Invited Talk by Dr. Saketha Nathm, IIT Bombay on Supervised Class Ratio Estimation
Title: Supervised Class Ratio Estimation
Speaker: Dr. Saketha Nath
Host Faculty: Dr. Vineeth N Balasubramanian
Room No:112, Academic Block-A
In this talk we present learning models for the class ratio estimation problem, which takes as input an unlabeled set of instances and predicts the proportions of instances in the set belonging to the diﬀerent classes. This problem has applications in social and commercial data analysis. Existing models for class-ratio estimation however require instance-level supervision. Whereas in domains like politics, and demography, set-level supervision is more common. We present a new kernel-based method for directly estimating class-ratios using set-level supervision.
We present learning bounds and use the insights obtained, to propose a novel convex formulation that automatically learns the kernel to be employed in the estimation. We design an efﬁcient cutting plane algorithm for solving this formulation. Finally, we empirically compare our estimator with several existing methods, and show signiﬁcantly improved performance under varying datasets, class ratios, and training sizes.
Saketh is an associate professor in the Department of Computer Science and Engineering at IIT Bombay. He is broadly interested in the area of machine learning, with focus on kernel methods and optimization. Currently he is on a sabbatical at Microsoft (Hyderabad), where he is working with the Bing team. For more information, please see: https://www.cse.iitb.ac.in/~saketh/.