Francis Bach - INRIA - ENS - PSL
Eugène Berta, co-advised with Michael Jordan Nabil Boukir, co-advised with Michael Jordan Sacha Braun, co-advised with Michael Jordan Bertille Follain Alexandre François, co-advised with Antonio Orvieto David Holzmüller Etienne Gauthier, co-advised with Michael Jordan Frederik Kunstner Marc Lambert, co-advised with Silvère Bonnabel Ivan Lerner, co-advised with Anita Burgun and Antoine Neuraz Simon
NIPS2010ã«ãŠã‘る発表論文ã«è¦‹ã‚‹ã€æ©Ÿæ¢°å¦ç¿’最å‰ç·š | gihyo.jp
ãªãŠã€åŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©ãƒ¼æ€§ã«ã¤ã„ã¦ã•ã‚‰ã«çŸ¥ã‚ŠãŸã„æ–¹ã¯ã€ãƒãƒ¥ãƒ¼ãƒˆãƒªã‚¢ãƒ«ï¼»3ï¼½ãŒå‚考ã«ãªã‚Šã¾ã™ã€‚ 昨年ã®NIPSã§ã®å‹•å‘ ãã‚Œã§ã¯ã€æ˜¨å¹´ã®NIPSã§ã®å‹•å‘を見ã¦ã¿ã¾ã—ょã†ã€‚ Bachï¼»4ï¼½ã¯ã€L∞ノルムãŒåŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©ãƒ¼é–¢æ•°ã®ãƒãƒ´ã‚¡ãƒ¼ã‚¹æ‹¡å¼µã‹ã‚‰å°Žå‡ºã§ãã‚‹ã“ã¨ã‚’示ã™ã“ã¨ã«ã‚ˆã‚Š, 劣モジュラー性ã¨ã‚¹ãƒ‘ース性ã¨ã®é–¢ä¿‚を示ã—ã¾ã—ãŸã€‚ã•ã‚‰ã«, ã“ã®æ´žå¯Ÿã‹ã‚‰æ•™å¸«ã‚ã‚Šå¦ç¿’ã§ç”¨ã„ã‚‹ã“ã¨ãŒã§ãã‚‹æ–°ã—ã„3ã¤ã®ãƒŽãƒ«ãƒ ã‚’æ案ã—ã¾ã—ãŸã€‚ã¾ãŸã€å‹¾é…法や近接法ãŒåŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©ãƒ¼é–¢æ•°æœ€é©åŒ–ã«ä½¿ãˆã‚‹ã“ã¨ã‚’示ã—, 実験ã«ã‚ˆã‚ŠL1,ã¨L2ノルムを用ã„るより精度ãŒè‰¯ã„ã“ã¨ã‚’示ã—ã¾ã—ãŸã€‚ Stobbe and Krauseï¼»5ï¼½ã¯ã€åŠ£ãƒ¢ã‚¸ãƒ¥ãƒ©ãƒ¼é–¢æ•°ã‚’凹関数ã®å’Œã¨ã—ã¦åˆ†è§£ã§ãã‚‹æ–°ã—ã„クラス（decomposable submodular function）を定義ã—, カットå•é¡Œ, ãƒžãƒ«ã‚³ãƒ•ç¢ºçŽ‡å ´ã®æœ€é©åŒ–, 集åˆè¢«è¦†å•é¡Œãªã©ãŒãã®æ–°ã—ã„クラスã®æœ€å°åŒ–å•
最長片é“ãã£ã·ã®çµŒè·¯ã‚’求ã‚ã‚‹
