Outlining my non-research, academic projects as well as hacks here. My research work can be seen here.

Implementation of Machine Learning algorithms and Deep Learning architectures :

  • Implemented standard machine learning algorithms like Bayes classifier, Naive Bayes classifier, k-nearest neighbours, Parzen window, k-means clustering, Expectation Maximization etc from scratch.
  • Implemented and trained state of the art image classification Neural Network architectures like LeNet, AlexNet, ResNet etc in PyTorch.
  • Studied and implemented Generative Models such as Generative Adversarial Networks (GANs), Wasserstein GANs, Variational Autoencoders, Gaussian Mixture Models etc.

Simultaneous Perturbation for Stochastic Approximation (SPSA)

  • Implemented Wulff et al.’s paper titled “SPSA for Layer-Wise Training of Deep Networks” from scratch.
  • Analyzed and compared the convergence and performance of SPSA with other popular algorithms like Adam, RMSprop, SGD etc for optimizing neural networks.

Field of View Tester and Contrast Tester Javascript Modules

  • Developed two Javascript modules for testing field view of vision and visible contrast in visually impaired.
  • Designed an algorithm to generate random points over complete area and fine-tuned the boundary of vision.
  • Implemented a logarithmic algorithm which converges in 7-8 steps to accurately find visible contrast in RGB.

Stock Prediction using Support Vector Regression

  • Used linear, polynomial and radial basis function to implement kernelized learning using Support Vector Regression and experimented with different thresholds.
  • Predicted and analysed stock prices by training monthly stock dataset fetched using Yahoo Finance API.

Data structure for Mobile Phone Tracking System

  • Implemented a central server consisting of phone exchange nodes at different levels and mobile phones of each exchange.
  • Created hierarchical Routing Map Tree Structure using k-ary trees, linked lists, sets etc to solve the mobile phone routing problem.

PS3 controlled Robotic Arm with Gripper

  • Designed a robotic arm with 4R-four-bar mechanism having 3 degrees of freedom to grip small objects.
  • Programmed Mega ADK to actuate shoulder and arm using servos and turntable base using Stepper motor.