Project: Adaptive Machine Learning

Many machine learning problems involve performing inference on statistical models that change over time.  We aim to develop algorithms and software for such dynamic problems.  Thus far, we have designed algorithms for adaptively updating computed inferences (e.g., marginals or MAP configurations) of models where observed evidence, conditional probabilities, and the structure of the dependencies change over time.  We have applied these algorithms to problems in vision and computational biology including protein modeling.  


See our publications.


  • This research is partially supported by funds from Intel Labs and Microsoft Research.