Mustafa Anil Kocak

     Research Assistant

      Department of Electrical and Computer Engineering
      NYU Tandon School of Engineering
      E-mail: kocak(at)nyu(dot)edu

 

 

Education

NYU Tandon School of Engineering, Brooklyn, NY
Ph.D. Candidate, Electrical and Computer Engineering, Fall 2011-Present                                                                                                                          
Concentration in Information Theory and Machine Learning
G.P.A. 3.98/4.00
Advisor: Prof. Elza Erkip

 

Bilkent University, Ankara, Turkey
Bachelor of Science, Electrical and Electronics Engineering, June 2011                                                                                                                      
Electrical and Electronics Engineering

G.P.A. 3.86/4.00


Research Interests

My research is focused on reliable machine learning methods, where the uncertainty associated with each prediction is assessed, and actions are taken accordingly. In particular, we develop meta-algorithms to predict the probability of failure for a given classification algorithm and let them refuse to predict when these probabilities are too high. 

Some of the projects I am currently working on or worked in the past are:

  • Conjugate conformal predictors: Well calibrated online predictors based on exchangeability of data streams
  • Variations and extensions of list decoding problem: Communicating Lists over a Noisy Channel
  • Mathematical methods for biological network inference: Designing Group Testing Experiments for Transcriptional Regulatory Networks (Joint work with Prof. Shasha)
  • Study of ensemble classifiers/regressors at non-stationary environments: Capturing the changes in the conditions by combining online learning and conformal prediction frameworks  (Joint work with Prof. Shasha)


Professional Experience

Summer Intern, Nokia Bell Labs

Murray Hill, NJ— Summer 2016 & Summer 2015

Reliable preamble detection methods for initial access with extended coverage with the emphasis of massive scale small-packet communications (IoT/mMTC) for new generation cellular networks.

Supervisor: Dr. Krishna Balanchandran

 

Summer Practice, TUBITAK

Ankara, Turkey — Summer 2010

Literature survey and simple Matlab implementations of image fusion algorithms with compressive sensing.

Supervisor: Dr. Tayfun Aytac
 


Publications

M.A. Kocak, D. E. Shasha, and E. Erkip, “Conjugate Conformal Prediction for Online Binary Classification”, Conference on Uncertainty in Artificial Intelligence, July 2016.

 

M. A. Kocak and E. Erkip, “Communicating Lists over a Noisy Channel,” 52nd Annual Allerton Conference on Communication, Control, and Computing, Urbana-Champaign, IL, October 2014.

 

M. A. Kocak and E. Erkip, “A Secrecy-Throughput Game on Deterministic Interference Channel,” North American School of Information Theory, Fields Institute, Toronto, CA, June 2014 (Poster).


M. A. Kocak and E. Erkip, “A Secrecy-Throughput Game on Deterministic Interference Channel,” North American Information Theory Summer School, Cornell University, Ithaca, NY, June 2012 (Poster).