最長片é“ãã£ã·ã®çµŒè·¯ã‚’求ã‚ã‚‹ Index & Overview ã‚らã¾ã— ã“ã®æ–‡æ›¸ã¯ã€ï¼ªï¼²ã®æœ€é•·ç‰‡é“ãã£ã·ã®çµŒè·¯ã‚’〠整数計画法ã¨å…¨æŽ¢ç´¢ã®ï¼’ã¤ã®æ–¹æ³•ã§æ±‚ã‚ãŸéŽç¨‹ã‚’ã¾ã¨ã‚ãŸã‚‚ã®ã§ã™ã€‚ å‰è€…ã§ã¯åŽ³å¯†ã«ã€å¾Œè€…ã§ã¯ã‚„やイイカゲンã«ã€ãã®çµŒè·¯ã‚’求ã‚ã‚‹ã“ã¨ã«æˆåŠŸã—〠２ã¤ã®æ–¹æ³•ã§æ±‚ã‚ãŸçµŒè·¯ã¯ä¸€è‡´ã—ã¾ã—ãŸã€‚ トピックス NHK ã®ç´€è¡Œç•ªçµ„ã€Œåˆ—å³¶ç¸¦æ– é‰„é“12000kmã®æ—…ã€ã‚’ãã£ã‹ã‘ã«ã“ã® Web ページを探ã—当ã¦ãŸæ–¹ã¯ã€ã¾ãšã€Œä»˜éŒ²ï¼’（2004年３月版）ã€ã‚’ã”覧ãã ã•ã„。 ç¾çŠ¶ã®æœ€é•·ç‰‡é“ãã£ã·ã®çµŒè·¯ã‚„ã€ã‚ã‚Šãã†ãªè³ªå•ã‚’ã¾ã¨ã‚ã¦ã‚ã‚Šã¾ã™ã€‚ ãµã¨æ€ã„ç«‹ã£ã¦ã€2006年５月版ã®æœ€é•·ç‰‡é“ãã£ã·çµŒè·¯å›³ï¼ˆPDF å½¢å¼ã€35,414 bytes）を作りã¾ã—ãŸã€‚ 2004年３月版ã®åœ°å›³ã¨ã®ç›¸é•ç‚¹ã¯ãŸã １点〠「富山港線を削除ã—ãŸã€ã“ã¨ã§ã™ï¼ˆ2006年２月28日廃æ¢ï¼‰ã€‚ ã‚‚ã¨ã‚‚ã¨æœ€é•·çµŒè·¯ã«å«ã¾ã‚Œã¦ã„ãªã‹ã£ãŸè·¯ç·šãŒå»ƒæ¢ã«ãª
確率的勾é…é™ä¸‹æ³•ã«ã‚ˆã‚‹è¡Œåˆ—分解を試ã—ã¦ã¿ãŸ - ã®ã‚“ã³ã‚Šèªæ›¸æ—¥è¨˜
å‰ã€…回ã®NMF(Non-negative Matrix Factorization)ã«ç¶šã„ã¦è¡Œåˆ—分解ãƒã‚¿ã§ã™ã€‚言語処ç†å¦ä¼šå…¨å›½å¤§ä¼šã®ãƒãƒ¥ãƒ¼ãƒˆãƒªã‚¢ãƒ«ã€ŒæŽ¨è–¦ã‚·ã‚¹ãƒ†ãƒ -機械å¦ç¿’ã®è¦–点ã‹ã‚‰-ã€ã§ç´¹ä»‹ã•ã‚Œã¦ã„ãŸã€ç¢ºçŽ‡çš„勾é…é™ä¸‹æ³•ã«ã‚ˆã‚‹è¡Œåˆ—分解を試ã—ã¦ã¿ã¾ã—ãŸã€‚ãƒãƒ¥ãƒ¼ãƒˆãƒªã‚¢ãƒ«ã®è³‡æ–™ã¯å…¬é–‹ã•ã‚Œã¦ã„ãªã„よã†ã§ã—ãŸã®ã§ã€å…ƒè«–æ–‡ã®æ–¹ã®ãƒªãƒ³ã‚¯ã‚’å¼µã£ã¦ãŠãã¾ã™ã€‚実際ã«ã¯åŒã˜è‘—者ã®åˆ¥ã®è«–文を引用ã•ã‚Œã¦ã¾ã—ãŸãŒã€åƒ•ã«ã¯ä¸‹ã®è«–æ–‡ã®æ–¹ãŒåˆ†ã‹ã‚Šã‚„ã™ã‹ã£ãŸã®ã§ã“ã£ã¡ã§ã€‚ MATRIX FACTORIZATION TECHNIQUES FOR RECOMMENDER SYSTEMS, Yehuda Koren, Rovert Bell, Chris Volinsky, IEEE Computer, Volume 42, Issue 8, p.30-37, 2009 作æˆã—ãŸã‚³ãƒ¼ãƒ‰ã¯ä»¥ä¸‹ã«ç½®ã„ã¦ã‚ã‚Šã¾ã™ã€‚行列演算ã«Eigenã‚’
CVX: Matlab Software for Disciplined Convex Programming | CVX Research, Inc.
