Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id IAA09857; Mon, 20 Mar 1995 08:48:07 +0100 Date: Mon, 20 Mar 1995 08:48:07 +0100 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: Maarten van Someren (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) Subject: icml workshop - 1 WORKSHOP CALL FOR PAPERS Twelfth International Conference on Machine Learning Applying Machine Learning in Practice Tahoe City, California, U.S.A. July 9, 1995 Applications of machine learning (ML) involve specifying the learning and performance tasks, constructing a dataset or data generator, and selecting and modifying appropriate learning algorithms for these tasks. Most published work on ML concerns the analyses of learning algorithms. Reading such work, one might be led to believe that little effort is required to obtain good performance with the selected algorithm(s) on the selected task. Machine learning practitioners and empiricists know that this view is fallacious. Successful applications or demonstrations of ML algorithms involve substantial user and domain expertise to guide multiple iterations of data transformation, algorithm selection, and/or algorithm modification. Unfortunately, such expertise is rarely communicated in the machine learning literature. OBJECTIVE The purpose of this workshop is to characterize the expertise used during the application of ML algorithms to real-world problems and, in doing so, to develop a better understanding of how to use ML tools successfully. We solicit descriptions of the expertise exhibited during the complete sequential decision process leading to successful ML applications and a discussion of what guided the decision making and selection of the approaches used at each step. Submissions involving the following or related topics are specifically welcome: 1. Papers documenting the expertise used during the sequential process of modifying/extending existing algorithms and/or domain data to solve complex learning tasks. 2. Novel models of the decision-making process describing how to apply expertise during ML problem solving. 3. Surveys or comparisons of different such process models for guiding the application of learning algorithms. 4. Reports on software tools and environments for capturing and/or automatically applying ML application expertise. 5. Empirical studies documenting the effect of expertise on the performance of machine learning algorithms. We encourage authors of submissions to discuss the explanations and motivations they used to guide the data transformation process, the selection of the experimental method, the selection of simulators, the selection of algorithms, their modifications, and/or parameter selection. Relevant discussions on previously published work are welcome, so long as the focus is on documenting the expertise used rather than, for example, predictive accuracy. Likewise, reports of informative negative results are also encouraged, with the caveat that lessons learned from these results can be used to guide the ML practitioner so as to avoid future pitfalls and instead lead to an ultimately successful application. Preference will be given to submissions discussing work with real-world problems - rather than case studies comparing the performance of a suite of algorithms on a set of datasets from the UCI repository - in which the authors discuss how one can define and solve such problems so as to attain good results on the performance task. We will request participants who wish to present papers to complete a simple questionnaire that we will soon provide via our World-Wide Web pages. The purpose of this questionnaire is to provide a mechanism for comparing how ML expertise was used in the workshop participants' projects. We will also request that authors of presented papers include in their paper a figure outlining the process model they chose. We will make examples of such process models available in our World-Wide Web pages. We plan to use the authors' figures and the completed questionnaires to characterize and summarize this workshop's contributions. Specifically, we hope to identify common themes that appear in the contributions. Our findings will be included as a paper in the working notes. FORMAT This workshop will consist of invited presentations, paper presentations and the presentation of our summary results. A panel discussion that focuses on common themes involved with using ML application expertise is being considered dependent on the submissions received. Time will be reserved for open discussions on this subject. Speakers will be selected based on the clarity of their contribution and by coverage of the different methods described among the submissions. Please see our World-Wide Web pages for up-to-date details. SUBMISSION INFORMATION For those wishing to present at this workshop, please submit a paper of at most ten pages (plus a completed copy of our questionnaire, when made available). For those wishing to simply attend this workshop, please instead submit a one