Skip to main content
Previous sectionNext section

ML Configurations

Important:

IntegratedML is a new feature of InterSystems IRIS®, coming soon to 2020.3.

An ML configuration is a collection of settings that IntegratedML uses to train a model. Primarily, a configuration specifies a machine learning provider that will perform training. Depending on the provider, the configuration may also specify requisite information for connection such as a URL and/or an API token.

You can use IntegratedML without any adjustment to your ML configuration necessary, as %AutoML is set as the system-default ML configuration upon installation.

Caution:

AutoML is currently not supported on Windows. %H2O is set as the system-default ML configuration for IntegratedML on Windows.

Creating ML Configurations

While you can use the system-default ML configuration upon installation, you can also create new ML configurations for model training.

Creating ML Configurations using the System Management Portal

To create an ML configuration:

  1. Log into the Management Portal.

  2. Go to System Administration > Configuration > Machine Learning Configurations.

  3. Select Create New Configuration and enter the following values for fields:

    • Name — The name of your ML configuration.

    • Provider — The machine learning provider your ML configuration connects to.

      • URL — The URL of a DataRobot endpoint

      • API Token — The API token of a DataRobot account

    • Description — Optional. A text description for your ML configuration.

    • Using Clause — Optional. A default USING clause for your ML configuration. See Parameter Customization for further discussion.

    • Owner — The owner of this ML configuration.

  4. Select Save to save this new ML configuration.

To set this new ML configuration as the system-default, see Setting the System Default ML Configuration.

Creating ML Configurations using SQL

You can create a new configuration using the CREATE ML CONFIGURATION command.

Syntax

The CREATE ML CONFIGURATION statement has the following syntax:

CREATE ML CONFIGURATION ml-configuration-name PROVIDER provider-name [ %DESCRIPTION description-string ] [ USING json-object-string ] provider-connection-settings
Copy code to clipboard
Examples

The following examples highlight use of different clauses for your CREATE ML CONFIGURATION statements:

Simplest Syntax

The following command creates an ML Configuration, H2OConfig, that uses the H2O provider. No provider connection settings are needed when connecting to H2O:

CREATE ML CONFIGURATION H2OConfig PROVIDER H2O
Copy code to clipboard
Selecting Training Parameters with USING

The following command creates an ML Configuration, H2OConfig, that uses the H2O provider and specifies a default USING clause:

CREATE ML CONFIGURATION H2OConfig PROVIDER H2O USING {"nfolds": 4}
Copy code to clipboard
See More

To set this new ML configuration as the system-default, see Setting the System Default ML Configuration.

For complete information about the CREATE ML CONFIGURATION command, see the reference.

Setting the ML Configuration

IntegratedML provides the following configurations for immediate use:

  • %AutoML

  • %H2O

  • %PMML

Upon installation, %AutoML is set as the system-default ML configuration. You can use IntegratedML without any adjustment to your configuration necessary. If you would like to specify a different ML configuration to use for your TRAIN MODEL statements, you can do so in one of the following methods:

  • SQL — you can set the ML configuration for your given process

  • System Management Portal — you can adjust the system-default ML configuration

You can see which ML configuration was used for your training run(s) by querying the INFORMATION_SCHEMA.ML_TRAINING_RUNS view.

Setting ML Configuration for the Given Process using SQL

You can use the SET ML CONFIGURATION statement to specify the ML configuration for your given process.

Syntax

The SET ML CONFIGURATION statement has the following syntax:

SET ML CONFIGURATION ml-configuration-name
Copy code to clipboard
See More

See the reference for more information about the SET ML CONFIGURATION statement.

Setting the System Default ML Configuration using the System Management Portal

You can set the system-default ML configuration in the Machine Learning Configurations page in the System Management Portal.

To set the system-default ML configuration:

  1. Log into the Management Portal

  2. Go to System Administration > Configuration > Machine Learning Configurations

  3. Next to System Default ML Configuration, select the ML configuration of your choice.

Maintaining ML Configurations

You can perform the following operations to maintain your ML configurations:

You can see which ML configuration was used for your training run(s) by querying the INFORMATION_SCHEMA.ML_TRAINING_RUNS view.

Altering ML Configurations

You can modify the properties of existing ML configurations.

Altering ML Configurations using the System Management Portal

To alter an ML configuration:

  1. Log into the Management Portal

  2. Go to System Administration > Configuration > Machine Learning Configurations

  3. Select the name of a listed ML configuration and adjust the values of your choice.

  4. Select Save to save this altered ML configuration.

Altering ML Configurations using SQL

You can alter a configuration using the ALTER ML CONFIGURATION statement.

Syntax

The ALTER ML CONFIGURATION statement has the following syntax:

ALTER ML CONFIGURATION ml-configuration-name alter-options
Copy code to clipboard

Where alter-options is one, or more, of the following:

  • PROVIDER provider-name

  • %DESCRIPTION description-string

  • USING json-object-string

  • provider-connection-settings

See More

For complete information about the ALTER ML CONFIGURATION command, see the reference.

Deleting ML Configurations

You can delete ML configurations.

Deleting ML Configurations using the System Management Portal

To delete an ML configuration:

  1. Log into the Management Portal

  2. Go to System Administration > Configuration > Machine Learning Configurations

  3. Find the row of the ML configuration you want to delete and select Delete.

Deleting ML Configurations using SQL

You can delete a configuration using the DROP ML CONFIGURATION statement.

Syntax

The DROP ML CONFIGURATION statement has the following syntax:

DROP ML CONFIGURATION ml-configuration-name
Copy code to clipboard
See More

For complete information about the DROP ML CONFIGURATION command, see the reference.