hunch_net hunch_net-2010 hunch_net-2010-403 knowledge-graph by maker-knowledge-mining
Source: html
Introduction: The papers which interested me most at ICML and COLT 2010 were: Thomas Walsh , Kaushik Subramanian , Michael Littman and Carlos Diuk Generalizing Apprenticeship Learning across Hypothesis Classes . This paper formalizes and provides algorithms with guarantees for mixed-mode apprenticeship and traditional reinforcement learning algorithms, allowing RL algorithms that perform better than for either setting alone. István Szita and Csaba Szepesvári Model-based reinforcement learning with nearly tight exploration complexity bounds . This paper and another represent the frontier of best-known algorithm for Reinforcement Learning in a Markov Decision Process. James Martens Deep learning via Hessian-free optimization . About a new not-quite-online second order gradient algorithm for learning deep functional structures. Potentially this is very powerful because while people have often talked about end-to-end learning, it has rarely worked in practice. Chrisoph
sentIndex sentText sentNum sentScore
1 The papers which interested me most at ICML and COLT 2010 were: Thomas Walsh , Kaushik Subramanian , Michael Littman and Carlos Diuk Generalizing Apprenticeship Learning across Hypothesis Classes . [sent-1, score-0.087]
2 This paper formalizes and provides algorithms with guarantees for mixed-mode apprenticeship and traditional reinforcement learning algorithms, allowing RL algorithms that perform better than for either setting alone. [sent-2, score-0.575]
3 István Szita and Csaba Szepesvári Model-based reinforcement learning with nearly tight exploration complexity bounds . [sent-3, score-0.123]
4 This paper and another represent the frontier of best-known algorithm for Reinforcement Learning in a Markov Decision Process. [sent-4, score-0.184]
5 James Martens Deep learning via Hessian-free optimization . [sent-5, score-0.162]
6 Brendan McMahan and Matthew Streeter Adaptive Bound Optimization for Online Convex Optimization and the almost-same paper John Duchi , Elad Hazan , and Yoram Singer , Adaptive Subgradient Methods for Online Learning and Stochastic Optimization . [sent-13, score-0.1]
7 These papers provide tractable online algorithms with regret guarantees over a family of metrics rather than just euclidean metrics. [sent-14, score-0.267]
8 The program chairs for ICML did a wide-ranging survey over participants. [sent-17, score-0.119]
9 The results seem to suggest that participants generally agree with the current ICML process. [sent-18, score-0.098]
10 I expect there is some amount of anchoring effect going on where participants have an apparent preference for the known status quo, although it’s difficult to judge the degree of that. [sent-19, score-0.182]
11 Some survey results which aren’t of that sort are: 7. [sent-20, score-0.119]
12 It would be interesting to know for which fraction of accepted papers reviewers had their mind changed, but that isn’t there. [sent-22, score-0.302]
13 4% of authors don’t know if the reviewers read their response, believe they read and ignored it, or believe they didn’t read it. [sent-24, score-0.66]
14 6% support growing the conference with the largest fraction suggesting poster-only papers. [sent-27, score-0.097]
15 Many possibilities are precluded by scheduling, but AAAI, IJCAI, UAI, KDD, COLT, SIGIR are all serious possibilities some of which haven’t been used much in the past. [sent-30, score-0.232]
16 My experience with Mark ‘s new paper discussion site is generally positive—having comments emailed to interested parties really helps the discussion. [sent-31, score-0.31]
17 There are a few comments that authors haven’t responded to, so if you are an author you might want to sign up to receive comments. [sent-32, score-0.383]
18 I think it would be great if ICML can continue to attract new challenge workshops in the future. [sent-36, score-0.25]
19 If anyone else has comments about the workshops, I’d love to hear them. [sent-37, score-0.132]
wordName wordTfidf (topN-words)
