variable-protocols 1.3.0

A protocol for machine learning variables

Variable Protocols

A way to specify common variables of data inputs in the field of data science and machine learning.

Installation

pip install variale_protocols

Specifications:

Principles

• Only one representation for each variable type
• See variable_protocols/variable_protocols.py

Guideline

• Use functions in variables.py to create variables
• 'VariableGroup' groups different type of Variables
• 'VariableTensor' groups same type of Variables

Examples

see test/examples.py

Goals

• specify common variables of data inputs in the field of data science and machine learning.
• structural typing, only sum types and product types (In the general sense, not ML family ADT)
• there should be one way to type one variable group
• good error messages
  • error (messages) oriented programming
• serialization in the future (probably in Dhall)

NonGoals

• compatability with any current stuff
• easy to optimize with current computer architectures
• easy to comprehend (simple is not easy)
• support nominal typing(only labels)