or two-page description summarizing your interests relevant to this workshop. The first page of each submission should include the title, the author's name(s) and affiliation(s), a complete mailing address, phone number, FAX number, e-mail address, an abstract of at most 250 words, and two to four keywords. All submissions (papers or descriptions of relevant work) will be reviewed by at least two members of the organizing committee. Please e-mail PostScript submissions to the contact address below by MAY 1, 1995. Two-sided hardcopy submissions are welcome for those who do not have PostScript capability. The revised versions of all accepted submissions will be made World-Wide Web accessible. Workshop attendees will be provided with hardcopies of the complete working notes. All workshop invitees should contact us by June 10, 1995 to verify their registration. SCHEDULE Submission deadline: MAY 1, 1995 Notification of acceptance: May 15 Workshop registration deadline: June 10 Revised submissions due: June 15 Working notes available on-line: June 22 Workshop date: July 9, 1995 ORGANIZING COMMITTEE David W. Aha, NRL (co-chair) Jason Catlett, AT&T Bell Laboratories Haym Hirsh, Rutgers University Patricia Riddle, Boeing Computer Services (co-chair) CONTACT ADDRESS Commanding Officer Naval Research Laboratory Attn: David Aha, Code 5510 4555 Overlook Ave, SW Washington, D.C. 20375-5337 aha@aic.nrl.navy.mil (202) 767-9006 (202) 767-3172 FAX http://www.aic.nrl.navy.mil/~aha/imlc95-workshop/home.html Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id IAA09829; Mon, 20 Mar 1995 08:45:24 +0100 Date: Mon, 20 Mar 1995 08:45:24 +0100 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: Maarten van Someren (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) Subject: icml workshop - 2 WORKSHOP CALL FOR PAPERS Twelfth International Conference on Machine Learning Agents that Learn from Other Agents Tahoe City, California, U.S.A. July 9, 1995 Agent-oriented learning is currently receiving a great deal of attention among machine learning researchers. The purpose of this workshop is to draw researchers from diverse areas of machine learning, such as learning in the context of distributed AI, planning and learning, software agents, knowledge acquisition, reinforcement learning, computational learning theory, neural networks, genetic algorithms, explanation-based learning, and multistrategy learning to address the unifying theme of agents that learn from other agents. This workshop is a special opportunity for empirically-oriented machine learning researchers to interact with the theoretical COLT community on a topic of mutual interest. There has been a growing trend in agent-oriented learning toward learning methods that involve interacting with other agents. One such interaction is via advice-taking (Suddarth and Holden 1991), which may well turn out to be the most efficient method for building software agent architectures. Instruction is steadily increasing in popularity as a method for agent learning (e.g., Gordon & Subramanian 1993, Huffman & Laird 1993, Lin 1993, Maclin & Shavlik 1994, Noelle & Cotrell 1994, and Tecuci, Hieb, Hille, & Pullen 1994). Meanwhile, there is also an active interest in agents that learn from observation (e.g., Iba and Langley 1992, Lin 1993), as well as agents that communicate and learn from each other (e.g., Etzioni and Weld 1994, Lashkari, Metral, & Maes 1994). Furthermore, in the COLT community there have been a number of papers on team learning (e.g., Daley, Kalyanasundaram, & Velauthapillai 1993). For these reasons, the main focus of our workshop is on learning from instruction or observation, rather than the environment. The instructors might be humans or other automated agents. This workshop will encourage papers on the following topics: - Learning from instruction or advice - Learning from observation of humans or other agents - Learning from multiagent interactions, including multiagent cooperation - Multistrategy learning methods for agents - The role of hybrid architectures in learning from instruction - The role of prior knowledge in learning from instruction - Theoretical foundations of team learning - Methodologies for building (knowledge-based) learning agents - Languages for inter-agent communication - Applications of learning agents, such as intelligent automated forces in simulated environments, softbots that find Internet papers or newsgroup postings of interest to another agent, resource-limited agent construction, learning apprentice systems, and intelligent information retrieval FORMAT This workshop will begin with a keynote talk by Dr. Tom Mitchell, followed by paper presentations on the main themes of the workshop. After each paper presentation, there will be time allotted for discussion. Furthermore, the presented papers will be grouped into theme-oriented sessions, and each session will be followed by an extended discussion period. If more papers are accepted than can be presented, a poster session will be