pangjac / JediMLSuite

Implement ML algorithm will give you a much better opportunity to see why things work.

JediML

• A simple machine learning algorithm suite, inspired by Professor Hsuan-Tien Lin, National Taiwan University, machine learning foundation technique course.
• The suite aims to help beginners to understand foundational machine learning algorithm, including Perceptron, Regression, Support Vector Machine, Decision Tree etc.

Installation

$ pip install JediML

- Note : The suite is still under development. Please check back if error occurs. Will fix it as soon as possible!

Algorithm

• Perceptron Learning

  • Perceptron Learning Algorithm Binary/Multi Classification
  • Pocket Perceptron Learning Algorithm Binary/Multi Classification

• Regression

  • Linear Regression Learning
  • Ridge Regression Learning
  • Kernel Ridge Regression Learning

• Logistic Regression

  • Logistic Regression Learning
  • L2 Regularized Logistic Regression Learning
  • Kernel Logistic Regression Learning

• Support Vector Machine

  • Hard Margin SVM
  • Polynomial/ Soft Polynomial Kernel SVM
  • Gaussian / Soft Gaussian Kernel SVM
  • Probabilistic SVM
  • Least Squares SVM
  • Support Vector Regression SVM

• Decision Tree

  • Decision Stump Based
  • AdaBoost Stump Based
  • AdaBoost Decision Tree
  • Gradient Boost Decision Tree
  • Decision Tree Classification/Regression
  • Random Forest Classification/Regression

• Neural Network

  • Neural Network Binary Classification

• Accelerator

  • Linear Regression Accelerator

• Feature Transform

  • Polynomial Feature/Legendre Feature Transform

• Validation

  • N Fold Cross Validation

• Blending

  • Uniform Blending for Classification/Regression
  • Linear Blending for Classification/Regression

Usage

>>> import numpy as np
>>> import JediML.PLA as pla

>>> your_input_data_file = '/path/to/your/data/file'
# Suite also provide sample data from Professor Hsuan-Tien Lin course material

>>> pla_bc = pla.BinaryClassifier()
>>> pla_bc.load_train_data(your_input_data_file)
>>> pla_bc.set_param()
>>> pla_bc.init_W()
>>> W = pla_bc.train()

>>> test_data = 'Format : Each feature of data x separated with spaces, and the ground truth y at the end of line.'
# assign test data, format like this '0.97681 0.10723 0.64385 ........ 0.29556 1'

>>> prediction = pla_bc.prediction(test_data)

>>> print prediction['input_data_x']
>>> print prediction['input_data_y']
>>> print prediction['prediction']

PEP8

pep8 JediML/*.py --ignore=E501

License

The MIT License (MIT)