Unrolled Optimization & Matrix Completion

We started this project with Dr.Muhammad Tahir (Linkedin) in Summer 2023 and it concluded in May 2024. A brief compilation of what we did is shown below.

• Self-taught fundamental methods and techniques in Advanced Signal Processing
• Implemented some popular Deep Learning Algorithms (Github Link)
• Did a reading course on High Dimensional Data Analysis and Compressed Sensing. Primary text followed was Book Link by John Wright and Yi Ma.
• Self-taught foundations of preliminaries like Duality Theory and optimization techniques like Augmented Lagrange Multiplier - ALM to understand the problem formulation and solve Matrix Completion.
• Replicated results of following papers Paper 1, Paper 2 and prominent MC algorithms - (Github Link)
• Completed and refined a proposed unfolded ALM algorithm ConvMC-Net for standard matrix completion problem. Github Link
• Proposed unfolded M-estimation (Paper 3) based algorithm ConvHuberMC-Net for robust matrix completion in the event of impulsive Gaussian noise. Github Link