Modeling and diagnosing the exercise of market power in the wholesale electricity industry | Active Learning on Graphs

Label prediction on graphs, i.e., the prediction of the labels of the vertices of a given graph using the labels of a subset of vertices, is a problem that commonly occurs in many areas of machine learning and data analysis.
In this talk I will describe a simple and efficient algorithm, S^2, that achieves this goal. The algorithm functions sequentially in that at each step, the vertex to be labeled is selected based on the structure of the graph and all previously gathered labels. I will demonstrate theoretically that the number of queries S^2 makes is optimal in terms of a novel parametrization of the space of all possible problems. Finally, I will also show some promising performance of this algorithm on synthetic and real world data.
(This is joint work with Prof. Jerry Zhu and Prof. Rob Nowak. )
—————————-
Jesse’s Talk:
Title: Modeling and diagnosing the exercise of market power in the wholesale electricity industry

Diagnosing the exercise of market power in the wholesale electricity industry poses a special difficulty not found in other industries. Transmission line congestion provides a method of exercising market power that leads firms to play randomized strategies in equilibrium. Statistical procedures relying on pure equilibrium play thus fail to see the exercise of this type of market power. I will present a statistical method for recognizing the exercise of transmission-related market power through such randomized strategies.