Daniel E. Acuna [Available in PDF]
Center for Cognitive Sciences |
Office phone: (612)
– 626 9296 |
Placement in a post-doctoral or faculty research position
I study human sequential decision-making under uncertainty, from perceptual to high-level cognitive tasks. In particular, I develop rational models based on Bayesian Reinforcement Learning to investigate how people integrate structure learning, transfer learning, spatiotemporal abstraction and state factorization into exploration and exploitation. Additionally, I am interested in understanding how humans can solve real-world instantiations of computationally hard problems in a near-optimal fashion. I use statistical physics and computational complexity to uncover statistical regularities of problems and study whether humans exploit these regularities to structurally bias their search schedules.
Keywords: Sequential decision-making under uncertainty, Bayesian reinforcement learning, structure learning, human problem-solving, rational and bounded-rational analysis of behavior
University of Minnesota, Twin Cities |
Minneapolis, MN |
2010 |
|
Ph.D. candidate in Computer Science, GPA: 3.73 Thesis: Rational analysis of sequential decision-making in humans and machines Advisor: Paul Schrater |
|||
University of Santiago |
Santiago, Chile |
2003,2004 |
|
Master of Engineering Sciences in Information Technology Thesis: An algorithm for the Traveling Salesperson Problem based on players of a computer game |
|||
Bachelor in Engineering Sciences and Professional Degree of Engineering in Information Technology (Highest Honors) |
|||
NIH Neuro-physical-computational Sciences (NPCS) Graduate Training
Fellowship (1R90 DK71500-04) |
2008-Present |
International Graduate Student Fellowship of the Chilean Council of
Scientfic and Technological Research and the World Bank |
2006-Present |
NIPS 2009 travel award |
2008 |
Master Degree Fellowship, Universidad de
Santiago |
2003 |
University of Minnesota, Twin Cities |
Minneapolis, MN |
2007-Present |
|
o Spring 2007-Present: Invited member of the Video Game Research Group meetings, PI: Dr. Daniel Kersten o Spring 2007-Present: Member of the Computational Perception and Action Laboratory, PI: Dr. Paul Schrater o Summer 2009: Summer Research
Assistant. Carlson School of Management, Department of Finances. Project
title: ''Corporate board's decisions on CEO hiring: matching and learning'',
PI: Dr. Rajesh K. Aggarwal. o April 2008: Wrote IRB application ''Human sequential decision-making
under uncertainty'' that included all-online experiments, from collecting
consent forms to compensation payment through PayPal. o Summer 2008: Summer Research
Assistant. Project title: Indoor Magnetic Wayfinding for the
Visually Impaired, PI: Dr. Paul Schrater |
|||
University of Minnesota, Twin Cities |
Minneapolis, MN |
2007-2009 |
|
o Spring 2008: Teaching Assistant for ''Quality Assurance and Process Improvement'' and ''Software Testing and Verification'' and ''Exploring Dynamic Languages'' o Fall 2007: Teaching Assistant,
''Statistical Pattern Recognition'' o Summer 2007: Summer Research Assistant, Computational Perception and Action Lab, PI: Dr. Paul Schrater o Spring 2007: Teaching assistant ''Software engineering II'' and ''Quality Assurance and Process Improvement'' |
|||
University of Santiago |
Santiago, Chile |
2004-2006 |
|
o Instructor: ''Distributed Systems'' (4 semesters), ''Computer Game Development'' (1 semester), ''Optimization Methods for Engineering'' (1 semester) and ''Graduation Seminar I'' (1 semester). These courses required preparation of classes, homework, tests, projects and holding office hours; some projects required students' presentations. Average anonymous student evaluation: 6.7 out of 7. o Undergrad thesis advisor: Advised three students in yearlong research thesis projects. Two of them
graduated with highest distinction (maximum grade in thesis project, 7 out of
7) |
|||
Security SA (Holding company), Software engineer |
Santiago, Chile |
2004-2005 |
||
o Participated as Software Engineer and data modeler of large-scale data warehousing project that consolidated data from 11 businesses and about 300,000 customers combined o Responsible for learning and understanding decision makers' needs from different business models, including banking, insurance, insurance brokerage, mutual funds, real estate and travel agency o Participated in early development of data mining infrastructure on top
of the data warehouse for cross-selling and up-selling |
||||
Repsol YPF (Latin American oil company), Multimedia designer |
Santiago, Chile |
Sept 2003 |
||