[('apprenticeship', 0.168), ('authors', 0.167), ('optimization', 0.162), ('sigir', 0.156), ('comments', 0.132), ('icml', 0.131), ('ijcai', 0.126), ('read', 0.125), ('reinforcement', 0.123), ('survey', 0.119), ('reviewers', 0.118), ('possibilities', 0.116), ('adaptive', 0.114), ('changed', 0.111), ('aaai', 0.105), ('colt', 0.104), ('paper', 0.1), ('guarantees', 0.1), ('participants', 0.098), ('fraction', 0.097), ('uai', 0.091), ('challenge', 0.088), ('papers', 0.087), ('kdd', 0.086), ('haven', 0.085), ('receive', 0.084), ('emnlp', 0.084), ('szepesv', 0.084), ('bickel', 0.084), ('steffen', 0.084), ('generalizing', 0.084), ('formalizes', 0.084), ('diuk', 0.084), ('szita', 0.084), ('anchoring', 0.084), ('brendan', 0.084), ('frontier', 0.084), ('james', 0.084), ('martens', 0.084), ('workshops', 0.084), ('online', 0.08), ('emailed', 0.078), ('attract', 0.078), ('singer', 0.078), ('csaba', 0.078), ('yoram', 0.078), ('scheffer', 0.078), ('tobias', 0.078), ('walsh', 0.078), ('gentile', 0.078)]
simIndex simValue blogId blogTitle
same-blog 1 0.99999964 403 hunch net-2010-07-18-ICML & COLT 2010
Introduction: The papers which interested me most at ICML and COLT 2010 were: Thomas Walsh , Kaushik Subramanian , Michael Littman and Carlos Diuk Generalizing Apprenticeship Learning across Hypothesis Classes . This paper formalizes and provides algorithms with guarantees for mixed-mode apprenticeship and traditional reinforcement learning algorithms, allowing RL algorithms that perform better than for either setting alone. István Szita and Csaba Szepesvári Model-based reinforcement learning with nearly tight exploration complexity bounds . This paper and another represent the frontier of best-known algorithm for Reinforcement Learning in a Markov Decision Process. James Martens Deep learning via Hessian-free optimization . About a new not-quite-online second order gradient algorithm for learning deep functional structures. Potentially this is very powerful because while people have often talked about end-to-end learning, it has rarely worked in practice. Chrisoph
2 0.22974727 468 hunch net-2012-06-29-ICML survey and comments
Introduction: Just about nothing could keep me from attending ICML , except for Dora who arrived on Monday. Consequently, I have only secondhand reports that the conference is going well. For those who are remote (like me) or after the conference (like everyone), Mark Reid has setup the ICML discussion site where you can comment on any paper or subscribe to papers. Authors are automatically subscribed to their own papers, so it should be possible to have a discussion significantly after the fact, as people desire. We also conducted a survey before the conference and have the survey results now. This can be compared with the ICML 2010 survey results . Looking at the comparable questions, we can sometimes order the answers to have scores ranging from 0 to 3 or 0 to 4 with 3 or 4 being best and 0 worst, then compute the average difference between 2012 and 2010. Glancing through them, I see: Most people found the papers they reviewed a good fit for their expertise (-.037 w.r.t 20
3 0.20538136 454 hunch net-2012-01-30-ICML Posters and Scope
Introduction: Normally, I don’t indulge in posters for ICML , but this year is naturally an exception for me. If you want one, there are a small number left here , if you sign up before February. It also seems worthwhile to give some sense of the scope and reviewing criteria for ICML for authors considering submitting papers. At ICML, the (very large) program committee does the reviewing which informs final decisions by area chairs on most papers. Program chairs setup the process, deal with exceptions or disagreements, and provide advice for the reviewing process. Providing advice is tricky (and easily misleading) because a conference is a community, and in the end the aggregate interests of the community determine the conference. Nevertheless, as a program chair this year it seems worthwhile to state the overall philosophy I have and what I plan to encourage (and occasionally discourage). At the highest level, I believe ICML exists to further research into machine learning, which I gene
4 0.19224717 437 hunch net-2011-07-10-ICML 2011 and the future