Version 2.2, January 2020, Build 1148 New: Professor Stephen Boyd recently recorded a video introduction to CVX for Stanford’s convex optimization courses. Click here to watch it. CVX 3.0 beta: We’ve added some interesting new features for users and system administrators. Give it a try! CVX is a Matlab-based modeling system for convex optimization. CVX turns Matlab into a modeling language, allowi
singular value thresholding
Singular Value Thresholding (SVT) is an algorithm to minimize the nuclear norm of a matrix, subject to certain types of constraints. It has been successfully used in many matrix-completion problems (for more on the matrix completion problem, see Exact matrix completion via convex optimization by E.J. Candès and B. Recht). The SVT algorithm is described in the paper A singular value thresholding al
SPGL1 : Home Page
This page has moved. You should have been automatically redirected to its new location. If you aren't forwarded to the new page, click here.
機械å¦ç¿’è«– 2009年度
æ‹…å½“ï¼šå’Œç”°å±±æ£ è¬›ç¾©ã®æ¦‚è¦ å¤šãã®çµ±è¨ˆçš„機械å¦ç¿’ã®æ‰‹æ³•ã¯é€£ç¶šæœ€é©åŒ–ã‚’ãã®åŸºç¤Žã¨ã—ã¦ã„る。特ã«å‡¸æœ€é©åŒ–（凸計画法）ã¯è¿‘å¹´ã®çµ±è¨ˆçš„機械å¦ç¿’ã«ãŠã„ã¦å¿…é ˆã®çŸ¥è˜ã¨ãªã‚Šã¤ã¤ã‚る。凸最é©åŒ–ã¯ã€æ©Ÿæ¢°å¦ç¿’ã®ã¿ã«ã¨ã©ã¾ã‚‰ãšä¿¡å·å‡¦ç†ã€æƒ…å ±ä¼é€ã€ç”»åƒå‡¦ç†ã€åˆ¶å¾¡å·¥å¦ãªã©æ§˜ã€…ãªåˆ†é‡Žã§æ´»èºã®å ´ã‚’広ã’ã¤ã¤ã‚る。本講義ã§ã¯ã€å‡¸æœ€é©åŒ–ã®åŸºç¤Žã‚’押ã•ãˆãŸä¸Šã§çµ±è¨ˆçš„機械å¦ç¿’ã«é–¢é€£ã™ã‚‹è©±é¡Œã‚’ã„ãã¤ã‹å¦ã¶ã€‚ 教科書ã¯åˆ©ç”¨ã—ãªã„ãŒS. Boyd and L. Vandenberghe,``Convex optimization''ã«æ²¿ã£ã¦è¬›ç¾©ã‚’進ã‚る。ã“ã®æœ¬ã®PDFファイル(出版社ã®è¨±å¯ã®ã‚‚ã¨ã«å…¬é–‹ã•ã‚Œã¦ã„ã‚‹)を手元ã«ã‚³ãƒ”ーã—ã¦ãŠãã€å‚考ã«ã™ã‚‹ã“ã¨ã‚’勧ã‚る。 ã¾ãŸã€å‚考プãƒã‚°ãƒ©ãƒ を実行ã€ãƒ‘ラメータ変更ã€æ”¹é€ ãªã©ã—ã¦ã¿ã¦å‹•ã‹ã—ã¦"éŠã‚“ã§ã¿ã‚‹"ã“ã¨ãŒç†è§£ã‚’æ·±ã‚ã‚‹ãŸã‚ã«éžå¸¸ã«æœ‰ç›Šã§ã‚る。 評価ã«ã¤ã„㦠期末試験（８ï¼ï¼…〜９ï¼ï¼…）＋数回ã®ãƒ¬ãƒ
Tim Kelley
(919) 515-7163 (office), (919) 515-3798 (FAX) My current research interests are in linear/nonlinear equations, mixed-precision algorithms, neutron transport problems, and computational quantum chemistry and physics. In the past I worked on multilevel methods for integral equations, quasi-Newton methods, semiconductor modeling, optimal control, optimization of noisy functions, and flow in porous me
OOQP: Object-Oriented Software for Quadratic Programming
