Meraj Hashemizadeh

M.Sc. CS Student at University of Montreal

About me

PROFESSIONAL PATH

Hello there! My name is Meraj Hashemizadeh; I'm a Master student studying Computer Science at the University of Montreal & MILA. Before joining MILA, I received my B.Sc. in both Electrical Engineering and Physics from Sharif University of Technology in 2019.

My current research interests include the use of tensor methods in deep learning, optimization for machine learning, and high-dimensional statistics.

PYTHON
MATLAB
R
C/C++

Education

ACADEMIC CAREER

M.Sc. in Computer Science

MILA - University of Montreal
Sep 2019 - Present

B.Sc. in Electrical Engineering

Sharif University of Technology
Sep 2014 - Jul 2019

B.Sc. in Physics

Sharif University of Technology
Sep 2014 - Jul 2019

Projects

RESEARCH EXPERIENCE

Theory of Deep Learning

“Acceleration and variance reduction in min-max optimization.”  In this work, I reviewed existing algorithms for min-max optimization and proposed an accelerated method for stochastic min-max optimization with last-iterate convergence results.

Probabilistic Graphical Models

“Language Modeling with Spectral HMM Training.”  In this project, we explored the spectral method of training a Hidden Markov Model to predict sequences of emissions. We evaluated this approach compared to the more widely used expectation-maximization technique for training HMMs, and with a non-Markovian model for language modeling, the bidirectional LSTM.

Foundations of Machine Learning

“Deep Image Classification in Fourier Domain.”  Motivated by the success of CNNs in image classification and the fact that convolutions in the spatial domain are equivalent to multiplication in the Fourier domain. We proposed new architectures for image classification without relying on the computationally expensive convolution operation.

B.Sc. Project

“A Low-rank and Sparse Tensor Regression Method for Estimating Hand Trajectories using ECoG Signals.”  In this project, I considered a sparse and low-rank CP structure for the weights of a tensor regression model. The optimization problem was solved with ADMM, and the resulting algorithm was implemented in MATLAB to test the model on the ECoG/ Hand trajectory data.

Honors

PRIZES
Sep 2019

Tuition Exemption Scholarship, University of Montreal

Dec 2017

Fourth prize in the First Iranian BCI Competition, National Brain Mapping Lab, Iran

Aug 2014

Passed the national universities entrance examination with overall rank 182 among 222,000 participants, Iran

Jul 2014

Passed Iran pre university exams with average score of 19.76 out of 20

Sep 2013

Silver medal in The Iranian National Physics Olympiad

Sep 2012

National Elite Foundation Scholarship

Sep 2012

Bronze medal in The Iranian National Physics Olympiad

Get in Touch

CONTACT INFO