Bite-sized machine learning models built using Python, Numpy, Scikit-Learn, and TensorFlow - with practical examples to solidfy the theory behind the models. Use these Jupyter (and Colaboratory) notebooks as learning resources to discover the inner workings of essential predictive algorithms.

$ git clone https://github.com/jissac/ScratchML
cd ScratchML

Predictive models used in this repository:

We'll build a logistic regression model from scratch in Python and use it in a binary classification problem - predicting if an image contains a cat or a dog:
Notebook
Blog Post

We'll then extend the binary image cat vs. dog classification task to a CNN architecture:
Notebook
Blog Post

We'll use the Iris dataset to understand how SVMs can be used for classification
Notebook

We'll extend the application of SVMs on the Gender Recognition by Voice dataset
Notebook

We'll use the basic Iris dataset to understand how Decision Trees can be used for classification
Notebook

Blog Post
Notebook

We'll build a GAN trained on the Fashion dataset.
Notebook

Datasets used in this repository.

A collection of recorded voice samples from male and female speakers. Hosted by Kaggle

Articles of clothing classified into 10 labels. Provided by Zalando Research. Hosted by Kaggle

Chest X-Ray images of patients with/without pneumonia. Hosted by Kaggle

A subset of data from the Microsoft Research team hosted by Kaggle with images that contain either cats or dogs.

A classic dataset comparing the structural variation of three related Iris flower species - Iris-setosa, Iris-Versicolor, and Iris-Virginica.