oriyonay / kiwiml

A tiny, simple, instructional machine learning library in python

kiwi v0.0.5

A tiny, simple, instructional machine learning library in python, written by Ori Yonay. Mostly written as a personal reference for classic ML algorithms. Follows similar syntax to sklearn :)

To install: pip3 install kiwiml

Current Features:

• Machine Learning Models:
  • KNN
  • Linear Regression
  • Logistic Regression
  • Naive Bayes
  • Perceptron
  • Single Dimensional Analysis (experimental)
• Utilities:
  • Accuracy score function for calculating model accuracy
  • train_test_split
  • A function to plot cost histories
  • PCA Decomposition (will be in its own dimensionality reduction class in the future)
• Error Functions:
  • Mean-Squared Error
  • Cross-Entropy Loss
• Dataset Loaders:
  • Boston dataset
  • Breast cancer dataset
  • Iris dataset
  • MNIST dataset

TODO:

• Error Functions:
  • Add mean absolute deviation
  • Change error functions to take parameters (w, X, predicted) instead of (X, y, predicted) so can be differentiated with autodiff
• Machine Learning Models:
  • SVM
  • Decision Tree
  • Random Forest
  • K-Means Clustering
  • Neural Network Library
    • Fully-Connected layer
    • Convolutional Layer
    • ...
• More dataset importers
• Automatic learning