Oracle Machine Learning for Python, with Demos

We learned about the Oracle Machine Learning for Python and how it integrates the advantages of Python with the scalability and performance of Oracle Database while enabling Python functionality on database data as if they were native Python objects.

Users transparently move from Python functions written for single core to overloaded functions leveraging database parallelism and scalability. Oracle Machine Learning for Python allows users to manipulate data in Oracle Database tables and views using Python syntax and functions, but translating Python functionality into SQL for in-database execution.

Users develop and operationalize comprehensive scripts for analytical applications without leaving the Python environment. Directly integrate user-defined Python scripts into applications and dashboards by immediately invoking Python scripts from SQL. This drastically reduces time-to-market by eliminating porting Python code and developing custom infrastructure, while enabling immediate updates to application code.

The Slides used in the presentation can be found in the Resources section below.

Video highlights:
03:30 Oracle Machine Learning for Python Introduction
06:50 Traditional Python and Database Interaction
09:40 OML4Py Features
13:20 Demo of OML4Py in Zeppelin Notebooks
16:36 Demo of AutoML - Automatic Algorithm Selection
20:44 Demo of AutoML - Automatic Feature Selection
22:32 Demo of AutoML - Automatic Hyperparameter Tuning
26:55 Demo of In-Database Machine Learning algorithms
35:30 Demo of Transparency Layer
40:52 Demo of Overloaded Data Visualization functions
45:05 Demo of Data Stores
48:35 Demo of Embedded Python Execution
55:58 Demo of Creating user-defined functions in the Python repository
57:10 Demo of Scoring data and Building Models in parallel
1:12:50 Demo of returning Images from Embedded Python execution
1:15:40 Demo of SQL Developer creating and invoking Python scripts via SQL and PL/SQL
1:18:08 Q&A

AskTOM Office Hours offers free, monthly training and tips on how to make the most of Oracle Database, from Oracle product managers, developers and evangelists.

[hummus_boss]

Ok I built the model and deployed it. So how can I predict on a new dataset? Seriously, this hasn't been explained at all. Even I can't find docs to talk about this. So far, what is the point of all this if my model can't predict on a new dataset?

Also, notebooks need to have imblearn package or at least let us install any package we want.

[pavelsjo]

When will be available? or how can test it?