From rsun at rpi.edu Sat Jul 1 20:13:53 2017 From: rsun at rpi.edu (Professor Ron Sun) Date: Sat, 1 Jul 2017 20:13:53 -0400 Subject: [ACT-R-users] Special issue on deep reinforcement learning: deadline extended to July 15, 2017 Message-ID: <7BD7DB90-CD15-4B60-99AC-3EB8E3823BB6@rpi.edu> Special Issue on Deep Reinforcement Learning in Neural Networks: deadline extended to July 15, 2017! https://www.journals.elsevier.com/neural-networks/call-for-papers/special-issue-on-deep-reinforcement-learning-in-neural-netwo Deep learning (DL) has become highly popular in recent years, among theoretically minded and application-focused researchers alike. Moreover, the idea of deep learning has been combined with reinforcement learning (RL), leading to deep reinforcement learning, which has achieved notable successes in tackling difficult problems. However, there are many open questions and issues that need to be addressed with regard to deep RL. Open questions with regard to deep RL include: ? How do we extend RL algorithms or systems to make them suitable for deep learning? How do we make RL (typically centered on values of states or state-action pairings) appropriately deep? ? How do we do so without jeopardizing useful characteristics of RL? ? What modification and enhancements to learning algorithms are necessary to accomplish deep RL in an effective and/or efficient manner? ? How can we make knowledge within deep RL systems explicit (generating explicit, symbolic, usable knowledge) and enable metacognitive reflection and regulation to some extent? ? How can deep learning help facilitate planning or model-based reinforcement learning? ? How can hierarchical or modular approaches be applied to deep RL? ? What theoretical/mathematical properties can be obtained with regard to deep RL (e.g., convergence, stability, robustness, and optimality)? ? How do we apply deep RL in real-world scenarios? The aim of this special issue is to showcase state-of-the-art work in the field of deep RL, addressing some of the above questions and beyond. Although there have no doubt been advances in addressing these questions, there is clearly room for further development. This special issue will provide a platform for deep learning and reinforcement learning researchers to share their work, for the sake of more rapid advances on a solid footing, fully realizing the potential of infusing reinforcement learning and deep learning. It also intends to showcase more effective applications in a variety of fields (robotics, control engineering, data analysis, and so on). We invite original research contributions on deep reinforcement learning (broadly defined). Possible topics for this special issue include, among others: ? New and better deep RL algorithms ? New and better neural network architectures for deep RL ? Better combinations of existing algorithms and techniques for deep RL ? Theories regarding deep RL ? Transfer learning and prior knowledge within deep RL ? Coping with uncertainty in deep RL ? Combining policy learning, value learning, and model-based search ? Symbolic structures from or within deep RL ? Planning and deep RL ? Mathematical analysis of deep RL (regarding convergence, optimality, stability, robustness, and so on) ? Hierarchical or modular RL ? Multi-agent RL ? Applications of deep RL algorithms, architectures, and systems to robotics, control, data analysis, prediction and forecast, modeling and simulation, and so on ? Applications of deep RL to cognitive-psychological or social modeling and analysis Survey papers are welcome also. Submission Procedure: Prospective authors should follow the standard author instructions for Neural Networks, and submit manuscripts online at http://ees.elsevier.com/neunet/. During submission, authors should indicate that their papers are for the special issue. Important Dates ? July 15, 2017 ? Deadline for submission ? December 1, 2017 ? Notification of review decisions to authors ? February 1, 2018 ? Deadline for submission of revised versions ? April 1, 2018 ? Final acceptance decision Guest Editors: Ron Sun, David Silver, Gerald Tesauro, Guang-Bin Huang ======================================================== Professor Ron Sun, Ph.D., FIEEE, FAPS, FPsyS Cognitive Science Department Rensselaer Polytechnic Institute 110 Eighth Street, Carnegie 302A Troy, NY 12180, USA phone: 518-276-3409 fax: 518-276-3017 email: dr.ron.sun [AT] gmail.com web: http://sites.google.com/site/drronsun ======================================================= From db30 at andrew.cmu.edu Wed Jul 12 11:04:17 2017 From: db30 at andrew.cmu.edu (db30 at andrew.cmu.edu) Date: Wed, 12 Jul 2017 11:04:17 -0400 Subject: [ACT-R-users] Updated ACT-R 7 software Message-ID: <04E914C7C4FB53D82BDCE90F@actr6b.psy.cmu.edu> The ACT-R 7 software available from the ACT-R website at has been updated. The current version is now 7.5-<2244:2017-07-11>. The two most notable changes are listed below. More details are available in the commit log found on the ACT-R website at: , but only the changes in the actr7 branch are relevant to the current software. A bug in the !bind! action of productions was fixed. When used on the RHS it was allowing a return value of nil to be bound to the variable which caused problems for anything that then tried to use that variable. Now, if nil is returned to a !bind! on the RHS it will print a warning about it and bind the variable to the value t instead to avoid errors downstream. The aural module was inconsistent in how it filled in the offset and duration values for audio-event chunks in the aural-location buffer. Now, it always fills in those slots appropriately if the corresponding sound ends while the chunk is still in the buffer. If you have any questions or problems with the new version please let me know. Dan From cl at cmu.edu Thu Jul 13 10:24:58 2017 From: cl at cmu.edu (Christian Lebiere) Date: Thu, 13 Jul 2017 10:24:58 -0400 Subject: [ACT-R-users] 2017 ACT-R Workshop Registration FINAL DEADLINE Message-ID: The 2017 ACT-R workshop will take place at University College London on July 26, 2017, between the ICCM and Cognitive Science conferences. The workshop agenda is available on the workshop home page , as is a link for registration. The FINAL registration deadline is this SUNDAY JULY 16. Please register by the end of the weekend if you are planning to attend. Looking forward to seeing you in London, Christian -------------- next part -------------- An HTML attachment was scrubbed... URL: From cl at cmu.edu Thu Jul 20 17:03:47 2017 From: cl at cmu.edu (Christian Lebiere) Date: Thu, 20 Jul 2017 17:03:47 -0400 Subject: [ACT-R-users] CFP Extension: A Standard Model of the Mind Message-ID: *The deadline for this CFP has been extended to July 30.* Invitation to submit a position paper to the AAAI 2017 Fall Symposium on A Standard Model of the Mind The purpose of this symposium is to engage the international research community in developing a standard model of the mind, with a focus specifically on human-like minds, which include human minds but also artificial minds that are either inspired by human ones or are similar because of common functional goals. The notion of a standard model has its roots in particle physics, where it is assumed to be internally consistent, yet still have major gaps; and serves as a cumulative reference point for the field while driving efforts to both extend and revise it. A standard model of the mind could yield similar benefits while also guiding experimentation, application, extension, interpretation, evaluation, and comparison. The intent is not to develop a single implementation, model or theory that everyone would abide by and agree is correct. What is sought is a statement of the best consensus given the community's current understanding of the mind, plus a sound basis for further refinement as more is learned. A beginning was made at the 2013 AAAIFall Symposium on Integrated Cognition, followed by an effort to capture and extend that initial consensus. Truly creating a standard model requires participation by researchers from across the community; hence this symposium. Format Working sessions will focus on the concept, framework, major components, and initial draft of a standard model; on mapping of existing architectures onto the model; and on summarizing the results and looking to the future. Each session will consist of an introduction, brief statements by 3-4 panelists on their position papers, and a moderated panel discussion. The focus will be on interactions that lead to a written summary document. Submissions Position papers (up to 6 pages) can be submitted to sm at ict.usc.edu by July 30, 2017. They should address fundamental issues with the concept of a standard model, outline proposals for such a model, or suggest specific contents. While contributions from all perspectives will be considered, those arising from a cognitive architecture approach ? and yielding implications for the computational structure and function of the mind and its parts ? are expected to be most directly relevant. Organizing Committee John Laird (University of Michigan, laird at umich.edu), Christian Lebiere (Carnegie Mellon University, cl at cmu.edu), Paul S. Rosenbloom (University of Southern California,rosenbloom at usc.edu) For More Information People considering writing position papers are encouraged to visit the symposium website (http://sm.ict.usc.edu), which has additional background resources. You