Incorporating Pruning Methods in GAN and Diffusion-based Models

Due to the widespread use of credit cards as a payment method, numerous fraudulent credit card transactions occur, resulting in significant losses for financial institutions. At a high level, the fundamental objective of the project is to improve the performance of fraud detection systems by providing high-quality synthetic data generated by improved deep generative models. This research aims to reduce the size of deep generative models to decrease their computing cost and/or improve the quality of generated data. Then, we up sample fraud patterns with synthetic training data to improve the performance of machine learning classification algorithms.

Faculty Supervisor:

Hamid Usefi

Student:

Partner:

NASDAQ Canada Inc

Discipline:

Computer science

Sector:

Artificial Intelligence; Finance and Insurance; Technology

University:

Memorial University of Newfoundland

Program: