machine-learning

A repository which contains all of my snippets and projects related to Machine Learning.

• classical_ml : Consists of all the basic machine learning algorithms. All of them have been first coded without using sklearn in order to understand how the algorithm actually works. Later, they have been coded using sklearn.
  Libraries used : numpy, sklearn and pandas

• deep_learning : Consists of snippets of various deep learning libraries like Tensorflow and Keras. It also includes my projects in deep learning.
  Frameworks used : Tensorflow, Keras, Theano

The various projects that I have done are:

• Image Classifier model :
  

1. First made my own Image Classifier model using Tensorflow and Keras on a small dataset. Achieved 90% accuracy. Needed to use Image Augumentation and a heavy Dropout in order to achieve this.
   
2. Applied Transfer Learning on the VGG 16 model by training my model just on the final fully connected layer of VGG16 model. Accuracy > 95%

• Google Dinosaur using CNN and Reinforcement Learning:
  

1. Model is still in development phase (It has some bugs). Want to develop a model that is able to play the Google Dinosaur Game on its own.

• Sentiment Analysis of Movie Reviews:
  

1. Given any movie review, the model is able to predict whether the review was "positive" or "negative".
   
2. Accuracy > 80%

Sources

• Andrew NG's Machine learning course on coursera: The most basic course. Everybody does it. Hello world of machine learning.
• Stanford's CS231n: Introduction to Deep learning and Convolutional Neural Networks
  
• Sentdex' playlist for Machine Learning with Python : Awesome Channel in general for any Python related stuff. This playlist especially focuses on how to use Python for Machine Learning.
• Jeremy Howard's fast.ai : An awesome MOOC which teaches the different frameworks in Python available for Deep Learning.
• Andrew NG's new course on Deep learning(paid) : New course being offered by Andrew NG in Deep learning on coursera. It is paid though. However financial help is available like for any other coursera course.

Installation Tutorials (Just Google it):

• Tensorflow
• Keras
• sklearn, numpy, pandas : Can be installed using pip

NOTE: In order to train various deep learning models, it is recommended that you have a GPU which supports CUDA framework to speed up things.