Comorbidity Clusters in Autism Spectrum Disorders: A Time-Series Analysis using Electronic Health Records. Finale Doshi-Velez, Yaorong Ge, Isaac Kohane. Pediatrics. To appear.
Bayesian Nonparametric Methods for Reinforcement Learning in Partially Observable Domains. Finale Doshi-Velez, David Pfau, Frank Wood, Nicholas Roy. Transactions of Pattern Analysis and Machine Intelligence. To appear.
Improving Safety and Operational Efficiency in Residential Care Settings with WiFi-based Localization. Finale Doshi-Velez, William Li, Yoni Battat, Jun-geun Park, Ben Charrow, Dorothy Curtis, Sachi Hemachandra, Bryan Reimer, Javier Velez, Cynthia Walsh, Don Fredette, Nicholas Roy, Seth Teller. Journal of the American Medical Directors Association 13(6), July 2012, 558-563. preprint
Reinforcement Learning with Limited Reinforcement: Bayes Risk for Active Learning in POMDPs. Finale Doshi-Velez, Joelle Pineau, Nicholas Roy. AI Journal, Volumes 187–188, August 2012, 115–132. preprint
A Bayesian nonparametric approach to modeling motion patterns. Joshua Joseph, Finale Doshi-Velez, Albert Huang, Nicholas Roy. Autonomous Robots 31(4): 383-400 (2011). preprint
Spoken Language Interaction with Model Uncertainty: An Adaptive Human-Robot Interaction System. Finale Doshi, Nicholas Roy. Connection Science November 2008. preprint
Accelerated Gibbs Sampling for the Indian Buffet Process. Finale Doshi-Velez, Zoubin Ghahramani. ICML 2009 paperslidesposter
Variational Inference for the Indian Buffet Process. Finale Doshi-Velez, Kurt Miller, Jurgen Van Gael, Yee Whye Teh. AISTATS 2009 (Best Paper Nominee) paper, slides (Presented by Kurt Miller). See below for extended tech report version and code.
Reinforcement Learning with Limited Reinforcement: Using Bayes-Risk for Active Learning in POMDPS. Finale Doshi, Joelle Pineau, Nicholas Roy. ICML 2008 paper, slides , video
The Permutable POMDP: Fast Solutions to POMDPs for Preference Elicitation. Finale Doshi, Nicholas Roy. AAMAS 2008 (Best Paper Nominee) paperslidesposter
Collision Detection in Legged Locomotion using Supervised Learning. Finale Doshi, Emma Brunskill, Alex Shkolnik, Tom Kollar, Khash Rohanimanesh, Russ Tedrake, Nicholas Roy. IROS 2007 paper
Efficient Model Learning for Dialog Management. Finale Doshi, Nicholas Roy. HRI 2007 paperslidesvideo
Phenotype discovery using combined curated ontologies and electronic health data. Finale Doshi-Velez (joint work with Ryan Adams, Isaac Kohane). MUCMD 2013.
Transfer Learning by Discovering Latent Task Parameterizations. Finale Doshi-Velez* and George Konidaris*. NIPS 2012 Workshop: Bayesian Nonparametric Models for Reliable Planning and Decision-Making Under Uncertainty paperslides
An Analysis of Activity Changes in MS Patients: A Case Study in the Use of Bayesian Nonparametrics. Finale Doshi-Velez, Nicholas Roy. NIPS 2011 Workshop: Bayesian Nonparametrics, Hope or Hype? paperposter
Reports of the AAAI 2011 Spring Symposia. Mark Buller, Paul Cuddihy, Ernest Davis, Patrick Doherty, Finale Doshi-Velez, Esra Erdem, Douglas Fisher, Nancy Green, Knut Hinkelmann, Mary Lou Maher, James McLurkin, Rajiv Maheswaran, Sara Rubinelli, Nathan Schurr, Donia Scott, Dylan Shell, Pedro Szekely, Barbara Thönssen, Arnold B. Urken. AI Magazine 32(3): 119-127 (2011).
Bayesian Nonparametric Approaches to Reinforcement Learning in Partially Observable Domains. Finale Doshi-Velez. AAAI Doctoral Consortium 2010 paperslides
Nonparametric Bayesian Methods for Finding Software Bugs. Finale Doshi and Jurgen Van Gael. CRISM Workshop on High Dimensional Data, 2008. poster
Thinking Machines. Marc Deisenroth, Finale Doshi, and Mate Lengyel. Cambridge Horizons Seminar, 2008. (for a broader audience) poster
Reinforcement Learning with Limited Reinforcement: Using Bayes Risk for Active Learning in POMDPs. Finale Doshi and Nicholas Roy. ISAIM 2008 paperslides
Learning User Models with Limited Reinforcement: An Adaptive Human-Robot Interaction System. Finale Doshi and Nicholas Roy. LANGRO 2007 paperslides
Efficient Model Learning for Dialog Management. Finale Doshi and Nicholas Roy. AAAI 2007 Spring Symposium paperslides
Model Learning for Dialog Management. Finale Doshi and Nicholas Roy. NIPS 2006 Workshop on Reinforcement Learning. poster
Bayesian Nonparametric Approaches for Reinforcement Learning in Partially Observable Domains. PhD Thesis, MIT, 2012 thesis and slides; if those documents are too long, check out the 2.5 minute musical summary
The Indian Buffet Process: Scalable Inference and Extensions. Masters Thesis, Cambridge, 2009 thesis
Variational Inference for the Indian Buffet Process. Extended version of AISTATS 2009 paper with all equations fully described and derived. tech report
Efficient Model Learning for Dialog Management. Masters Thesis, MIT, 2007 thesis
Here are slides from a few of my presentations. For a list of my invited talks, please see my CV.
Presentation at the CSAIL Student Workshop on Bayesian Nonparametric Approaches to Reinforcement Learning in Partially Observable Domains. (Includes more introductory material than conference presentations.) September 2010. (Best Paper Award Winner) slides
Presentation to the MIT Spoken Language Systems group on POMDPs for dialog management. (Includes more introductory material than conference presentations.) October 2006. slides
iPOMDP model and test-bench (Matlab). This is a rather large set of files that contain both the iPOMDP code (with flags for FFBS/EM as well as several action-selection strategies) and several benchmark domains. Not as tidy as my other code, but you can use it to run many of the iPOMDP experiments from my thesis. code
Samplers for nonparametric correlated latent feature models (Matlab). Contains Gibbs samplers for the models described in Nonparametric Correlated Latent Feature Models for linear-Gaussian, binary, and exponential likelihood models. samplers and utils (needed for the samplers).
Gibbs sampling for the IBP (Matlab). Contains two different collapsed samplers (including the fast sampler described in Accelerated Gibbs Sampling for the Indian Buffet Process), one semi-collapsed sampler, and a fully uncollpased sampler for the linear-Gaussian model. samplers and utils (needed for the samplers). November 25, 2009: Changed test matrix format to handle large, sparse matrices AND fixed MH bug when sampling adding new features
Variational inference for the IBP (Matlab). Contains simplified code described in Variational Inference for the Indian Buffet Process; current version contains only the linear-Gaussian likelihood model. vibp
Point Based Value Iteration (PBVI) POMDP solver (Matlab). pbvi