Official Content

Represents the complete definition of a model in the context of the GeneXus Cognitive API for custom models.

Members

  • Name: VarChar(64) -- Model's name
  • Dataset (structure)
    • Loader: ObjectName, GeneXus -- Fully qualified object name that generates training data samples.
    • Option (Structure)
      • Batch: Numeric(4.0) -- Describes how many elements will be processed time by time.
      • Size: (structure) -- Proportion reservation in range [0,1].
        • Training: Numeric(5.3)
        • Testing: Numeric(5.3)
      • Seed: Numeric(4.0) -- Useful when you want to repeat an experiment (same random splitting).
  • Input (structure)
    • Features (collection) -- Defines every data input by name and type (i.e. every column names/types).
  • Output (structure)

Notes

  • Dataset.Loader field must be set with the fully-qualified name of your dataset 'generator' object (Procedure or Data Provider).
  • Features' field count in the Input field must be equal to every Data.Input Features' field count of your data-set.
  • Features' field in Definition.Input and Data.Input must be correlative in order to describe the name and type of every feature of your data item (i.e. the k-th element of Data.Input.Feature should have the Type of the k-th element of Definition.Input.Feature).
  • If Data.Input.Features field has more items than Definition.Input.Features field, remain items in Data.Input.Features will be ignored.
  • If Data.Input.Features field has fewer items than Definition.Input.Features, means you have undefined Data.Input.Features and will produce an error.
  • When Name field is omitted in Data.Input.Features' item, the column name will be the Type field name, enumerated in order of appearance (e.g. "MEDIA_1"). In order to data summary being human readable, it is recommended to do not omit the Name field.

Scope

Availability

This data type is available as of GeneXus 16 upgrade 6.



Last update: February 2024 | © GeneXus. All rights reserved. GeneXus Powered by Globant