Title: Ensemble Return Forecasting: Practical Challenges and a Unified Theory for Solving Them.
Bio:
Sri Gollamudi is currently on garden leave from his role as Head of Risk Research at Citadel Global Quantitative Strategies, where he specialized in quantitative portfolio construction and risk modeling. Prior to Citadel, he held the same position at Two Sigma Investments. His research interests are in the general areas of buyside quantitative finance and machine learning.
Sri began his career at Bell Labs, conducting research and developing products in wireless communications before transitioning into finance, a move which included a stint at Bloomberg Portfolio and Risk Analytics. He holds a Ph.D. in Signal Processing and Communications Theory.
Abstract:
Asset return forecasts are key to quantitative portfolio construction, and a common strategy for producing a strongly predictive forecast is to combine several weak forecasts from different models. In practice, these models often produce varied forecasts: some forecast only subsets of assets or asset baskets, others provide buy/sell signals, rank expected returns, or define uncertainty regions. How can we sensibly combine such diverse forecasts? I present a new unified theoretical framework, derived from first principles, to ensemble these and other types of return forecasts without requiring additional modeling assumptions or parameters to estimate. I also provide numerical algorithms with proven convergence to efficiently compute the solution.