included. The workshop will conclude with a panel discussion on the central issues of the workshop. The issues to be discussed will be distributed in advance to all of the attendees in order to encourage their participation. SUBMISSION INFORMATION For those wishing to present papers, our selection will be made based on 5-7 page extended abstracts (using 12-pt font and double spacing). Submissions should be sent by email to the contact address below by MAY 1, 1995. PostScript is the preferable submission mode. The working notes will appear in a proceedings published on the World-Wide Web. Those who wish to attend the workshop without presenting a paper should send email to the contact address below stating that they intend to register and should include a one-paragraph description of their research interests. We need these items so that we can estimate the workshop attendance for room requirements and agenda preparation. IMPORTANT DATES Extended abstracts/research descriptions due: MAY 1, 1995 Notification of acceptance: May 22 Submission of final (revised) abstracts for working notes: June 15 Workshop: July 9, 1995 ORGANIZING COMMITTEE Diana Gordon, Naval Research Laboratory (chair) Jude Shavlik, University of Wisconsin Devika Subramanian, Cornell University Gheorghe Tecuci, George Mason University and Romanian Academy CONTACT ADDRESS Dr. Diana Gordon Naval Research Laboratory, Code 5514 4555 Overlook Avenue, S.W. Washington, D.C. 20375-5337 gordon@aic.nrl.navy.mil (202) 767-2686 (202) 767-3172 FAX Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id IAA09833; Mon, 20 Mar 1995 08:46:00 +0100 Date: Mon, 20 Mar 1995 08:46:00 +0100 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: Maarten van Someren (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) Subject: icml workshop - 3 >From schlimme@eecs.wsu.edu Thu Mar 2 07:10:31 1995 Received: from dns1.eecs.wsu.edu by swi.swi.psy.uva.nl with SMTP id AA01443 (5.65c/IDA-1.4.4/SWI-1.1m for ); Thu, 2 Mar 1995 07:10:29 +0100 Received: from stimpy.eecs.wsu.edu by dns1.eecs.wsu.edu (8.6.8.1/8.940801) id VAA04375; Wed, 1 Mar 1995 21:14:45 -0800 From: "Jeffrey C. Schlimmer" Received: by stimpy.eecs.wsu.edu (1.37.109.4) id AA03468; Wed, 1 Mar 95 21:14:45 -0800 Message-Id: <9503020514.AA03468@stimpy.eecs.wsu.edu> Subject: Machine Learning Conference: Genetic Programming CFP To: ml95@eecs.wsu.edu Date: Wed, 1 Mar 1995 21:14:44 -0800 (PST) X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 4228 Status: RO WORKSHOP CALL FOR PAPERS Twelfth International Conference on Machine Learning Genetic Programming - From Theory to Real-World Applications Tahoe City, California, U.S.A. July 9, 1995 The goal of the workshop is to shed light onto the methodology for understanding, explaining and controlling GP search and to show how these issues are reflected in GP frameworks and successful or innovative applications. The workshop will focus on three discussion and paper topics: 1. Theoretical issues. Questions about representation and operators have generated both theoretical and experimental work in the field of genetic algorithms (GA). Similar questions are relevant for GP. o Representation-- How do architectural choices influence performance? o Operators-- What operators are appropriate and why? o Search-- How does GP sample the space of programs? Why does GP work, even for tremendously huge search spaces? 2. GP algorithms. GP extensions show that the process of finding better solutions can be accelerated. We would like to stimulate other approaches to GP search by encouraging new architectural ideas and the formal analysis of such ideas. o What is the relationship with other machine learning techniques, such as inductive logic programming? o Application domain-- Are particular classes of applications facilitated by particular approaches? 3. Real-world and innovative applications. GP has proven that it can go beyond toy applications in problems of classification, image analysis, system identification, prediction and control. We encourage submissions reporting GP applications to real-world problems as well as innovative application ideas. The workshop will also explore GP-based approaches to real problems faced by the ML community. o Efficiency-- How do results of the GP approach compare with results of other machine learning or artificial intelligence approaches? What makes GP an attractive tool? o Scalability-- What makes an implementation scalable? o Biases-- What are the limits of particular approaches and representations? Are there desired features that cannot be implemented in GP? To allow for a pointed technical discussion of the issues, each submission should address the questions relevant for its appropriate category. Be very specific about the experiments reported. Discuss why you have chosen a particular problem representation. Show the influence of the various architectural decisions. The workshop will consist of invited talks, presentations of selected papers and discussions on the proposed topics. We will encourage