Home > Research
Computational Intelligence Laboratory
Department of Computer Science, Stanford University
The Computational Intelligence Laboratory (CIL) investigates the foundations of machine learning and automated reasoning. We are a group of 4 Ph.D. students, 2 postdocs, and several M.S. students working on problems at the intersection of statistics, logic, and computation.
Fig. 1: Feedforward architecture used in our hybrid neuro-symbolic system
| Current Projects |
| Bayesian Structure Learning |
Learning the structure of Bayesian networks from sparse, incomplete data. We have developed a novel scoring criterion that incorporates domain-specific structural priors, achieving 40% better reconstruction accuracy than BDe on biomedical datasets.
Funded by NSF IIS-9702847
|
| Hybrid Neuro-Symbolic Reasoning |
Can neural networks learn to respect logical constraints? We are training feedforward networks augmented with penalty terms derived from first-order logic, enabling a form of "soft theorem proving." Early results on the MYCIN knowledge base are promising.
Funded by DARPA HPKB Program
|
| Knowledge Compilation for Planning |
Compiling propositional theories into DNNF (Decomposable Negation Normal Form) for efficient online query answering. Applications in autonomous agent planning under uncertainty.
Funded by ONR N00014-96-1-0718
|
| Inductive Logic Programming |
Extending Progol with statistical significance tests to prevent overfitting in relational learning. Applied to protein structure prediction in collaboration with the Stanford Biochemistry Dept.
Joint work with Prof. R. Altman
|
| Lab Members (1997-98) |
Postdocs: Dr. Kenji Takahashi (Bayesian methods), Dr. Ana Soares (ILP)
Ph.D. Students: David Park (structure learning), Sarah Mitchell (neuro-symbolic), Raj Gupta (knowledge compilation), Lisa Wong (planning)
M.S. Students: Tom Bradley, Jennifer Nakamura, Carlos Reyes
Visiting: Prof. Luc De Raedt (K.U. Leuven, Belgium) — Winter Quarter 1998
|
Interested in joining? I am looking for motivated Ph.D. students for the 1998-99 academic year. Strong mathematical background required. Email me with your CV and a brief description of your research interests.
Home | Research | Courses | Publications | Hot Links
|