Learning to Solve Multiple Goals

نویسندگان

  • Jonas Karlsson
  • Dana H. Ballard
  • Shaun Gittens
چکیده

1997 ii Acknowledgments I thank my advisor, Dana Ballard, for his unwavering patience, support, and assistance throughout my eeorts on this work. As a constant source of enthusiasm, ideas, and insight, he allowed me to reach goals that would have been unattainable without his guidance. In addition, his willingness to read and comment on a seemingly endless succession of drafts of this document greatly helped to improve its scope, clarity, and readability. I would also like to thank my committee, for their penetrating questions and comments, forcing me to think hard about the assumptions and motivations that lay as a basis for this work. I am doubly indebted to Josh Tenenberg who rst set me on the path of learning to solve multiple goals, and shares my excitement for the concepts and ideas involved. He has also proven to be a good friend and of a genuinely warm spirit. Steve Whitehead is the other collaborator on the initial work on learning multiple goal, and was the rst person to get me interested in Machine Learning. To have been allowed the privilege of working with Josh and Steve is something for which I will always be grateful. Designing and implementing a driving simulator, as well as running experiments , involves a lot of hard, sometimes tedious work. I was fortunate to have several helpers along the way. Andrew Kachites McCallum collaborated on all versions of the driving simulator as well as lending me the NeXTstation on which much of the initial work was done. Andrew McCallum was a kindred spirit from the beginning: mon semblable, mon fr ere. His presence here made all the diier-ence, and I am greatly in his debt for his generosity. Tim Becker was also an integral part of the design and implementation of the Virtual Driving Simulator that was used in all experiments described herein. He taught me much about programming and design, and without his eeorts none of this work would have been possible. He has also been a good friend and supporter throughout my presence at Rochester, and our hikes in the Colorado Rockies and our many long conversations will always be a highlight of my memories. Eric Ringger implemented some ideas for using lookahead, which I'm sorry we never had time to pursue and was always happy to give thoughtful responses to all sorts of wild ideas. I would iii also like to …

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Bayesian Multitask Inverse Reinforcement Learning

We generalise the problem of inverse reinforcement learning to multiple tasks, from multiple demonstrations. Each one may represent one expert trying to solve a different task, or as different experts trying to solve the same task. Our main contribution is to formalise the problem as statistical preference elicitation, via a number of structured priors, whose form captures our biases about the ...

متن کامل

Five Phases to Pbl: Mita (multiple Intelligence Teaching Approach) Model for Redesigned Higher Education Classes

On January 24, 2000 at UCLA's Higher Education Research Institute, surveys of more than 260,000 full-time college freshmen reported boredom, drudgery and disengagement in class. This paper reports several reasons for lack of interest in higher education, and introduces a PBL model to address and help resolve this problem. The MITA (Multiple Intelligence Teaching Approach) model is applied to re...

متن کامل

Learning to Achieve Goals

Temporal diierence methods solve the temporal credit assignment problem for reinforcement learning. An important subproblem of general reinforcement learning is learning to achieve dynamic goals. Although existing temporal diierence methods, such as Q learning, can be applied to this problem, they do not take advantage of its special structure. This paper presents the DG-learning algorithm, whi...

متن کامل

پیش‌بینی نتایج یادگیری بر اساس تجربه دوره تحصیلی در دانشجویان دانشگاه علوم پزشکی رفسنجان در سال 1392

  Background and Objectives: Learners attitude of teaching - learning environment have a particular effect on learning skills and improving their performance during the course and their satisfaction after professional education. This study was designed to Predict learning outcomes based on the perceptions of the courses in Rafsanjan University of Medical Sciences.   Material and Methods: This c...

متن کامل

مقایسه الگوی طراحی مدارس متداول با مدارس بدون کلاس از منظر کارآمدی محیط یادگیری

Although the learning approach is gradually changing from teacher-centered to student-centered, the pattern of designing school in Iran still is following the inefficient approach of teacher-centered. Learning environments nowadays are being less formally timetabled and increasingly collaborative and socially participating oriented.  Inductive reasoning has been applied in this qualitativ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1997