interaction and discussion of ideas among participants, reserving additional time for this at the end of each presentation session. Based on submissions in the above categories, we will select discussion leaders. SUBMISSION INFORMATION Potential attendees should submit either an extended abstract (5 pages) or a full paper (which should not exceed 20 pages) by APRIL 24, 1995. We encourage everyone to submit their papers or extended abstracts electronically, Postscript or ASCII only. E-mail submissions should be sent to the contact address below. If e-mail submission is not possible, please send three copies of the paper or abstract to the contact address below. We intend to create working notes to be distributed at the time of the workshop. IMPORTANT DATES Submission of extended abstracts/papers: APRIL 24, 1995 Acceptance notification: May 12 Submission of final camera-ready papers: June 9 Workshop: July 9, 1995 ORGANIZING COMMITTEE Justinian Rosca (Chair) - University of Rochester Frederic Gruau - Stanford University Kim Kinnear - Adaptive Computing Technologies John Koza - Stanford University Walter Tackett - Neuromedia CONTACT ADDRESS ML-95 GP Workshop Justinian Rosca 615 Computer Studies Building Computer Science Department University of Rochester Rochester, NY 14627 rosca@cs.rochester.edu (716) 275-1174 (716) 461-2018 FAX Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id IAA09839; Mon, 20 Mar 1995 08:46:51 +0100 Date: Mon, 20 Mar 1995 08:46:51 +0100 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: Maarten van Someren (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) Subject: icml workshop - 4 >From schlimme@eecs.wsu.edu Thu Mar 2 07:34:20 1995 Received: from dns1.eecs.wsu.edu by swi.swi.psy.uva.nl with SMTP id AA01523 (5.65c/IDA-1.4.4/SWI-1.1m for ); Thu, 2 Mar 1995 07:34:18 +0100 Received: from stimpy.eecs.wsu.edu by dns1.eecs.wsu.edu (8.6.8.1/8.940801) id VAA05224; Wed, 1 Mar 1995 21:47:57 -0800 From: "Jeffrey C. Schlimmer" Received: by stimpy.eecs.wsu.edu (1.37.109.4) id AA04181; Wed, 1 Mar 95 21:47:56 -0800 Message-Id: <9503020547.AA04181@stimpy.eecs.wsu.edu> Subject: Machine Learning Conference: Prog. by Demo CFP To: ml95@eecs.wsu.edu Date: Wed, 1 Mar 1995 21:47:55 -0800 (PST) X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 3793 Status: RO WORKSHOP CALL FOR PAPERS Twelfth International Conference on Machine Learning Learning from Examples versus Programming by Demonstration: Is Interaction the Key to (Better) Applications? Tahoe City, California, U.S.A. July 9, 1995 Inductive Learning from Examples (LfE) is a well established subject in Machine Learning. "Pure" LfE is performed automatically without any human interaction. Programming by Demonstration (PbD) on the other hand can be seen as some kind of "extreme" form of user-supported LfE where the user continually interacts with a PbD system. The nearly exclusive focus of it is the learning of programs. BACKGROUND PbD researchers have been disappointed with standard ML algorithms. They require too many examples or too strong a domain theory. Moreover they do not learn "useful" concepts that deterministically generate or modify data, but rather learn how to classify. Finally they easily get out of the user's control. Therefore, PbD applications to date have relied on pragmatic assumptions, application-specific heuristics, and manual intervention, to make plausible inferences from very few examples. Although this approach has yielded some successful early results, the systems require considerable engineering and have strict and often counter-intuitive limits on their adaptability. ML techniques integrated with user interaction hold out the promise of simpler, more capable systems. But the dialog between ML and PbD is just beginning. SCOPE The workshop aims at bringing together researchers from the Machine Learning as well as different application areas. These include robotics, graphics interface development, office automation, and interactive software design. Topics of interest include, but are not limited to: - Combining LfE with user interaction to create working PbD applications - Acquisition and maintenance of control-knowledge using interaction - Relations between an agent's background knowledge, reasoning methods, and user interaction abilities - Which results of LfE are not used in PbD systems today, and why? - Effects of PbD system development on existing ML techniques and problems - User interfaces and system architectures for PbD systems - Why does PbD receive only little attention in the ML community so far? SUBMISSION OF PAPERS Paper submissions are limited to 5000 words. The title page must contain the title of the talk, name(s) and affiliation(s) of the author(s) and a list of keywords as well as the full address (including e-mail) of the first author. A sample paper, a LaTeX style file and a MS-Word file are available via ftp://ftpipr.ira.uka.de/pub/conferences/ML95-PBD/ . A laser-quality copy of the paper must be received by the workshop organizers at