Agam Chopra’s Deep Learning and Data Science repository

What is the Deep Learning repository about?

“Deep learning is a class of machine learning algorithms that uses multiple layers to progressively extract higher-level features from the raw input. For example, in image processing, lower layers may identify edges, while higher layers may identify the concepts relevant to a human such as digits or letters or faces.”

This repo is a collection of all my publicly available Deep Learning projects. Please feel free to fork and program your versions of the code.

Projects in this repository

• Intro to Deep Learning Series of lectures, derivations, code, and other useful materials to get an in-depth and hands-on understanding of parametric nonlinear models such as Deep Learning.

  

• Fully Convolutional Neural Network for Embryo Classification The goal of this project is to correctly classify embryo viability based on a 2D microscope image of the embryo.

  

• X-Ray Pneumonia diagnostic Convolutional Neural Network A Convolutional neural network was developed that analyzes a patient’s chest X-ray image to give an accurate Pneumonia diagnosis and the type of Pneumonia (viral or bacterial) if applicable. ResNet50 architecture gave test accuracy of 1.00 and Val accuracy of 0.98.

  

• Simple Stock Perfomance Predictor is a deep learning model developed to predict whether one should invest and hold a given stock for a year or not for profit.

  Click here to read a brief report for this project.

  

• I have created a visualization of all the activation functions and their first derivatives that I am familiar with using the Desmos graphing tool click here.

Useful resources:

• Neural Networks and Deep Learning by Michael Nielson is a free online book and a great resource to gain intuition about Deep Learning.

• I also recommend enrolling in the Deep Learning Specialization by Andrew Ng on Coursera

• ADAM optimization

• Learning rate decay

• Input normalization

• Swish over LeakyReLU activation function

• Vectorization

• Data Visualization

• Decision threshold optimization

• Random initialization

License

The MIT License