o Developed multimedia interactive presentation for chief of Chilean ads
campaign to be presented at Repsol YPF International Summit in Buenos Aires,
Argentina |
||||
CONAF & GTZ, Multimedia designer |
Santiago, Chile |
July 2003 |
||
o Developed multimedia presentation in German and Spanish about
sustainability of forestry in Southern Chile |
||||
Telefonica Movistar (world's third-largest mobile company), Developer |
Santiago, Chile |
2002-2003 |
||
o Developed online simulator of 20 cell phones based on Adobe Flash. The
simulators were used 10,000 times per day on average by both the company's
Customer Support Department and customers |
||||
Journal article |
|
Acuna, D., Parada, V. (2010) ''People exploit the structure of the
computationally intractable Traveling Salesman Problem to find near-optimal
solutions'', PLoS ONE (In press) |
|
Conference article |
|
Acuna, D., Schrater, P. (2009) ''Improving Bayesian Reinforcement Learning using transition abstraction'', ICML/UAI/CLT Workshop on Abstraction in Reinforcement Learning 2009 Acuna, D., Schrater, P. (2009) ''Structure learning in human sequential decision-making'', NIPS 2008 Acuna, D., Schrater, P. (2008) ''Bayesian modeling of human sequential
decision-making on the Multi-Armed Bandit Problem'', COGSCI 2008 |
|
Submitted |
|
Acuna, D., Schrater, P. ''Structure learning in human sequential
decision-making'', PLoS Computational
Biology |
|
Other publications (non-competitively peer-reviewed) |
|
Acuna, D., Green, C.S., Schrater, P. (2010) ''The rational control of aspiration in learning'', COSYNE 2010 (Abstract and poster presentation) Acuna, D., Green, CS., Schrater, P. (2010) ''Decision-making in unbounded environments using nonparametric Bayesian Reinforcement Learning'', NIPS 2010 Workshop on Bounded-rational analyses of human cognition: Bayesian models, approximate inference, and the brain (Poster presentation) Acuna, D., Parada, V., Schrater, P. (2009) ''Skill acquisition and performance on the Traveling Salesman Problem'', Center for Cognitive Science, Spring Research Day (Poster presentation) Acuna, D., Schrater P. (2009) ''Structure learning in human sequential
decision-making'', COSYNE 2009 (Abstract and poster presentation) |
|
In preparation |
|
Acuna, D., Green, C.S., Schrater, P. ''The rational control of aspiration in learning'' Acuna, D., Schrater, P. ''Dynamic factorization of unknown Markov
decision processes using Beta process and Pitman-Yor process'' |
|
Invited talk, (June 2010) "Rational analysis of human sequential decision-making under uncertainty and human problem solving", Department of Brain and Cognitive Sciences, MIT
Roadmap Talk NIH NPCS Fellowship (April 2009) ''Rational analysis of human sequential decision-making under uncertainty'', Department of Neuroscience, University of Minnesota
NIPS 2009, NIPS 2010, CogSci 2009
Training Committee member of the Center for Cognitive Science, organized panel discussion on ''Job hunting, hiring process and setting up a new lab in academia''
Pre-doctoral full member of the Center for Cognitive Science, University of Minnesota
Cognitive Science Society
Bayesian modeling and learning, control theory, Markov chain Monte Carlo methods, mathematical programming, operations research, computational complexity theory, evolutionary optimization methods (''meta-heuristics''), linear and non-linear hierarchical and non-hierarchical regression using Bayesian statistics or maximum likelihood estimation
o MATLAB: Toolbox: Curve fitting, database, genetic algorithm and direct search, optimization, parallel computing, statistics, symbolic math; Simulink, Stateflow; connection to Winbugs
o C/C++: code optimization, standalone dynamically or statically linked libraries, arbitrary-precision arithmetic.
o Mathematica: symbolic mathematics, graphics, dynamic manipulation, demonstrations; automated theorem proving (Theorema package); parallel programming
o Flash & PHP: development of web-based psychological experiments (see http://www.cs.umn.edu/~acuna/game/)
o Winbugs: modeling and learning of complex static and dynamic Bayesian networks including nonparametric objects such as Dirichlet, beta and Bernoulli processes
o R: multilevel/hierarchical regression, exploratory and confirmatory data analysis; connection to Winbugs
o Others: Infer.NET framework, Actionscript, Maple, C#, Java, Scheme, LISP, Haskell, Prolog, Excel VBA; Game programming (2D & 3D)
Available upon request
The views and opinions expressed in this page are strictly those of the page author.
The contents of this page have not been reviewed or approved by the University of Minnesota.