ml-experiment 0.0.3

Machine Learning Experiment Framework

Introduction

This Python package facilitates the fast prototyping of machine learning model with great scalability and flexibility.

Characteristics of this package:

• Flexibility of Feature Engineering: it is convenient to define a function to put to feature-processing pipeline;
• Flexibility of Models: there is no restriction about whether you have to use scikit-learn, TensorFlow, or PyTorch;
• Few Specifications on Models: user only need to worry about the fit and predict_proba;
• Training Job Specifications: features, data locations, model specifications can be specified in a Python dictionary or JSON, facilitating potential MapReduce or parallelism;
• Scalability: data is stored temporarily in disks in batch to save memory space;
• Statistics: statistical measures of the performance of the models and their class labels are calculated;
• Cross Validation: cross validation option is available.
• Ready Adaptation to Production: data pipelines and algorithms can be adapted into production codes with little changes.

There will be tutorials and documentations.

News

• 05/03/2020: '0.0.3' released.
• 04/29/2020: '0.0.2' released.
• 04/24/2020: '0.0.1' released.