Exploring machine learning techniques

This repository contains python notebooks of some machine learning algorithms.

Notebooks:

1. Basic python (Numpy, matplotlib, scikit-image): a patch extraction techique, that may be used for future medical patch-based classification/segmentation, is provided at the end.

2. Feed-forward neural networks: a touch of theory + a digit classification application.

3. Convolutional neural networks: Definition + Cifar10 object classification.

4. Fully Convolutional neural networks: Retinal blood vessel segmentation.

5. Generative adversarial networks: MNIST image generation.

6. Gradient descent for deep learning: contains the following

   • The standard Gradient Descent (GD) algorithm
   • The GD+Momentum algorithm
   • The AdaDelta algorithm
   • The Adam algorithm

7. Loss Landscape: visualizing the loss landscape on the MNIST database using 2 random directions.

8. Dictionary learning: contains the following

   • The Iterative Shrinkage and Thresholding Algorithm (ISTA)
   • The Coordinate Descent (CD) algorithm for $\ell_1$ sparse coding
   • The Block-Coordinate Descent (BCD) algorithm for dictionary learning
   • The Online Dictionary Learning (OLD) algorithm

Environment: The following software/libraries may be needed to run all the notebooks:

• TensorFlow 1.12.0
• numpy 1.15.1
• h5py 2.7.0
• matplotlib 2.0.2
• scikit-image 0.14.1
• scikit-learn *.19.1