Timezone: »

Finding the Homology of Decision Boundaries with Active Learning
Weizhi Li · Gautam Dasarathy · Karthikeyan Natesan Ramamurthy · Visar Berisha

Wed Dec 09 09:00 PM -- 11:00 PM (PST) @ Poster Session 4 #1210

Accurately and efficiently characterizing the decision boundary of classifiers is important for problems related to model selection and meta-learning. Inspired by topological data analysis, the characterization of decision boundaries using their homology has recently emerged as a general and powerful tool. In this paper, we propose an active learning algorithm to recover the homology of decision boundaries. Our algorithm sequentially and adaptively selects which samples it requires the labels of. We theoretically analyze the proposed framework and show that the query complexity of our active learning algorithm depends naturally on the intrinsic complexity of the underlying manifold. We demonstrate the effectiveness of our framework in selecting best-performing machine learning models for datasets just using their respective homological summaries. Experiments on several standard datasets show the sample complexity improvement in recovering the homology and demonstrate the practical utility of the framework for model selection.

Author Information

Weizhi Li (Arizona State University)
Gautam Dasarathy (Arizona State University)
Karthi Natesan Ramamurthy (IBM Research)
Visar Berisha (Arizona State University)

More from the Same Authors