Skip to content

Usage and database design blueprint #2

@pebeto

Description

@pebeto

This issue contains the usage design, and will serve as the starting point of the database design.

The idea behind this project is to bring a simple but effective machine learning tracking system. The initial idea is the next:

  1. A User is the entity reflecting an individual capabilities
    • Users can have access to projects
    • Users can't have individual access to anything below a project
    • There are two types of users: Administrator and Regular
      • Administrators are able to manage projects and are non-limited by permissions
      • Regulars actions to anything below a project are limited by permissions. They can't manage projects
      • The permissions are: create, read, update, and delete
  2. A Project is an user directory of experiment executions
    • Projects contain the user permissions
    • Projects can contain experiments
    • A project deletion brings a cascade deletion
  3. An Experiment is the core entity in the machine learning project structure
    • Experiments can contain resources (documents, artifacts, graphics, etc...)
    • Experiments can contain iterations
    • An experiment deletion brings a cascade deletion
  4. An Iteration is the product of a model execution. It saves results information
    • Iterations can contain parameters and metrics
      • A parameter is a value passed at the start of the model training process
      • A metric is a measurement got during or after the model training process
    • Iterations can contain results
    • Iterations can contain notes
    • An iteration deletion brings a cascade deletion

Metadata

Metadata

Assignees

Labels

documentationImprovements or additions to documentationhelp wantedExtra attention is needed

Type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions