Project Grant 2228910

This two-year, $200,000 project grant from the National Science Foundation's Division of Information and Intelligent Systems, under the Computer and Information Science and Engineering program (CFDA 47.070), will support research to develop new approaches for integrating domain knowledge into deep learning models. Specifically, the awardee, the Research Foundation of the City University of New York for Brooklyn College, will demonstrate the value of incorporating domain expertise into structured prediction for temporal deep learning models in complex domains like music. The project will experiment with distilling established music pedagogies and expressions of skilled musical practice to help machines learn more efficiently by mimicking human learning processes. It aims to develop models that can learn with less data and counteract biases in training datasets. Outcomes will include best practices for encoding pedagogical and performance expertise into machine-readable formats to reduce complex musical signals into their essential structural components in a more accurate and interpretable manner. Findings may also apply to natural language understanding and other domains involving complex temporal data.