Stochastic Systems Group
Home Research Group Members Programs  
Demos Calendar Publications Mission Statement Alumni

SSG Seminar Abstract


Performance guarantees for sensor management

Jason Williams
SSG, MIT


Many machine learning problems involve sensors which can be actively controlled to alter the information received and utilized in the underlying inference task. In this talk, we discuss performance guarantees for greedy heuristic algorithms for adaptive sensor control, where the inference criterion is mutual information. Examples confirm that the bounds are tight, and counterexamples illuminate the barriers preventing wider application. Finally, we demonstrate the performance of our tighter online computable performance guarantees through computational simulations.



Problems with this site should be emailed to jonesb@mit.edu