1.4 OML4Py Updates

Describes changes in Oracle Machine Learning for Python for Oracle Database 23ai.

New Features in 23ai

Table 1-1 New Features

Features Description

Support for in-database machine learning Non-Negative Matrix Factorization algorithm, Exponential Smoothing Method algorithm and XGBoost algorithm.
The following functions are new in the package that use in-database algorithms:
  • oml.nmf, Non-Negative Matrix Factorization model
  • oml.esm, Exponential Smoothing Method model
  • oml.xgb, XGboost model
Support for data types that enable you to manipulate date, time and integer.

The following data types are supported by OML4Py:

  • oml.Datetime, to create date.
  • oml.Timezone, to create time and timezone that includes hour, minute, second, microsecond, and tzone.
  • oml.Timedelta, to perform simple arithmetic operations.
  • oml.Integer to represent the integer data type.

Convert Pretrained Models to ONNX Format

Oracle Machine Learning for Python supports ONNX format models. To learn more, see ONNX Pipeline Models : Text Embedding.

Topic:

The following topic tells about new features added in 23ai.

Parent topic: Changes in This Release