Introduction: Unfortunately, I ended up sick for much of this ICML. I did manage to catch one interesting paper: Richard Socher , Cliff Lin , Andrew Y. Ng , and Christopher D. Manning Parsing Natural Scenes and Natural Language with Recursive Neural Networks . I invited Richard to share his list of interesting papers, so hopefully we’ll hear from him soon. In the meantime, Paul and Hal have posted some lists. the future Joelle and I are program chairs for ICML 2012 in Edinburgh , which I previously enjoyed visiting in 2005 . This is a huge responsibility, that we hope to accomplish well. A part of this (perhaps the most fun part), is imagining how we can make ICML better. A key and critical constraint is choosing things that can be accomplished. So far we have: Colocation . The first thing we looked into was potential colocations. We quickly discovered that many other conferences precomitted their location. For the future, getting a colocation with ACL or SIGI
5 0.18876611 452 hunch net-2012-01-04-Why ICML? and the summer conferences
Introduction: Here’s a quick reference for summer ML-related conferences sorted by due date: Conference Due date Location Reviewing KDD Feb 10 August 12-16, Beijing, China Single Blind COLT Feb 14 June 25-June 27, Edinburgh, Scotland Single Blind? (historically) ICML Feb 24 June 26-July 1, Edinburgh, Scotland Double Blind, author response, zero SPOF UAI March 30 August 15-17, Catalina Islands, California Double Blind, author response Geographically, this is greatly dispersed and the UAI/KDD conflict is unfortunate. Machine Learning conferences are triannual now, between NIPS , AIStat , and ICML . This has not always been the case: the academic default is annual summer conferences, then NIPS started with a December conference, and now AIStat has grown into an April conference. However, the first claim is not quite correct. NIPS and AIStat have few competing venues while ICML implicitly competes with many other conf
6 0.18808562 453 hunch net-2012-01-28-Why COLT?
7 0.1765793 343 hunch net-2009-02-18-Decision by Vetocracy
8 0.17130174 361 hunch net-2009-06-24-Interesting papers at UAICMOLT 2009
9 0.15927394 251 hunch net-2007-06-24-Interesting Papers at ICML 2007
10 0.14590794 304 hunch net-2008-06-27-Reviewing Horror Stories
11 0.14497788 116 hunch net-2005-09-30-Research in conferences
12 0.14476791 484 hunch net-2013-06-16-Representative Reviewing
13 0.14393586 65 hunch net-2005-05-02-Reviewing techniques for conferences
14 0.13988352 325 hunch net-2008-11-10-ICML Reviewing Criteria
15 0.13950621 309 hunch net-2008-07-10-Interesting papers, ICML 2008
16 0.13826682 461 hunch net-2012-04-09-ICML author feedback is open
17 0.13716364 385 hunch net-2009-12-27-Interesting things at NIPS 2009
18 0.13568954 320 hunch net-2008-10-14-Who is Responsible for a Bad Review?
19 0.13491207 174 hunch net-2006-04-27-Conferences, Workshops, and Tutorials
20 0.12989897 192 hunch net-2006-07-08-Some recent papers
topicId topicWeight
[(0, 0.314), (1, -0.152), (2, 0.114), (3, -0.107), (4, 0.16), (5, 0.006), (6, -0.012), (7, -0.083), (8, -0.016), (9, 0.069), (10, 0.082), (11, 0.079), (12, -0.147), (13, -0.009), (14, 0.069), (15, 0.056), (16, -0.033), (17, -0.02), (18, 0.001), (19, 0.031), (20, 0.052), (21, -0.029), (22, -0.054), (23, 0.021), (24, 0.008), (25, -0.026), (26, -0.009), (27, 0.008), (28, 0.049), (29, 0.087), (30, 0.052), (31, 0.043), (32, -0.015), (33, 0.038), (34, 0.01), (35, 0.017), (36, 0.018), (37, 0.062), (38, 0.014), (39, 0.017), (40, -0.006), (41, -0.006), (42, 0.036), (43, -0.013), (44, -0.017), (45, 0.056), (46, 0.035), (47, 0.018), (48, -0.031), (49, -0.006)]
simIndex simValue blogId blogTitle
same-blog 1 0.96505183 403 hunch net-2010-07-18-ICML & COLT 2010
Introduction: The papers which interested me most at ICML and COLT 2010 were: Thomas Walsh , Kaushik Subramanian , Michael Littman and Carlos Diuk Generalizing Apprenticeship Learning across Hypothesis Classes . This paper formalizes and provides algorithms with guarantees for mixed-mode apprenticeship and traditional reinforcement learning algorithms, allowing RL algorithms that perform better than for either setting alone. István Szita and Csaba Szepesvári Model-based reinforcement learning with nearly tight exploration complexity bounds . This paper and another represent the frontier of best-known algorithm for Reinforcement Learning in a Markov Decision Process. James Martens Deep learning via Hessian-free optimization . About a new not-quite-online second order gradient algorithm for learning deep functional structures. Potentially this is very powerful because while people have often talked about end-to-end learning, it has rarely worked in practice. Chrisoph
2 0.75323766 468 hunch net-2012-06-29-ICML survey and comments