the contact address below by APRIL 7, 1995. The electronic submission of papers in Postscript format to ftp://ftpipr.ira.uka.de/ is encouraged. Accepted papers (about 9) will be published in the Workshop notes, the best papers will be selected for a special issue of AI and Engineering Applications. IMPORTANT DATES Submission deadline: APRIL 7, 1995 Notification of acceptance: May 2 Camera-ready paper: June 9 Workshop: July 9, 1995 PROGRAM COMMITTEE S. Bocionek (Munich, Germany) R. Dillmann (Karlsruhe, Germany) A. Giordana (Turin, Italy) Y. Kuniyoshi (Tsukuba, Japan) D. Maulsby (Cambridge, USA) CONTACT ADDRESS Holger Friedrich University of Karlsruhe Institute for Real-Time Computer Systems & Robotics D-76128 Karlsruhe, Germany friedric@ira.uka.de ftp://ftpipr.ira.uka.de/ This announcement is also available in PostScript form in the URL ftp://eecs.wsu.edu/pub/ml95/workshop-lfevpbd.ps . Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id IAA09846; Mon, 20 Mar 1995 08:47:36 +0100 Date: Mon, 20 Mar 1995 08:47:36 +0100 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: Maarten van Someren (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) Subject: icml workshop - 5 >From schlimme@eecs.wsu.edu Thu Mar 2 07:59:33 1995 Received: from dns1.eecs.wsu.edu by swi.swi.psy.uva.nl with SMTP id AA01564 (5.65c/IDA-1.4.4/SWI-1.1m for ); Thu, 2 Mar 1995 07:59:31 +0100 Received: from stimpy.eecs.wsu.edu by dns1.eecs.wsu.edu (8.6.8.1/8.940801) id WAA05806; Wed, 1 Mar 1995 22:13:25 -0800 From: "Jeffrey C. Schlimmer" Received: by stimpy.eecs.wsu.edu (1.37.109.4) id AA04793; Wed, 1 Mar 95 22:13:24 -0800 Message-Id: <9503020613.AA04793@stimpy.eecs.wsu.edu> Subject: Machine Learning Conference: Reinforcement CFP To: ml95@eecs.wsu.edu Date: Wed, 1 Mar 1995 22:13:24 -0800 (PST) X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 5743 Status: RO WORKSHOP CALL FOR PAPERS Twelfth International Conference on Machine Learning Value Function Approximation in Reinforcement Learning Tahoe City, California, U.S.A. July 9, 1995 This workshop will explore the issues that arise in reinforcement learning when the value function cannot be learned exactly, but must be approximated. It has long been recognized that approximation is essential on large, real-world problems because the state space is too large to permit table-lookup approaches. In addition, we need to generalize from past experiences to future ones, which inevitably involves making approximations. In principle, all methods for learning from examples are relevant here, but in practice only a few have been tried, and fewer still have been effective. The objective of this workshop is to bring together all the strands of reinforcement learning research that bear directly on the issue of value function approximation in reinforcement learning. We hope to survey what works and what doesn't, and achieve a better understanding of what makes value function approximation special as learning from examples problem. The key computational idea underlying reinforcement learning is the iterative approximation of the value function---the mapping from states (or state-action pairs) to an estimate of the long-term future reward obtainable from that state. For large problems, the approximation of the value function must involve generalization from examples to reduce memory requirements and training time. Moreover, generalization may help beat the curse of dimensionality for problems in which the complexity of the value function increases sub-exponentially with the number of state variables. Generalizing function approximators have been used effectively in reinforcement learning as far back as Samuel's checker player, which used a linear approximator, and Michie and Chambers' BOXES system, which used state aggregation. Tesauro's TD-Gammon, which used a backpropagation network, provides a tantalizing recent demonstration of just how effective this can be. However, almost all of the theory of reinforcement learning is dependent on the assumption of tabular representations of the value function, for which generalization is impossible. Moreover, several researchers have argued that there are serious hazards intrinsic to the approximation of value functions by reinforcement learning methods. It is an important and still unclear question as to just how serious these hazards might be. In this workshop we will survey the substantial recent and ongoing work pertaining to this question. We will seek to answer it or, failing that, to identify the remaining outstanding issues. If existing theory of reinforcement learning no longer applies when an approximate value function is learned, what are we to do? The workshop will explore the problem and following possible responses, among others: (1) ignore the problem and proceed empirically, perhaps discovering as we create applications that some seem to work and some don't. A danger with this approach is that successes will be reported loudly and failures whispered quietly, and we may become a community