Hello! I am Amir Khalilian, a machine learning and signal processing engineer with a deep commitment to advancing interpretable, scalable algorithms for real-world applications. My background bridges electrical engineering, computer vision, and computational neuroscience, with experience developing and applying advanced optimization, deep learning, graph-based signal processing methods. My research and engineering work have focused on designing innovative frameworks for neural signal analysis, video and image restoration, and speech decoding, resulting in multiple peer-reviewed publications, patents, and successful collaborations across academia and industry.
Currently as a postdoctoral researcher at NYU Langone Health, I leverage my expertise to investigate the neural mechanisms of speech production and perception using electrophysiological data, developing novel tools to map and decode brain activity. Beyond research, I am passionate about creating solutions that integrate rigorous mathematical foundations with machine learning to produce explainable deep-learning technologies. Whether building neuroprosthetic systems, enhancing image processing pipelines, or designing new brian connectivity analysis techniques, I aim to deliver engineering solutions that are both interpretable and transformative. Here, I post updates about my research and other topics of interest.
July 2025: I will be presenting a turtorial at IEEE MIPR 2025, Directly Parameterized Neural Network Construction for Generalization and Robustness, August 6th-8th San Jose CA, 2025.
June 2025: I will be presenting an IEEE SPS Webinar with Dr. Nikola Janjušević on Directly Parameterized Neural Network Construction for Generalization and Robustness in Imaging Inverse Problems, July 17th 11:30am EST. Register with the link!
Feb. 2025 GroupCDL: Interpretable Denoising and Compressed Sensing MRI via Learned Group-Sparsity and Circulant Attention published in IEEE Transactions on Computational Imaging.
Dec. 2024: A corollary discharge circuit in human speech, published in PNAS. Read the press release here.
A solver for the NMF problem with missing data points. ...
This model provides the means to separate the data into two classes ...
We forgo the additive assumption and instead propose a formulation of an overlaying model, which acknowledges that the foreground object is overlaid on top of the background and is occluding it ...
We present an automatic pipeline designed to detect and correct striping artifacts while minimally degrading the unknown artifact-free image ...
Scanning acoustic microscopy (SAM) is an imaging modality used to obtain 2D maps of acoustical and mechanical properties of soft tissues and uses ultrasound transducers operating at very high-frequencies. ...