Introduction: Just about nothing could keep me from attending ICML , except for Dora who arrived on Monday. Consequently, I have only secondhand reports that the conference is going well. For those who are remote (like me) or after the conference (like everyone), Mark Reid has setup the ICML discussion site where you can comment on any paper or subscribe to papers. Authors are automatically subscribed to their own papers, so it should be possible to have a discussion significantly after the fact, as people desire. We also conducted a survey before the conference and have the survey results now. This can be compared with the ICML 2010 survey results . Looking at the comparable questions, we can sometimes order the answers to have scores ranging from 0 to 3 or 0 to 4 with 3 or 4 being best and 0 worst, then compute the average difference between 2012 and 2010. Glancing through them, I see: Most people found the papers they reviewed a good fit for their expertise (-.037 w.r.t 20
3 0.72006625 361 hunch net-2009-06-24-Interesting papers at UAICMOLT 2009
Introduction: Here’s a list of papers that I found interesting at ICML / COLT / UAI in 2009. Elad Hazan and Comandur Seshadhri Efficient learning algorithms for changing environments at ICML. This paper shows how to adapt learning algorithms that compete with fixed predictors to compete with changing policies. The definition of regret they deal with seems particularly useful in many situation. Hal Daume , Unsupervised Search-based Structured Prediction at ICML. This paper shows a technique for reducing unsupervised learning to supervised learning which (a) make a fast unsupervised learning algorithm and (b) makes semisupervised learning both easy and highly effective. There were two papers with similar results on active learning in the KWIK framework for linear regression, both reducing the sample complexity to . One was Nicolo Cesa-Bianchi , Claudio Gentile , and Francesco Orabona Robust Bounds for Classification via Selective Sampling at ICML and the other was Thoma
4 0.71474981 452 hunch net-2012-01-04-Why ICML? and the summer conferences
Introduction: Here’s a quick reference for summer ML-related conferences sorted by due date: Conference Due date Location Reviewing KDD Feb 10 August 12-16, Beijing, China Single Blind COLT Feb 14 June 25-June 27, Edinburgh, Scotland Single Blind? (historically) ICML Feb 24 June 26-July 1, Edinburgh, Scotland Double Blind, author response, zero SPOF UAI March 30 August 15-17, Catalina Islands, California Double Blind, author response Geographically, this is greatly dispersed and the UAI/KDD conflict is unfortunate. Machine Learning conferences are triannual now, between NIPS , AIStat , and ICML . This has not always been the case: the academic default is annual summer conferences, then NIPS started with a December conference, and now AIStat has grown into an April conference. However, the first claim is not quite correct. NIPS and AIStat have few competing venues while ICML implicitly competes with many other conf
5 0.69303334 437 hunch net-2011-07-10-ICML 2011 and the future
Introduction: Unfortunately, I ended up sick for much of this ICML. I did manage to catch one interesting paper: Richard Socher , Cliff Lin , Andrew Y. Ng , and Christopher D. Manning Parsing Natural Scenes and Natural Language with Recursive Neural Networks . I invited Richard to share his list of interesting papers, so hopefully we’ll hear from him soon. In the meantime, Paul and Hal have posted some lists. the future Joelle and I are program chairs for ICML 2012 in Edinburgh , which I previously enjoyed visiting in 2005 . This is a huge responsibility, that we hope to accomplish well. A part of this (perhaps the most fun part), is imagining how we can make ICML better. A key and critical constraint is choosing things that can be accomplished. So far we have: Colocation . The first thing we looked into was potential colocations. We quickly discovered that many other conferences precomitted their location. For the future, getting a colocation with ACL or SIGI
6 0.65778244 439 hunch net-2011-08-01-Interesting papers at COLT 2011
7 0.6553244 309 hunch net-2008-07-10-Interesting papers, ICML 2008
8 0.6454711 188 hunch net-2006-06-30-ICML papers
9 0.6415869 318 hunch net-2008-09-26-The SODA Program Committee
10 0.62649912 453 hunch net-2012-01-28-Why COLT?