of tweakers. (2) analyze what properties function approximators need in order to work well with current reinforcement learning algorithms: are there function approximators specially suited to reinforcement learning? (3) invent new reinforcement learning methods that are specifically designed to work well with function approximators. (4) invent new theory for value function approximation. What bounds can be placed on the errors? Can particular function approximators give us better bounds or stability guarantees? Can online training get us better bounds or stability guarantees? (5) What can we learn from the literature of other fields: - optimal control theory and LQG regulation - heuristic function generation techniques in computer game playing - operations research and differential games WORKSHOP FORMAT Several focussed sessions with short (~15 minute) talks followed by moderated discussion. WHO SHOULD ATTEND All researchers with empirical or theoretical experience with value function approximation in reinforcement learning or dynamic programming. As an expression of interest, please submit as soon as possible a short (one paragraph to one page) statement of your research interests in value function approximation, or a copy of a paper you have written in this area. We will use this material to help select and organize the sessions. If you would like to make a presentation at the workshop, please also submit an extended abstract (2-5 pages) by MAY 1, 1995. SUBMISSION INFORMATION Electronic submissions (ASCII or postscript) are preferred. E-mail or post submissions should be sent to the contact address below by MAY 1, 1995. IMPORTANT DATES Submission of statements of interest: ASAP Submission of extended abstracts/papers: MAY 1, 1995 Acceptance notification: May 19 Submission of final camera-ready papers: June 9 Workshop: July 9, 1995 ORGANIZERS Andrew Moore (chair) - Carnegie Mellon University Rich Sutton - Stow Research Justin Boyan - Carnegie Mellon University PROGRAM COMMITTEE John Tsitsiklis - MIT, Lab for Information and Decision Sciences Satinder Singh - MIT, Brain and Cognitive Sciences Leemon Baird - Wright Patterson Air Force Base CONTACT ADDRESS Andrew W. Moore Smith Hall 221 Robotics Institute Carnegie Mellon University Pittsburgh, PA 15213 awm@cs.cmu.edu http://www.cs.cmu.edu:8001/Web/Groups/reinforcement/ml95/ Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id LAA22609; Tue, 28 Mar 1995 11:37:37 +0200 Date: Tue, 28 Mar 1995 11:37:37 +0200 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: Maarten van Someren (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) Subject: ICML 95 - papers An initial list of papers accepted for ML-95 is available on the conference's World-Wide Web page in http://www.eecs.wsu.edu/~schlimme/ml95.html Congratulations to the successful authors! We're going to have an interesting technical program. A schedule for the conference and registration information will be available shortly. --Jeff Schlimmer Return-Path: jq@lirmm.lirmm.fr Received: from [193.49.104.48] ([193.49.104.48]) by lirmm.lirmm.fr (8.6.10/8.6.4) with SMTP id PAA10270; Mon, 10 Apr 1995 15:39:47 +0200 Date: Mon, 10 Apr 1995 15:39:47 +0200 X-Sender: jq@lirmm.lirmm.fr Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Eudora F1.5.1 To: gascuel, hr, reitz, js, mephu, pierre, pompidor, vignal, cdlh, gracy@EMBL-Heidelberg.DE, jappy, vismara From: " (MLnet Admin)" (transmis par jq@lirmm.lirmm.fr (Joel Quinqueton)) ------- Forwarded Message PRELIMINARY ANNOUNCEMENT Twelfth International Conference on Machine Learning Granlibakken Resort, Tahoe City, California, U.S.A. July 9-12, 1995 Early Registration Deadline is MAY 26, 1995. Housing Reservation Deadline is JUNE 8, 1995. The Twelfth International Conference on Machine Learning (ML-95) will be held at the Granlibakken Resort in Tahoe City, California during July 9-12, 1995, with informal workshops and tutorials on July 9. The conference will include presentations of refereed papers and three invited talks. This preliminary announcement, which omits the final technical program, is being provided so that travel arrangements can be made as early as possible. An updated announcement, including the technical program, will be distributed sometime in late April. CONFERENCE HIGHLIGHTS Invited Lectures by: - David Heckerman, Microsoft Research, "Machine Learning and Uncertainty in AI". - Bruce Croft, UMass, Amherst, "Machine Learning and Information Retrieval". - Dean Pomerleau, CMU, "Machine Learning for Autonomous Driving and Collision Warning". 68 Papers Accepted to ML-95 Two Tutorials Five Workshops Student Discount on Registration FOR FURTHER INFORMATION Please consult the conference's World-Wide Web pages in http://www.eecs.wsu.edu/~schlimme/ml95.html or send email to ml95@cs.ucdavis.edu . #| Jeffrey C. Schlimmer, Asst. Prof., School of EE & CS, Washington State University, Pullman, WA 99164-2752, (509) 335-2399, (509) 335-3818 FAX http://www.eecs.wsu.edu/~schlimme/ |# ------- End of Forwarded Message