OOQP is an object-oriented C++ package, based on a primal-dual interior-point method, for solving convex quadratic programming problems (QPs). It contains code that can be used "out of the box" to solve a variety of structured QPs, including general sparse QPs, QPs arising from support vector machines, Huber regression problems, and QPs with bound constraints. OOQP also can be used as a framework
Welcome - OpenOpt
You may be interested in our amazing software, alternative to commercial frameworks with obsolete and/or banausic programming constructs. It is completely free (license: BSD) and cross-platform (Linux, Windows, Mac etc) Python language modules. The software is published quarterly since 2007 and already has some essential applications. We expect it to become even more popular when NumPy will got fu
ベルマンフォードã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã§å®Ÿè¡Œã•ã‚Œã‚‹çµæžœã‚‚é€æ¬¡è¡¨ç¤º - yasuhisa's blog
離散最é©åŒ–ç†è«–ã®èª²é¡ŒãŒå‡ºã¦ã„ãŸã®ã§ã€ãƒ™ãƒ«ãƒžãƒ³ãƒ•ã‚©ãƒ¼ãƒ‰ã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ を実装ã—ã¦ã¿ã‚‹ã“ã¨ã«ã—ãŸã€‚アルゴリズムãŒå®Ÿè¡Œã•ã‚Œã¦ã„ã様åã®ä¾‹ã‚‚レãƒãƒ¼ãƒˆã«è²¼ã‚ã†ã¨æ€ã£ãŸã‚“ã ã‘ã©ã€ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¯ã‚‚ã†ã‚ã‚‹ã‚“ã ã‹ã‚‰ã€ãã®æ§˜åã‚’ruby-graphvizã¨ã‹ã§åã‘ã°ã„ã„ã˜ã‚ƒã‚“ã¨ã„ã†ã“ã¨ã§ã‚„ã£ã¦ã¿ãŸã€‚ pngファイルをアニメーションgifã«å¤‰æ›ã™ã‚‹ã®ã¯ã“ã‚“ãªæ„Ÿã˜ã§ã€‚ã“ã®è¾ºã‚’å‚考ã«ã€‚ convert -geometry 320x500! -delay 150 -loop 0 bellman_ford_example_a_uniq*.png bellman_ford_example_a.gif 俺ã¯Rubyã§æ›¸ã„ãŸã‚ã‘ã ã‘ã©ã€ã“ã‚“ãªã“ã¨ã‚’ã‚„ã£ã¦ã‚‹id:mickey24ã«ã€Œãã‚ŒRã§ã§ãるよ!!ã€ã¨è¨€ã‚ã‚Œãã†ã§ã‚る。 å˜ä¸€å§‹ç‚¹æœ€çŸè·¯å•é¡Œã«å¯¾ã™ã‚‹ãã®ä»–ã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ベルマンフォードã®ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã¯å˜ä¸€å§‹ç‚¹æœ€çŸè·¯å•é¡Œã«å¯¾ã™ã‚‹
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
http://www.ibis.t.u-tokyo.ac.jp/RyotaTomioka/Notes/DerivingDual
http://www.ece.ucsb.edu/~roy/cgi-bin/makepage.pl?nav=course_271a
http://www.cs.caltech.edu/~krausea/publications.html
http://www.cs.caltech.edu/~krausea/sfo/
http://www.ann.jussieu.fr/~plc/
http://www.stanford.edu/~boyd/ee263/oldvideos.html
http://www.acm.caltech.edu/~nesta/
MATLAB mfiles for Working Analysis
{{#tags}}- {{label}}
{{/tags}}