11 0.62521726 304 hunch net-2008-06-27-Reviewing Horror Stories
12 0.62314409 192 hunch net-2006-07-08-Some recent papers
13 0.60928535 251 hunch net-2007-06-24-Interesting Papers at ICML 2007
14 0.60734898 221 hunch net-2006-12-04-Structural Problems in NIPS Decision Making
15 0.6007297 343 hunch net-2009-02-18-Decision by Vetocracy
16 0.58979303 385 hunch net-2009-12-27-Interesting things at NIPS 2009
17 0.58492923 463 hunch net-2012-05-02-ICML: Behind the Scenes
18 0.57799232 466 hunch net-2012-06-05-ICML acceptance statistics
19 0.57788575 454 hunch net-2012-01-30-ICML Posters and Scope
20 0.57065827 406 hunch net-2010-08-22-KDD 2010
topicId topicWeight
[(3, 0.03), (9, 0.074), (27, 0.193), (38, 0.076), (48, 0.037), (51, 0.019), (53, 0.073), (55, 0.134), (65, 0.148), (67, 0.033), (94, 0.059), (95, 0.057)]
simIndex simValue blogId blogTitle
same-blog 1 0.94147199 403 hunch net-2010-07-18-ICML & COLT 2010
Introduction: The papers which interested me most at ICML and COLT 2010 were: Thomas Walsh , Kaushik Subramanian , Michael Littman and Carlos Diuk Generalizing Apprenticeship Learning across Hypothesis Classes . This paper formalizes and provides algorithms with guarantees for mixed-mode apprenticeship and traditional reinforcement learning algorithms, allowing RL algorithms that perform better than for either setting alone. István Szita and Csaba Szepesvári Model-based reinforcement learning with nearly tight exploration complexity bounds . This paper and another represent the frontier of best-known algorithm for Reinforcement Learning in a Markov Decision Process. James Martens Deep learning via Hessian-free optimization . About a new not-quite-online second order gradient algorithm for learning deep functional structures. Potentially this is very powerful because while people have often talked about end-to-end learning, it has rarely worked in practice. Chrisoph
2 0.90956753 398 hunch net-2010-05-10-Aggregation of estimators, sparsity in high dimension and computational feasibility
Introduction: (I’m channeling for Jean-Yves Audibert here, with some minor tweaking for clarity.) Since Nemirovski’s Saint Flour lecture notes , numerous researchers have studied the following problem in least squares regression: predict as well as (MS) the best of d given functions (like in prediction with expert advice; model = finite set of d functions) (C) the best convex combination of these functions (i.e., model = convex hull of the d functions) (L) the best linear combination of these functions (i.e., model = linear span of the d functions) It is now well known (see, e.g., Sacha Tsybakov’s COLT’03 paper) that these tasks can be achieved since there exist estimators having an excess risk of order (log d)/n for (MS), min( sqrt((log d)/n), d/n ) for (C) and d/n for (L), where n is the training set size. Here, “risk” is amount of extra loss per example which may be suffered due to the choice of random sample. The practical use of these results seems rather limited to trivial
3 0.88374424 249 hunch net-2007-06-21-Presentation Preparation
