principal subspace, and constructing a gradient estimate of this loss whose bias can be controlled. We complement our theoretical analysis with a series of experiments on synthetic matrices, the MNIST dataset \citep{lecun2010mnist} and the reinforcement learning domain PuddleWorld \citep{sutton1995generalization} demonstrating the usefulness of our approach. |
format | Article |
fullrecord | <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2748627045</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2748627045</sourcerecordid><originalsourceid>FETCH-proquest_journals_27486270453</originalsourceid><addsrcrecordid>eNqNi0sKwjAUAIMgWLR3CLguxKS_bfG7kCLovsT42qbUpOalnt8KHsDVLGZmRgIuxCbKY84XJETsGGM8zXiSiICUBS3tG3p69Va1Er1W9OjkQ4PxdAeoviz6xjrt2yetraNnkM5o09CL00bpQU7zeMdBKsAVmdeyRwh_XJL1YX_bnqLB2dcI6KvOjs5MquJZnKc8Y3Ei_qs-prA-BA</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2748627045</pqid></control><display><type>article</type><title>A Novel Stochastic Gradient Descent Algorithm for Learning Principal Subspaces</title><source>Free E- Journals</source><creator>Charline Le Lan ; Greaves, Joshua ; Farebrother, Jesse ; Rowland, Mark ; Pedregosa, Fabian ; Agarwal, Rishabh ; Bellemare, Marc G</creator><creatorcontrib>Charline Le Lan ; Greaves, Joshua ; Farebrother, Jesse ; Rowland, Mark ; Pedregosa, Fabian ; Agarwal, Rishabh ; Bellemare, Marc G</creatorcontrib><description>Many machine learning problems encode their data as a matrix with a possibly very large number of rows and columns. In several applications like neuroscience, image compression or deep reinforcement learning, the principal subspace of such a matrix provides a useful, low-dimensional representation of individual data. Here, we are interested in determining the \(d\)-dimensional principal subspace of a given matrix from sample entries, i.e. from small random submatrices. Although a number of sample-based methods exist for this problem (e.g. Oja's rule \citep{oja1982simplified}), these assume access to full columns of the matrix or particular matrix structure such as symmetry and cannot be combined as-is with neural networks \citep{baldi1989neural}. In this paper, we derive an algorithm that learns a principal subspace from sample entries, can be applied when the approximate subspace is represented by a neural network, and hence can be scaled to datasets with an effectively infinite number of rows and columns. Our method consists in defining a loss function whose minimizer is the desired principal subspace, and constructing a gradient estimate of this loss whose bias can be controlled. We complement our theoretical analysis with a series of experiments on synthetic matrices, the MNIST dataset \citep{lecun2010mnist} and the reinforcement learning domain PuddleWorld \citep{sutton1995generalization} demonstrating the usefulness of our approach.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Algorithms ; Datasets ; Deep learning ; Image compression ; Machine learning ; Mathematical analysis ; Matrices (mathematics) ; Neural networks ; Subspaces</subject><ispartof>arXiv.org, 2022-12</ispartof><rights>2022. This work is published under http://creativecommons.org/licenses/by/4.0/ (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>777,781</link.rule.ids></links><search><creatorcontrib>Charline Le Lan</creatorcontrib><creatorcontrib>Greaves, Joshua</creatorcontrib><creatorcontrib>Farebrother, Jesse</creatorcontrib><creatorcontrib>Rowland, Mark</creatorcontrib><creatorcontrib>Pedregosa, Fabian</creatorcontrib><creatorcontrib>Agarwal, Rishabh</creatorcontrib><creatorcontrib>Bellemare, Marc G</creatorcontrib><title>A Novel Stochastic Gradient Descent Algorithm for Learning Principal Subspaces</title><title>arXiv.org</title><description>Many machine learning problems encode their data as a matrix with a possibly very large number of rows and columns. In several applications like neuroscience, image compression or deep reinforcement learning, the principal subspace of such a matrix provides a useful, low-dimensional representation of individual data. Here, we are interested in determining the \(d\)-dimensional principal subspace of a given matrix from sample entries, i.e. from small random submatrices. Although a number of sample-based methods exist for this problem (e.g. Oja's rule \citep{oja1982simplified}), these assume access to full columns of the matrix or particular matrix structure such as symmetry and cannot be combined as-is with neural networks \citep{baldi1989neural}. In this paper, we derive an algorithm that learns a principal subspace from sample entries, can be applied when the approximate subspace is represented by a neural network, and hence can be scaled to datasets with an effectively infinite number of rows and columns. Our method consists in defining a loss function whose minimizer is the desired principal subspace, and constructing a gradient estimate of this loss whose bias can be controlled. We complement our theoretical analysis with a series of experiments on synthetic matrices, the MNIST dataset \citep{lecun2010mnist} and the reinforcement learning domain PuddleWorld \citep{sutton1995generalization} demonstrating the usefulness of our approach.