Mehrdad Aksari Mahabadi

I'm studying computer science at Amirkabir University of Technology.

Previously I worked as a software engineer at Snapp. Snapp is the leading company in providing online taxi services with over 30 million users in Iran. As part of the Routing team, we were responsible for providing estimated time of arrival (ETA) and navigation services. To this end, we integrated traffic data into our routing engines and enabled time-dependent routing, which significantly improved the accuracy of our ETA.

I am also a member of the algorithms, graphs, and machine learning laboratory (AGM) at Amirkabir University. Currently, the main focus of the lab is graph neural networks. We discuss the cutting edge research in this area during weekly sessions.

Projects

In 2013, a deep learning model successfully learned control policy directly from high-dimensional sensory input. The model was applied to seven Atari 2600 games from the Arcade Learning Environment. It surpassed human experts in four of them. Here I used the same method to learn to land on the moon !

Evolucop is an evolutionary agent that masters a simplistic helicopter game. It uses a three layers neural net, to choose wether to push the helicopter. The challenge is in training the net. There are many ways to train a network that can solve this effectively. As the name suggests Evolocop uses Evolution.

writing and training generative adverserial neural networks from scratch in Pytorch

singular Value Decomposition (SVD) is a factorization of a matrix. In this report, we will show how to classify hand-written digits leveraging the aforementioned decomposition. In order to classify an image, we will see which principal component can best estimate the given image.

implementation of a simple artificial neural network framework in pure numpy

Visual Cryptography is a cryptographic technique which allows visual image to be encrypted in such a way that the decrypted information is a visual image. One of the best-known schemes is by Adi Shamir & Moni Naor developed 1994.

Camel is an interpreted programming language written in Golang

Template