Reza Arabpour's Projects
Config files for my GitHub profile.
Here I will create a website which is will be another way for us know each other better :)))
Used pure C++, OOP, and no pre-defined data structures, created a database capable of storing bank account data and handling different types of queries at O(logN) and creating log files online.
Used clustering algorithms such as K-Means, Fuzzy C-Means, and Density-Based Algorithms like DBScan to cluster three datasets and reported result of the best algorithm after 200 random starting points.[part of my data mining course]
C++ implementation of some of the most well-known data structures in the simplest way. *** Note : This repo will update as the course goes on ! [I am currently TA for Data Structures course] ***
Used different pre-trained models -such as ResNet, VGG, Alexnet- to check if an image is a picture of a dog and then recognize its breed. [Note: This project is part of Udacity's AI programming nano degree.]
Used different pre-trained models like ResNet, VGG, Alexnet, and own designed architecture, to recognize grape breeds based on their leaves. Ended up in about 95% on validation data and 90% on out-of-sample test data. Besides, we tried to improve the accuracy by using image denoising and dimension reduction autoencoder networks.
C++ implementation of some of the most well-known graph algorithms in the simplest way. Mostly in non-OOP style because the algorithms itself and its performance were the main points.
In this project, we first review the paper βA Recurrent Neural Network for Game Theoretic Decision Makingβ by Sudeep Bhatia and Russell Golman, 2014, introducing a computational model to solve the outcome of strategic games. Then we go through the codes which have been implemented in Matlab.
Getting an image of a signature and converting it into the underlying graph structure the signature which can be used for late classification tasks.
Used Python to implement the Soft Regularized Markov Clustering (SR-MCL) algorithm and tested it on an extensive real-world dataset, weighted yeast proteins interactome, for time complexity and performance check beside a small random graph for showing the nodes' predicted clusters.