Introduction: A big part of doing research is presenting it at a conference. Since many people start out shy of public presentations, this can be a substantial challenge. Here are a few notes which might be helpful when thinking about preparing a presentation on research. Motivate . Talks which don’t start by describing the problem to solve cause many people to zone out. Prioritize . It is typical that you have more things to say than time to say them, and many presenters fall into the failure mode of trying to say too much. This is an easy-to-understand failure mode as it’s very natural to want to include everything. A basic fact is: you can’t. Example of this are: Your slides are so densely full of equations and words that you can’t cover them. Your talk runs over and a moderator prioritizes for you by cutting you off. You motor-mouth through the presentation, and the information absorption rate of the audience prioritizes in some uncontrolled fashion. The rate of flow of c
4 0.84518939 432 hunch net-2011-04-20-The End of the Beginning of Active Learning
Introduction: This post is by Daniel Hsu and John Langford. In selective sampling style active learning, a learning algorithm chooses which examples to label. We now have an active learning algorithm that is: Efficient in label complexity, unlabeled complexity, and computational complexity. Competitive with supervised learning anywhere that supervised learning works. Compatible with online learning, with any optimization-based learning algorithm, with any loss function, with offline testing, and even with changing learning algorithms. Empirically effective. The basic idea is to combine disagreement region-based sampling with importance weighting : an example is selected to be labeled with probability proportional to how useful it is for distinguishing among near-optimal classifiers, and labeled examples are importance-weighted by the inverse of these probabilities. The combination of these simple ideas removes the sampling bias problem that has plagued many previous he
5 0.82228655 466 hunch net-2012-06-05-ICML acceptance statistics
Introduction: People are naturally interested in slicing the ICML acceptance statistics in various ways. Here’s a rundown for the top categories. 18/66 = 0.27 in (0.18,0.36) Reinforcement Learning 10/52 = 0.19 in (0.17,0.37) Supervised Learning 9/51 = 0.18 not in (0.18, 0.37) Clustering 12/46 = 0.26 in (0.17, 0.37) Kernel Methods 11/40 = 0.28 in (0.15, 0.4) Optimization Algorithms 8/33 = 0.24 in (0.15, 0.39) Learning Theory 14/33 = 0.42 not in (0.15, 0.39) Graphical Models 10/32 = 0.31 in (0.15, 0.41) Applications (+5 invited) 8/29 = 0.28 in (0.14, 0.41]) Probabilistic Models 13/29 = 0.45 not in (0.14, 0.41) NN & Deep Learning 8/26 = 0.31 in (0.12, 0.42) Transfer and Multi-Task Learning 13/25 = 0.52 not in (0.12, 0.44) Online Learning 5/25 = 0.20 in (0.12, 0.44) Active Learning 6/22 = 0.27 in (0.14, 0.41) Semi-Superv
6 0.81742483 463 hunch net-2012-05-02-ICML: Behind the Scenes
7 0.81707966 437 hunch net-2011-07-10-ICML 2011 and the future
8 0.81646174 204 hunch net-2006-08-28-Learning Theory standards for NIPS 2006
9 0.80697948 116 hunch net-2005-09-30-Research in conferences
10 0.80616832 343 hunch net-2009-02-18-Decision by Vetocracy
11 0.80324411 464 hunch net-2012-05-03-Microsoft Research, New York City
12 0.80265528 452 hunch net-2012-01-04-Why ICML? and the summer conferences
13 0.79988706 194 hunch net-2006-07-11-New Models
14 0.79973495 484 hunch net-2013-06-16-Representative Reviewing
15 0.79948705 225 hunch net-2007-01-02-Retrospective
16 0.79945672 406 hunch net-2010-08-22-KDD 2010
17 0.79894793 318 hunch net-2008-09-26-The SODA Program Committee
18 0.79782444 309 hunch net-2008-07-10-Interesting papers, ICML 2008
19 0.79700965 40 hunch net-2005-03-13-Avoiding Bad Reviewing
20 0.7962243 490 hunch net-2013-11-09-Graduates and Postdocs