Machine-Learning-and-Data-Science

Overview

This is a repository which contains all my work and experience with machine learning, AI and data science. I am an aspiring data scientist and have learnt that the only way to learn the techniques and methods of anything is to get your hands dirty by doing lots of projects. This repo is my attempt to learn what I am the most passionate about: AI, ML and Data Science. The repository is structured as follows:

Table of Contents

Learning Resources and Articles

A collection of articles, books and research papers spanning different topics in data science and machine learning.

Implementation of Machine Learning Algorithms

Implementations of major machine learning algorithms using only Numpy and Python - without using any other external libraries like sklearn, tensorflow, pytorch etc...

Implementation of Reinforcement Learning Algorithms

Implementations of major RL algorithms from the book - Reinforcement Learning: An Introduction. Also includes Pytorch implementations of deep RL algorithms like DQN, DDPG, A3C etc...

Machine Learning Competitions

Code for the online ML competitions I participated in. This includes Kaggle competitions, Analytics Vidhya Competitions and some random datasets I worked on.

Statistics-101

Important statistical algorithms implemented in R - Maximum Likelihood Estimator, Bayesian Parameter Estimator, Method of Moments etc..

Image Colorizer using Neural Networks

Using neural networks to colorise grayscale images. The neural network is written using only Numpy and Python.

Probablistic Search and Destroy

Use Bayesian Networks to create an agent for optimally searching a target within a simulated environment.

Minesweeper AI Bot

Create an AI bot to play minesweeper at different difficulty levels. Written using Numpy, Python and Matplotlib.

Mazerunner - Analysing AI Search Algorithms

Analysis and comparison of different search algorithms - DFS, BFS and the A* algorithms. Visualisations are done using matplotlib to understand the working of each algorithm.

Music Genre Belief Recognition using Neural Networks

Identify the genre of the song using direct audio files as input and a real-time web-based GUI tool to visualise it. The model identifies and predicts the changing genre of a song as it plays.

Deep Learning

Some small implementations of autoencoders, CNNs, RNNs using tensorflow.

Exploratory Data Analysis

A detailed data analysis and visualisation of some random Kaggle datasets using R's tidyverse package.