</description><subject>Algorithms</subject><subject>Datasets</subject><subject>Deep learning</subject><subject>Image compression</subject><subject>Machine learning</subject><subject>Mathematical analysis</subject><subject>Matrices (mathematics)</subject><subject>Neural networks</subject><subject>Subspaces</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNi0sKwjAUAIMgWLR3CLguxKS_bfG7kCLovsT42qbUpOalnt8KHsDVLGZmRgIuxCbKY84XJETsGGM8zXiSiICUBS3tG3p69Va1Er1W9OjkQ4PxdAeoviz6xjrt2yetraNnkM5o09CL00bpQU7zeMdBKsAVmdeyRwh_XJL1YX_bnqLB2dcI6KvOjs5MquJZnKc8Y3Ei_qs-prA-BA</recordid><startdate>20221208</startdate><enddate>20221208</enddate><creator>Charline Le Lan</creator><creator>Greaves, Joshua</creator><creator>Farebrother, Jesse</creator><creator>Rowland, Mark</creator><creator>Pedregosa, Fabian</creator><creator>Agarwal, Rishabh</creator><creator>Bellemare, Marc G</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>20221208</creationdate><title>A Novel Stochastic Gradient Descent Algorithm for Learning Principal Subspaces</title><author>Charline Le Lan ; Greaves, Joshua ; Farebrother, Jesse ; Rowland, Mark ; Pedregosa, Fabian ; Agarwal, Rishabh ; Bellemare, Marc G</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_27486270453</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Algorithms</topic><topic>Datasets</topic><topic>Deep learning</topic><topic>Image compression</topic><topic>Machine learning</topic><topic>Mathematical analysis</topic><topic>Matrices (mathematics)</topic><topic>Neural networks</topic><topic>Subspaces</topic><toplevel>online_resources</toplevel><creatorcontrib>Charline Le Lan</creatorcontrib><creatorcontrib>Greaves, Joshua</creatorcontrib><creatorcontrib>Farebrother, Jesse</creatorcontrib><creatorcontrib>Rowland, Mark</creatorcontrib><creatorcontrib>Pedregosa, Fabian</creatorcontrib><creatorcontrib>Agarwal, Rishabh</creatorcontrib><creatorcontrib>Bellemare, Marc G</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</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>Charline Le Lan</au><au>Greaves, Joshua</au><au>Farebrother, Jesse</au><au>Rowland, Mark</au><au>Pedregosa, Fabian</au><au>Agarwal, Rishabh</au><au>Bellemare, Marc G</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>A Novel Stochastic Gradient Descent Algorithm for Learning Principal Subspaces</atitle><jtitle>arXiv.org</jtitle><date>2022-12-08</date><risdate>2022</risdate><eissn>2331-8422</eissn><abstract>Many machine learning problems encode their data as a matrix with a possibly very large number of rows and columns. In several applications like neuroscience, image compression or deep reinforcement learning, the principal subspace of such a matrix provides a useful, low-dimensional representation of individual data. Here, we are interested in determining the \(d\)-dimensional principal subspace of a given matrix from sample entries, i.e. from small random submatrices. Although a number of sample-based methods exist for this problem (e.g. Oja's rule \citep{oja1982simplified}), these assume access to full columns of the matrix or particular matrix structure such as symmetry and cannot be combined as-is with neural networks \citep{baldi1989neural}. In this paper, we derive an algorithm that learns a principal subspace from sample entries, can be applied when the approximate subspace is represented by a neural network, and hence can be scaled to datasets with an effectively infinite number of rows and columns. Our method consists in defining a loss function whose minimizer is the desired principal subspace, and constructing a gradient estimate of this loss whose bias can be controlled. We complement our theoretical analysis with a series of experiments on synthetic matrices, the MNIST dataset \citep{lecun2010mnist} and the reinforcement learning domain PuddleWorld \citep{sutton1995generalization} demonstrating the usefulness of our approach.</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, 2022-12 |
issn | 2331-8422 |
language | eng |
recordid | cdi_proquest_journals_2748627045 |
source | Free E- Journals |
subjects | Algorithms Datasets Deep learning Image compression Machine learning Mathematical analysis Matrices (mathematics) Neural networks Subspaces |
title | A Novel Stochastic Gradient Descent Algorithm for Learning Principal Subspaces |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-17T10%3A29%3A51IST&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=A%20Novel%20Stochastic%20Gradient%20Descent%20Algorithm%20for%20Learning%20Principal%20Subspaces&rft.jtitle=arXiv.org&rft.au=Charline%20Le%20Lan&rft.date=2022-12-08&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2748627045%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2748627045&rft_id=info:pmid/&rfr_iscdi=true |