In this project, we reproduce the work of Jia Luo et. al., and make some modifications. The overall task is to use GPU acceleration technique for faster solving Flow Shop Scheduling problem with genetic algorithm (GA), which is proved to be NP-Hard. By coding with CUDA-C on NVIDIA Tesla V100 GPU card, we have achieved 750 times speedup comparing to Python codes running on Intel Core i7-8750H CPU.

You can access the report via this link.

This project studies using machine learning to predict the 0/1 integer decision variables for capacitated lot sizing problem. It is plausible since as long as we can determine its integer decision variable, the problem becomes linear and can be solved efficiently. We tried several classic algorithms, and the best ones were logistic regression and decision trees.

You can access the report via this link.

The ride sharing company, Uber, wants to optimize its profit by increasing the number of successful rides. In this project, we model Uber's ride sharing system in Atlanta, GA as a continuous time Markov chain (CTMC) in both theoretical and numerical approaches. Also, we consider "Empty Car Routing" policy to improve ride-match efficiency.

You can access the report via this link.