Spectral Feature Extraction for RL in Financial Markets
Combining frequency-domain with reinforcement learning for asset trading.
PI / Advisor: Dr. Catia Silva
Institution / Department: University of Florida, Department of Electrical and Computer Engineering
Timeframe: August 2025 – April 2026
Research Focus
This project focuses on reinforcement learning for financial trading, with an emphasis on building agents that make decisions from informative market features rather than raw price movements alone. We explore how spectral and time-domain representations of market behavior can improve generalization, robustness, and interpretability in trading policies.
Responsibilities
I conducted the literature review, designed and implemented the reinforcement learning experiments, and analyzed how different feature and reward formulations affected trading performance. I also synthesized the results into a formal research paper, including the experimental methodology, evaluation, and discussion of the project’s broader implications.