Cost-sensitive Dynamic Feature Selection
He He, Hal Daum, Jason M. Eisner · 2012
We present an instance-specific test-time dy-namic feature selection algorithm. Our al-gorithm sequentially chooses features given previously selected features and their val-ues. It stops the selection process to make a prediction according to a user-specified accuracy-cost trade-off. We cast the sequen-tial decision-making problem as a Markov Decision Process and apply imitation learn-ing techniques. We address the problem of learning and inference jointly in a simple mul-ticlass classification setting. Experimental results on UCI datasets show that our ap-proach achieves the same or higher accuracy using only a small fraction of features than static feature selection methods. 1.