Notable Projects

Transfer Learning in Deep Reinforcement Learning

Published:

In this project, we proposed a new nested neural network architecture to be used for deep reinforcement learning. To do this, we combined two recently proposed architectures, Actor-Mimic and Progressive Neural Networks, to speed-up the learning process, to save space, and to improve transfer learning. We have tested our approach on a set of Atari Games and our results show some interesting trends and improvement in comparison to the state-of-the-art models. More details maybe found in our final paper.

Telematics Data Analysis to Study Driving Behavior

Published:

This project is about analysis of drivers behavior by learning their significant driving patterns. The goal is to provide an end-to-end solution to identify risky versus safe drivers. Our identification process is based on extracting driving patterns and then analyzing each pattern within its context. To identify driving pattens, we leverage a novel trajectory segmentation approach which is presented at the 3rd ACM SIGSPATIAL PhD Symposium (2016). Besides, we study each pattern with respect to extrinsic causes which may have relationship with the exatracted pattern, like Traffic condition, Weather condition, Physical properties of routes, etc. The results of current project may be used for Usage Based Insurance (UBI) programs. For more information, please visit our paper which is presented at the 25th ACM SIGSPATIAL Conference (2017).