Changes in This Release for Oracle Machine Learning for SQL User's Guide

Describes changes in Oracle Machine Learning for SQL User’s Guide for Oracle AI Database 26ai.

New Features

• Computer Vision (CV) models in your database: Integrate CV models in your database and generate embeddings for images. See Image Transformer ONNX Format Models for more details.

• Vector Data Support: Oracle Machine Learning supports vector data type for clustering, classification, regression, anomaly detection, and feature extraction. See Vector Data Type to learn more.

• Oracle Machine Learning for SQL supports ONNX format models with the integration of ONNX Runtime. To learn more, see Integration of ONNX Runtime.

• BOOLEAN data type is supported. For more information, see Convert Column Data Types, Numericals, Categoricals, and Unstructured Text, and Target Attribute.

Enhancements

Oracle AI Database supports in-memory sharing of external initializers for ONNX models, reducing memory usage and improving scalability for concurrent inference workloads. The database loads large initializers once into global memory and shares them across sessions. This feature applies to ONNX models imported with external initializers and supports monitoring and control through dictionary views. See External Initializers and In-Memory Sharing and Enable and Use External Initializers for more details.

Model Views

• Model detail views for ONNX models are introduced. The DM$VX and DM$VM model detail views are added. See Model Detail Views for ONNX Models

• Model view for Exponential Smoothing is enhanced. See Model Detail Views for Exponential Smoothing.