can also contact any member of the organizing committee. -------------- next part -------------- An HTML attachment was scrubbed... URL: From cl at cmu.edu Sat Jul 22 12:11:35 2017 From: cl at cmu.edu (Christian Lebiere) Date: Sat, 22 Jul 2017 12:11:35 -0400 Subject: [ACT-R-users] ACT-R Workshop Directions Message-ID: Here are the directions for the ACT-R Workshop on Wednesday at UCL: Room 4.05 66-72 Gower Street London WC1E Google Maps: https://goo.gl/maps/rWmP7x3RMsz For public transport ? closest stations: ? Goodge Street on the Northern Line (4 min). ? Euston Square on the Circle, Metropolitan and Hammersmith & City Lines (4 min). ? Euston on the Northern and Victoria Lines (5 min). ? Warren Street on the Northern and Victoria Lines (5 min). The nearest mainline stations are Euston (5 min walk) and King's Cross St Pancras (20 min walk or one stop on the underground to Euston Square). Christian -------------- next part -------------- An HTML attachment was scrubbed... URL: From j.p.borst at rug.nl Thu Jul 27 08:21:55 2017 From: j.p.borst at rug.nl (Jelmer Borst) Date: Thu, 27 Jul 2017 14:21:55 +0200 Subject: [ACT-R-users] PhD Scholarship on Model-based Cognitive Neuroscience Message-ID: PhD Scholarship on Model-based Cognitive Neuroscience The Institute of Artificial Intelligence of the University of Groningen offers a four-year scholarship for a PhD position on model-based cognitive neuroscience. Model-based cognitive neuroscience is the area of research that bridges the disciplines of computational cognitive modeling and cognitive neuroscience. Cognitive models ? be it symbolic process models, mathematical models, or neural network models ? are notoriously hard to evaluate based on behavioral measures alone. For that reason, researchers have turned to neuroscience (M/EEG, fMRI) as an additional source of information. At the same time, neuroimaging data is often so complex that it is difficult to fully account for with traditional analysis methods. As a solution, cognitive and mathematical models have been used within the analysis stream to interpret neural measures directly. This dual approach, using neuroscience to inform models and models to inform neuroimaging analyses, is very powerful, and has led to the emerging field of model-based cognitive neuroscience (see a recent special issue of the Journal of Mathematical Psychology for an introduction: Palmeri, Love, & Turner, 2017; Turner, Forstmann, Love, Palmeri, & van Maanen, 2017). The PhD position is available in the cognitive modeling group, under supervision of Jelmer Borst. The goal of our group is to better understand cognitive processes in the human mind. To achieve this, we combine computational modeling with fMRI, EEG, and MEG data, and also apply machine learning techniques to analyze neural data. Top candidates will be invited to write and develop their own research project within the general scope of this research line. The selection procedure for the scholarship is as follows: 1. Candidates apply by submitting a cover letter and CV (see below for details). 2. Based on the cover letters, several candidates will be invited to write a short research proposal within the topic of model-based neuroscience. 3. These candidates will be asked to present this proposal as part of the job interview. 4. After the selection, the candidate?s proposal will be further developed in collaboration with the PhD advisors: Jelmer Borst, Niels Taatgen, and Hedderik van Rijn. Qualifications Successful candidates will have completed a Master?s degree (or equivalent) in Cognitive Neuroscience, Artificial Intelligence, or another field of science relevant for the position. The ideal candidate has experience with M/EEG or fMRI and cognitive modeling, and has strong programming skills. Conditions The PhD student will be enrolled in the PhD Scholarship Programme and receive a scholarship of ? 2027 per month (gross) from the University of Groningen for a period of four years. Date The preferred start date is November 1, 2017, but this can be postponed to February 1, 2018. Application Please see the attachment for details on the application procedure. The application deadline is August 30. For more information, contact Jelmer Borst (j.p.borst at rug.nl ). http://www.rug.nl/education/phd-programmes/phd-scholarship-programme/phd-scholarships?details=00347-02S0005SGP -- Jelmer Borst Assistant Professor University of Groningen Dept. of Artificial Intelligence E: j.p.borst at rug.nl W: http://www.jelmerborst.nl/ -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: phd_modelbased_neuro.pdf Type: application/pdf Size: 48648 bytes Desc: not available URL: -------------- next part -------------- An HTML attachment was scrubbed... URL: