Machine-Learning-From-Zero-to-Hero

Some example codes of widely used machine learning algorithms

Linear Regression

The original data looks like this:

After running gradient descent or normal equation:

The contour of the cost function:

Support Vector Machine (Using SMO Algo to Optimize)

Linear Kernel

The original data looks like this:

After training the SVM with a linear kernel, we can get the following decision boundary:

Gaussian Kernel

The original data looks like this:

After training the SVM with a Gaussian kernel, we can get the following decision boundary:

K-Means

2D clustering

Image Compression (From Original 0 ~ 255 (8bit) to 0 ~ 16 (4bit))

Principle Component Analysis

2D-1D

EigenFaces

Gaussian Mixture Model

Visualize the clusters

The loglikelihood

Neural Network

Visualize the decision boundary

The loss and training acc

MCMC sampling

Bayesian Linear Regression

(Fully Convolutional Neural Network) U-Net-Based-Heart-Sound-Segmentation

This is the project that I done in the lab during my undergraduate study.

A given heart sound can be segment into four stages which is shown as below:

Our goal is to train a deep neural network to segment a given heart sound into 4 stages.

The architecture that we use is 1-D U-Net used in the literature.