Machine Learning from Scratch

I have implemented common ML algorithms with Numpy, and verified the correctness of my implementations using sk-learn.

Inside each algorithm's code, you'll always find this 4-step experiment:

1. load the relevant dataset
2. solve the problem using sk-learn and 1.
3. solve the problem using my own implementation and 1.
4. assert that 2. and 3. are equal, by:
   • comparing the accuracy on the entire dataset in a classification scenario.
   • taking one sample from the dataset randomly and comparing the predictions in a regression scenario.

Datasets

• For binary classification, I used the Pima Indians Diabetes.
• For regression, I used the Boston Housing Dataset.