Module core.Project

Handles the serialization and loading of projects. A project involves the neural network and the attached train / test data, plus other informations.

Functions

Project.new () Constructor.
Project:load (folder) Loads the fields from a folder.
Project:save (folder, export, fixed) Saves the project to an existing folder (the folder creation is handled by the ProjectSaveDialog).
(private) load_helper (folder) Loads a project from a folder.
(private) save_helper (prj, folder, fixed) Saves a project to a folder.

Tables

Project Format of the in memory project.
network.nv Format of the serialized project file.


Functions

Project.new ()
Constructor.
Project:load (folder)
Loads the fields from a folder.

Parameters

  • folder: Folder where the project is stored

Return value:

true, or nil plus an error message in case of errors

See also:

Project:save (folder, export, fixed)
Saves the project to an existing folder (the folder creation is handled by the ProjectSaveDialog). Overwrites existing files.

Parameters

  • folder: Folder where the project will be stored
  • export: If true, the project will not be marked as 'not changed'
  • fixed: If true, the project will be saved in fixed mode. Also, it will not mark as 'not changed'

Return value:

true, or nil plus an error message in case of errors

See also:

(private) load_helper (folder)
Loads a project from a folder. Can raise errors, so use with pcall.

Parameters

  • folder: Path of the project folder to be loaded

Return value:

Project table with the loaded info
(private) save_helper (prj, folder, fixed)
Saves a project to a folder. Can raise errors, so use with pcall. If no error has been raised, then the save succeeded.

Parameters

  • prj: Project to be saved
  • folder: Folder wher the project will be saved
  • fixed: true if the project will be saved in fixed mode

Tables

Project
Format of the in memory project.

Parameters

  • changed: True if the network changed since the last save
  • description: User-defined description of the network
  • data: Array of {desc, obj} that holds the project's data files
  • network: The neural network
  • folder: The path where the network was loaded
  • training_params: Holds the last used training parameters (can be nil)
  • training_status: Holds the last training status (can be nil)
  • weight_init: Holds the last weight initialization params
network.nv
Format of the serialized project file.

Parameters

  • fixed: True it the network has been saved to a fixed point format
  • description: User-defined description of the network
  • data: Array that holds the project's data file descriptions. The files must be named index.data, like 1.data, 2.data, 3.data, relative to the description array
  • filters: Array that holds the project's data filters (to be used to filter input and output neuron values)
  • file: The filename (relative) of the serialized neural network from the core library

Valid XHTML 1.0!