An AI model asset represents the essential parameters of an AI model. These parameters also include the environment where the model is run, the input parameters and the output data, and the business purpose and rules for the model. AI models provide insights for many activities so that you can make decisions faster and easier or provide prediction modeling.
For example, if you have a regular expression that you run on a large set of web browser clickstream data to predict an Internet user's navigation on your organization web site, you can create a business representation of this expression as an AI model in Data Governance and Catalog.
You can rate an AI Model asset to indicate your assessment of the quality of the asset, or add your comments to the asset to provide your inputs on the data present within the asset. This enables different users to collaborate on assets and make collective decisions on using the data. For more information about collaborating on an asset, see the Asset Management help.
If you implement a workflow, you can view the number of open tickets and overdue tickets. Open tickets indicate the tickets pending in progress to track changes to the AI Model. Overdue tickets indicate the tickets beyond due date that require your action. You can also view the lifecycle of the AI Model asset to know the particular stage of the lifecycle in the asset workflow.
Overview tab
On the Overview tab, you can view the system and custom attributes of the AI model along with other basic information about the AI model.
The system attributes are the system-defined properties of the asset. The custom attributes are user-configured properties of the asset that allow you to specify and store specific information relevant to your assets beyond the system-defined attributes.
The following table describes the system attributes and other information about the AI model that are displayed on the Overview tab:
Field
Description
Description
Functional definition of the AI model. The size of the field should not exceed 1 MB.
Reference ID
Unique reference identifier for the AI model.
AI Model Purpose
Purpose of the AI model.
Input
Data elements that represent the input parameters the AI model.
Output
Data elements that represent the output after the AI model is run.
Environment
Indicates whether the AI model belongs to the training, validation or production environment.
Model Rules
Rules that govern the use of the AI model.
Architecture Type
The architecture and algorithms that are used in the AI model.
Libraries
The software libraries that are used by the AI model.
Model format
The format in which the AI model is maintained.
Source Model Repository
The repository in which the source of the AI model is maintained.
Contains
Displays all the AI models that have a direct relationship with the source AI model.
Is Part Of
Displays all the AI models that are part of the source AI model.
Data
Displays the data sets that are associated with the AI model.
Evaluation Metrics
Record additional data that pertain to an AI Model asset. You can use evaluation metrics to compare various models, identify potential biases, and monitor model behavior over time.
An AI Model asset has the following predefined evaluation metrics:
- Drift Score. Measure of degradation of the model. A higher value indicates greater degradation. Use the drift score to determine if the model can provide results within an acceptable range of accuracy. Drift score is measured in percentage.
Enter a value between 0 and 100. Ensure that the value you specify has a maximum precision of 5 digits and a scale of 2 decimal places.
- Bias Score. Measure of bias in the model. A higher value indicates a greater difference between the predicted result and the actual result. Use the bias score to determine if the assumptions in the model are correct. Bias score is measured in percentage.
Enter a value between 0 and 100. Ensure that the value you specify has a maximum precision of 5 digits and a scale of 2 decimal places.
In Metadata Command Center, your administrator can define additional metrics. For more information, see the Administration help in Metadata Command Center.
Associated AI Model Versions
Displays the AI Model Core Version technical asset that was used to create the AI Model asset.
Stakeholders
Users responsible for the AI model or interested in the AI model.
Asset Groups
One or more asset groups assigned to this asset. Asset groups allow users access to a set of assets in the organization.
You can assign asset groups to an asset if you are the stakeholder of the asset or if your organization administrator has granted you the Manage Access permission on the asset through access policies in Metadata Command Center.
Data Marketplace
Details of an asset's presence in Data Marketplace.
If an asset is added to a data collection in Data Marketplace, you can view the details of the data collection.
For more information about how you can connect Data Governance and Catalog to Data Marketplace, see the Set Up Data Marketplace help in Data Marketplace. For more information about how you can add an asset to a data collection, see the Working With Data Collections help in Data Marketplace.
Created By
User that created the AI model and the date on which the AI model was created.
Updated By
User that last updated the AI model and the date on which it was updated.
Model Lineage
On this tab, you can view a visual representation of the data sets and other AI models that are associated with the AI model.
For more information about model lineage, see the Asset Discovery help.
Relationships tab
On this tab, you can switch between grid view or graph view to view all the relationships that the AI model has with other assets. Create relationships with other assets if they can be used together to make business or governance decisions.
The following table describes the fields that appear when you switch to the grid view on the Relationships tab:
Field
Description
Name
Name of the related asset.
Type
Type of the related asset.
How It Is Related
Relationship between the other assets and the asset that you have opened.
Hover the mouse over the arrows to see if the relationship is direct, indirect, inbound or outbound. The term "this asset" refers to the asset that is open.
Hierarchy
Hierarchical path of the asset.
Note: This column is hidden by default. To display the column, right-click on the column header and select Hierarchy.
You can perform the following activities on the Relationships tab:
•Click the Filters icon on the top of the tab to apply filters to narrow the list of relationships. You can apply filters such as asset type to view related assets of a specific type, direction to view related assets of a particular origin, lifecycle to view assets in a particular stage of the lifecycle, relationship type to view assets that are directly or indirectly connected to the AI model, or relationship label.
The direction of the asset can be inbound or outbound. A relationship created from the AI model to the related asset is called an outbound relationship. A relationship created from the related asset to the AI model is called an inbound relationship.
•Search for a related asset using the search box.
•Click the Add icon to add a new related asset, or the Delete icon to delete an existing related asset.
Note: You can delete only outbound direct related assets.
Relationship example
The following image shows a sample asset and the different types of relationships with other assets:
In the image, the business term 'Delivery Cost' is the open asset. The Relationships tab shows the grid view of all the assets related to 'Delivery Cost'.
1The highlighted policy asset 'Gov Policy 01' is related to the open asset through an inbound relationship, which is indicated by the orange arrow pointing to the left direction. This indicates that the relationship between the business term 'Delivery Cost' and the policy 'Gov Policy 01' originated in the policy asset 'Gov Policy 01'. The relationship type 'is Regulating' indicates that the policy asset is regulating the business term.
2The highlighted system asset 'Customer Call Analysis Reports' is related to the open asset through an outbound relationship, which is indicated by the blue arrow pointing to the right direction. This indicates that the relationship between the business term 'Delivery Cost' and the system 'Customer Call Analysis Reports' originated in the business term 'Delivery Cost'. The relationship type 'is a Strategic Source for' indicates that the business term is the preferred or recommended source of data for the system asset.
For more information about asset relationships, see the Relationships topic in the Asset Discovery help.
Data tab
View data sets that are used in the AI model.
On this tab, you can search for a specific data set and use filters to narrow the list of data sets. To add a new data set to the AI model, click the Add icon. To remove a data set, select the data set and click the Delete option from the Action menu. You can also update the assocation value for a data set. To update the association, select the data set and select the applicable association value in the Action > Update Association menu.
The following table describes the properties of the Data tab:
Field
Description
Name
Name of the data set.
Type
Type of the data set.
Association
The type of association between the AI model and the data set.
This field can have one of the following values:
- Unknown. The nature of association between the AI model and the data set has not been determined.
- used to Train model. The data set is used to train the model.
- used to Validate model. The data set is used to validate the model.
- used to Train and Validate model. The data set is used to train and validate the model.
Glossaries
Glossary assets that are associated with the data set.
Classifications
Data classifications associated with the data set.
For information about configuring data classification, see Administration in the Metadata Command Center help.
Note: To see this information, the organization administrator must grant the View Data Classifications feature privilege to your user role in Informatica Intelligent Cloud ServicesAdministrator. For more information about roles and users, see the Asset Management help.
Sensitivity
Data sensitivity levels for the data elements that belong to the data set.
Hierarchy
Hierarchy of the data set.
Description
Description of the data set.
Reference ID
Unique identifier for the data set.
Asset Groups
One or more asset groups assigned to the Data Set asset. Asset groups allow users access to a set of assets in the organization.
You can assign asset groups to an asset if you are the stakeholder of the asset or if your organization administrator has granted you the Manage Access permission on the asset through access policies in Metadata Command Center.
Business Name
Business name of the data set.
Inferred Data Type
Inferred data type of the data set.
Stakeholders tab
View users that have been designated as stakeholders for the AI model. Stakeholders are authorized users who are responsible for the model, and can approve or reject change requests for the model. If your role has the required permissions, you can add or remove stakeholders on this tab.
The following table describes the properties of the Stakeholders tab:
Field
Description
Name
Name of the user or user group.
Role
Role of the user or user group in the organization.
Email
Email address of the user.
Note: This field does not appear for user groups.
User Origin
The origin of the users. For example, when a user logs in using SSO, the value for this field is displayed as SAML.
You can also use filters to narrow your search for stakeholders.
Tickets tab
View change request tickets that users have created for the AI model.
The following table describes the properties of the Tickets tab:
Field
Description
Name
Name of the ticket.
Description
Description of the ticket.
Type
Type of ticket.
Status
Status of the change ticket.
Current Task Owner
The stakeholders who are assigned to the tasks in the ticket.
You can filter the list of tickets by selecting available filters options. For more information about tickets, see the Workflows and tickets topic in the Asset Management help.
History tab
View changes that users have made to the AI model. The History tab provides an audit log that you can use for compliance requirements.
The following table describes the properties of the History tab:
Field
Description
Date
Displays the created and last updated dates of the asset.
Event Type
Displays the type of the event that is performed on the asset:
- Asset. Displays changes in the name, description, certification, or the reference ID of the asset.
- Relationship. Displays additions, deletions, and the updates of the relationships between the assets.
- Stakeholder. Displays additions and deletions of stakeholders to the asset.
Action
Displays the following actions that are performed on the asset:
- Create. Displayed if the asset is created.
- Updated. Displayed if the asset is updated.
- Deleted. Displayed if the asset is deleted.
Current Attributes
Displays the recently changed attributes of the asset.
Previous Attributes
Displays the earlier attributes of the asset.
Changed By
Displays the user who has performed any action on the asset.
Asset Name
Displays the name of the asset.
Asset Path
Displays the origin path of the asset.
Asset Type
Displays the type of the asset.
From Asset Type
Displays the source asset type for which the change is applicable. This is applicable when the event type is Relationship.
To Asset Type
Displays the target asset type for which the change is applicable. This is applicable when the event type is Relationship.
