Khalil Elkhalil

University Appartments (A1B), 1500 Duke University Road · Durham, NC 27701 · khalil.elkhalil@duke.edu

I am a Tunisian engineer working as a postdoctoral fellow at Duke Microsoft Innovation Hub where I investigate topics related to Deep Learning and Artificial Intelligence in general under the supervision of Vahid Tarokh. I like to learn new things to sharpen my understanding of the world around us, because to understand is to be free as Spinoza once said.

Duke Microsoft

News
  • Our paper on Generative Archimedean copulas has been accepted for publication in UAI 2021. [Paper] [Github]
  • Our paper on Fisher Auto-Encoders is now accepted for publication in AISTATS 2021. [Paper] [Github]
  • Our virtual persentation at ICASSP 2020 is now available online [Presentation]
  • Our paper on kernel ridge regression has been accepted for publication in IEEE Trans. on signal processing. [Paper]
  • I am now a teaching assistant in the Introduction to Deep Learning (ECE 590) course at Duke.
  • I have successfully defended my thesis and I am finally a doctor 😁. Here are the slides of my presentation [Slides]
  • I am now in Brighton attending ICASSP to present our work on LDA with random projections.


  • Education

    KAUST
    MS/PhD.
    Electrical Engineering - Statistics

    GPA: 3.96/4.0

    August 2013 - May 2019
    supcom
    Undergraduate Engineering
    Telecommunication Engineering - Applied Mathematics

    GPA: 14.79/20

    September 2009 - June 2012


    Thesis

  • Random Matrix Theory: Selected Applications from Statistical Signal Processing and Machine Learning
  • K. Elkhalil [Thesis]

    Journals

    • Risk Convergence of Centered Kernel Ridge Regression with Large Dimensional Data
    • K. Elkhalil, A. Kammoun, X. Zhang, M. S. Alouini and T. Al-Naffouri [arxiv] [PDF] [code]

    • A Large Dimensional Study of Regularized Discriminant Analysis Classifiers
    • K. Elkhalil, A. Kammoun, R. Couillet, T. Y. Al-Naffouri and M. S. Alouini [arxiv] [PDF] [code]

    • Measurement Selection : A Random Matrix Theory Approach
    • K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [arxiv] [PDF]

    • Fluctuations of the SNR at the output of the MVDR with regularized Tyler estimators
    • K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [arxiv] [PDF]

    • Numerically Stable Evaluation of Moments of Random Gram Matrices with Applications
    • K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [arxiv] [PDF]

    • On the Feedback Reduction of Multiuser Relay Networks Using Compressive Sensing
    • K. Elkhalil, M. E. Eltayeb, A. Kammoun, T. Y. Al-Naffouri and H. Bahrami [arxiv] [PDF]

    • Analytical Derivation of the Inverse Moments of One-Sided Correlated Gram Matrices With Applications
    • K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [arxiv] [PDF]

    • Opportunistic Relay Selection With Limited Feedback
    • M. E. Eltayeb, K. Elkhalil, H. Bahrami and T. Y. Al-Naffouri [IEEE]

    Conferences

    • Fisher Auto-Encoders
    • K. Elkhalil, A. Hasan, J. Ding, S. Farsiu and V. Tarokh [Arxiv] [code]

    • Asymptotic Performance of Linear Discriminant Analysis with Random Projections
    • K. Elkhalil, A. Kammoun, R. Calderbank, T. Y. Al-Naffouri and M. S. Alouini [PDF] [code]

    • Regularized Discriminant Analysis: A Large Dimensional Study
    • Xiaoke Yang, K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [IEEE] [PDF] [Slides] [code]

    • Asymptotic Performance of Regularized Quadratic Discriminant Analysis Based Classifiers
    • K. Elkhalil, A. Kammoun, R. Couillet, T. Y. Al-Naffouri and M. S. Alouini [IEEE] [PDF] [Poster]

    • A Blind Antenna Selection Scheme for Single-Cell Uplink Massive MIMO
    • K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [IEEE] [PDF] [Poster]

    • Exact Closed-Form Expression for The Inverse Moments of One-Sided Correlated Gram Matrices
    • K. Elkhalil, A. Kammoun, T. Y. Al-Naffouri and M. S. Alouini [IEEE] [PDF] [Slides]

    • Block Compressed Sensing for Feedback Reduction in Relay-Aided Multiuser Full Duplex Networks
    • K. Elkhalil, M. Eltayeb, A. Kammoun, T. Y. Al-Naffouri and H, Bahrami [IEEE] [PDF] [Poster]

    • Distributed User Selection in Network MIMO Systems with Limited Feedback
    • K. Elkhalil, M. Eltayeb, H. Dahrouj and T. Y. Al-Naffouri [IEEE] [PDF] [Slides]

    • Relay Selection with Limited and Noisy Feedback
    • M. Eltayeb, K. Elkhalil, A. Mas'ud and T. Y. Al-Naffouri [IEEE]

    • Opportunistic Relay Selection in Multicast Relay Networks using Compressive Sensing
    • K. Elkhalil, M. Eltayeb, H. Shibli, H. Bahrami and T. Y. Al-Naffouri [IEEE] [PDF]

    • Joint Channel Estimation and Decision Directed Decoding for OFDM-IDMA Systems over Sparse Channels
    • K. Elkhalil, L. Najjar and Hichem Besbes [PDF]

    Interests

    Apart from research, I enjoy watching movies and reading books. I have just completed reading a book that has to some extent changed the way I look to the world, it is called "Sapiens" a masterpiece of the legendary Yuval Noah Harari . Although, it seems intimidating to most people, it studies the human history from a completely rational view. I don't want to spoil the book, so it is for you to go and read it. I have also been onto a new adventure with another exceptional author Tara Westover and her book "Educated" which describes her uncommon educational journey. Now, I am enjoying the book by Nassim Nicholas Taleb (trader, mathematician and essayist but most importantly a thinker) called the "The Black Swan" which explores the potential impact of rare events (that usually defy conventional wisdom). One of the inspiring and illuminating points raised in the book is that history doesn't crawl but actually jumps pretty much like a stock market. My next adventure will be "Men of Mathematics" and "Birth of a Theorem" by one of my scientific heroes Cédric Villani Sometimes, I like to think about puzzles and riddles, most of the times old interesting problems from the IMO. It is frustrating to do so, but extremely satisfying when you figure out the right answer, you kind of feel smarter if you understand what I mean. One of the most challenging problems which took me a great deal of time is problem number 6 in the IMO of 1988 (ironically the year I was born). Try to spend some time on it, you will feel the greatness when you solve it.

    Sapiens Educated Homodeus Flatland 21 mind genome black swan Men Math Cedric Villani Antifragile EVT Simons

    I also like watching documentaries from time to time. I still remember the first documentary I watched: "Einstein's big idea" broadcasted in 2005 celebrating 100 years after Einstein's special relativity. It is still a good one, go and watch it if you didn't; you will discover many other brilliant scientists such as Faraday, Lavoisier, Lise Meitner and Emilie du Chatelet. This is how I developed a passion for science.

    Einstein

    I am a big fan of Fairuz and Celine Dion two legendary singers. Hedi Jouini is my favourite Tunisian singer ever.

    Hedi_Jouini Fairuz Celine


    Awards & Certifications

    • 1st Place - National entrance examination to the Ecole Normale Superieure de Tunis, 2009
    • KAUST Graduate Fellowship
    • Finalist for the best student paper award in IEEE MLSP conference 2017