MSCS Candidate Apoorva Tamaskar Successfully Defends Deep Reinforcement Learning Thesis
On Monday, February 1, MS in Computer Science candidate Apoorva Tamaskar (MET ’21) defended his thesis, “Data Stream Query Optimization Using Deep Reinforcement Learning,” supervised by Assistant Professor of Computer Science Kia Teymourian, coordinator of programming languages.
As our world has become more connected, Tamaskar posited, we have generated more and more data streams. The rise in network traffic and proliferation of inter-connectivity has steepened the challenge of analyzing data and extracting key information. Now, more than ever, successfully executing queries to extract insights is essential to the process of making quick and efficient assessments and decisions.
Deep reinforcement learning is a technique that combines the flexibility of reinforcement learning, strong prediction properties, and precise mathematical model of Deep neural networks that serves as a robust and promising method to reduce query processing times.
While prior research has focused on query optimization on data streams, Tamaskar showcased the novel use of Deep reinforcement learning. Thanks to the continuous nature of data streams, conventional approaches are inapplicable, given the expansive increases in processing time. The method implemented in Tamaskar’s thesis was generic enough to apply to a wide range of data streams and queries to optimize the query processing time.
Tamaskar’s thesis, acting as a proof of concept, explored the use of Deep reinforcement learning on a query from Linear Road Benchmark data, demonstrating the method to be effective in picking optimal query plan and reducing the query execution time.
Congratulations to Apoorva Tamaskar for successfully defending the thesis and research contribution!