Hooshmand Shokri Razaghi

hooshmand shokri columbia

Columbia University
Ph.D. Candidate (2019)

E-mail: hooshmand@cs.columbia.edu

linkedin icon github icon


I'm a Ph.D. candidate in Computer Science at Columbia University (expected graduation date: May 2019). My research advisor is Liam Paninski. I work on machine learning and statistical methods for computational neuroscience. We've been developing spike sorting method and software (YASS) for large multi-electrode array recordings (NIPS paper). I'm currently working on auto-encoding variational inference for latent non-linear dynamical systems applied to modelling neural population activity.

Before, I used to work at CCLS on smart cities projects including forecast and optimization of energy consumption in smart buildings.

My research interests include Bayesian inference, deep generative models, and reinforcement learning.

My resume (as of August 2018).


Education

  • Ph.D. Computer Science, Columbia University, 2014-Present

  • M.S. Computer Science, Columbia University, 2012-2013

  • B.S. Computer Engineering-Software Engineering, Sharif University of Technology, 2007-2012

Employment

  • Graduate Research Assistant, Columbia University, 2014-Present

  • Software Engineering Intern, Google Inc., May 2017-Aug 2017

  • Research Coordinator, Columbia University, Feb 2014-Aug 2014

Projects

Neural Data Analysis

  • Processing dense multi-electrode array recordings of neural activity. We developed a method and a software (YASS) to efficiently and accurately carry out spike sorting (open source code).

Natural Language Processing (NLP)

  • Omnimixture: Enhancing topic models with graph based representation of textual data that leads to new features that are lowly correlated with topic model features.

Smart Cities

  • NYC public transport - Feasibility study for wireless charging technology. The goal of the study was to discover the optimal locations of charging stations for buses powered by electricity using probabilistic modeling and heuristic search.
  • Di-BOSS - comprehensive building operating system that aims to optimize energy consumption and increase security using real-time data streams to make predictions and provide recommendations for future operation.

Teaching

  • Head Teaching Assistant, Discrete Math, Columbia University, September-December 2015

  • Instructor, Introduction to Calculus, Barnard College, Columbia University, June-August 2015

Papers & Presentations

  • YASS: Yet Another Spike Sorter [PDF|Code]
  • JinHyung Lee, David Carlson, Hooshmand Shokri, Weichi Yao, Georges Goetz, Espen Hagen, Eleanor Batty, EJ Chichilnisky, Gaute Einevoll, Liam Paninski, NIPS 2017.
  • Omnimixture: Enriched Topic Modeling [Code]
  • Lauren A. Hannah, Rebecca J. Passonneau, Hooshmand Shokri Razaghi, Ruilin Zhong.
  • Adaptive Stochastic Controller for Smart Buildings
  • Roger N. Anderson, Albert Boulanger, Promiti Dutta, and Ashish Gagneja, Hooshmand Shokri Razaghi, New York Academy of Sciences ML Conference, 2014.
  • Di-BOSS: Digital Building Operating System Solution
  • Roger N. Anderson, Albert Boulanger, Vaibhav Bhandari, Jessica Forde, Ashwath Rajan, Vivek Rathod, Hooshmand Shokri Razaghi, NIPS, 2013.
  • An Efficient Simulated Annealing Approach to Traveling Tournament Problem [PDF]
  • Sevnaz Nourollahi, Kourosh Eshghi, Hooshmand Shokri Razaghi, American Journal of Operations Research (AJOR), Vol.2 No.3, September 2012.
Last Update: August 2018