Vikash K. Mansinghka
[my initials] at mit dot edu
My research focuses on uncertain reasoning, inductive
learning and stochastic computation. I try to make
computers smarter by generalizing the classic tower of computing
abstractions to the broader setting of probabilistic (rather than
deductive or Boolean) inference and stochastic (rather than
deterministic) languages, algorithms, and machines. My research draws
on tools from probability theory, Bayesian statistics, computational
statistical mechanics and formal approaches to knowledge
representation (both procedural - e.g. Scheme - and declarative -
e.g. graphs, grammars and logics).
In particular, I work on:
- Languages for compositionally specifying stochastic generative
processes that can do useful computational work and represent
uncertain beliefs
- Models written in these languages for finding patterns and making
actionable predictions, both commercially and cognitively motivated
- Algorithms for efficiently solving the inference and optimization
problems that arise in reasoning, learning and acting according to
these models
- Machines, both virtual and physical, for executing these
algorithms naturally, robustly and efficiently, based on distributed,
stochastic circuits
My academic work is driven by the engineering challenge of building
intelligent, autonomous machines and the scientific challenge of
explaining human mental life and behavior in computational terms. I am
a member of the Computational
Cognitive Science Group at MIT's
Brain and Cognitive Sciences
Department and Computer Science
and Artificial Intelligence Laboratory, where I am advised by Professor Joshua
Tenenbaum. My other academic interests include the integration of
programming with pedagogy, the procedural formalization of mathematics
and physical law, and the clear (and hopefully inspiring and
empowering) communication of the methods and products of human
knowledge to the general public.
I am also involved in an ongoing effort to develop several commercial
applications of this research, focused on specific problems of
inference and optimization at the core of many consumer internet and
biotechnology businesses, and to produce a commercially backed
software and hardware platform for natively stochastic computing.
Manuscripts in preparation (email me for a copy):
- Mansinghka, Jonas, Tenenbaum. Stochastic Digital Circuits for Probabilistic Inference.
- Mansinghka, Roy, Bonawitz, Jonas, Tenenbaum. Approximate Monte Carlo Inference by Systematic Stochastic Search
Some peer reviewed publications, reverse chronological order:
- [PDF]
Goodman, Mansinghka, Roy, Bonawitz, Tenenbaum (to appear). Church: a
language for generative models. In Proceedings of the
Twenty-Fourth Conference on Uncertainty in Artificial Intelligence
(UAI) 2008.
- [PDF]
Goodman, Mansinghka, Tenenbaum. Learning grounded causal
models. In Proceedings of the Twenty-Ninth
Annual
Conference of the Cognitive Science Society (COGSCI) 2007.
(The experiment demo is
here.)
[2007 Cognitive Science Society computational modeling prize
for Perception and Action.]
- [PDF]
Mansinghka, Roy, Rifkin, Tenenbaum. AClass: An online algorithm for
generative classification. In Proceedings of the 11th
International Conference on Artificial Intelligence and Statistics
(AISTATS) 2007.
- [PDF]
Roy, Kemp, Mansinghka, Tenenbaum. Learning annotated hierarchies
from relational data. In Neural Information Processing Systems
(NIPS) 19. December, 2006.
- [PDF]
Mansinghka, Kemp, Tenenbaum, Griffiths. Structured priors for
structure learning. In Proceedings of the Twenty-Second Conference
on Uncertainty in Artificial Intelligence (UAI) 2006.
- [PDF]
Shafto, Kemp, Mansinghka, Gordon, Tenenbaum. Learning cross-cutting
systems of categories. In Proceedings of the Twenty-Eighth Annual
Conference of the Cognitive Science Society (COGSCI) 2006.
- [PDF]
Goodman, Bonawitz, Baker, Mansinghka, Gopnik, Wellman, Schulz,
Tenenbaum. Intuitive theories of mind: a rational approach to false
belief. In Proceedings of the Twenty-Eighth Annual Conference of
the Cognitive Science Society (COGSCI) 2006.
Feel free to read my old homepage.