Ismail Alkhouri

Ismail Alkhouri is a Researcher in the Department of Computational Mathematics, Science and Engineering (CMSE) at Michigan State University's College of Engineering. His work focuses on machine learning, optimization algorithms, and medical image processing, with a particular emphasis on adversarial robustness and deep learning applications. He has contributed to advancements in inverse problems, medical imaging reconstruction, and algorithmic solutions for combinatorial optimization and control systems.

His research explores novel techniques such as diffusion-based models for image restoration, adversarial purification for MRI reconstruction, and dataless reinforcement learning approaches. Key areas of investigation include the theoretical underpinnings of untrained deep models and the development of efficient network compression strategies. His work bridges formal methods (e.g., controller synthesis for temporal logic) with practical machine learning challenges in healthcare and computer vision.

Notable contributions include the UGoDIT framework for unsupervised group deep image prior, the SITCOM method for inverse problem sampling, and studies on adversarial attacks in hierarchical classification systems. His research often addresses real-world problems in medical imaging and robust AI systems.

Current work trends emphasize improving training efficiency of diffusion models, developing robust algorithms against adversarial perturbations, and advancing combinatorial optimization through differentiable techniques. He actively explores applications of these methodologies in healthcare diagnostics and automated decision-making systems.