Akshay Rangamani

Akshay Rangamani is an Assistant Professor in Data Science at the New Jersey Institute of Technology (NJIT). His research focuses on deep learning theory, neural network dynamics, generalization bounds, and optimization algorithms. He explores topics such as neural collapse, low-rank layers in neural networks, and the interplay between architecture design and performance.

His work bridges theoretical insights with practical applications in signal processing, computer vision, and recurrent network analysis. Notable contributions include studies on skip connections enhancing associative memory capacity, weight decay effects on neural collapse, and generalization guarantees for interpolating kernel machines.

Rangamani has published widely in top conferences and journals, with over 190 citations and an h-index of 19. His research has been highlighted in media outlets, emphasizing advancements in understanding deep classifier dynamics and neural network training mechanisms.