Machine Learning for Finance

My research interest is in understanding financial market data using tools from signal processing and machine learning. In addition, my interest is in viewing the financial market itself as an information processing system and study asset price dynamics from the perspective of learning agents.

Search algorithms

My interest is in algorithms that solve optimization problems in a distributed fashion. Questions that arise in this context concern the relation between local and global search, the organizitation of collaboration among search agents, and the communication between agents when coordinating the task.

Conversation models

Two of my recent papers focus on the mechanism behind mutual understanding in conversations. A central question is how much modeling of the conversation partner is necessary in order to make sense of a stream of talk. In a related project, we decide whether a collection of weblogs belong to the same web--conversation.

Selected Publications

Global nonlinear control

Many concepts of linear control theory can be extended to the nonlinear domain provided one restricts oneself to a neighborhood of the equilibrium. But what happens beyond that neighborhood?

Multiple Models

If the parameters of a system are time--varying a mixture of models may be employed to identify and store the parameter values assumed by the system over time. We introduce the method and discuss an application to the analysis of fund exposures to risk factors.

Conversation Studies

How do participants in a conversation understand each other? My interest is in the cooperative mechanisms that establish coherence in a stream of talk. In the papers below (presented at the Yale workshop for adaptive and learning systems in '15 and '17) I discuss two of the main control variables: timing and choice of content.

Neural Networks

This is a book chapter on neural networks for the control of complex systems (first author: K.S. Narendra). We review important milestones and discuss fundamental questions from a systems theoretic perspective. These questions are resurfacing as the new field of data science attempts to capture structure in large and complex data sets using deep learning.