What If We Could Use Machine Learning Models as Tables?
summary
Claims that machine learning models behave like database tables and shows it running for MySQL and Postgres by exposing ai tables via federated storage. Training happens by inserting a row into a predictors table that points to a SELECT, AutoML builds and tests models, and a SELECT on the model returns predicted value, confidence and explainability metrics. Works with MariaDB, ClickHouse, MongoDB and SQL Server sources, supports bringing your own models on scikit-learn/pytorch, and includes a GUI for data quality warnings and insight.
AI-written summary · about
chapters (8)
- 0:00Start
- 1:33An Ai Table and a Regular Table
- 7:44Train the Machine Learning Model
- 10:06To Get the Passenger Satisfaction out of the Trained Model
- 12:21Important Features To Have while Training the Model
- 17:16Graphical User Interface
- 18:42Can I Train the Model with Other Other Data Sources and Query Them from the Database with Sql Queries
- 19:50Can I Tune Parameters of the Model or Can I Use Other Machine Learning Machine Learning Libraries like Scikit-Learn
related talks
28mCharge Your Database with Predictive ML Capabilities using Open-Source
31mIn-database machine learning: Automate, scale, deploy
50mLeveraging AI as a PostgreSQL DBA
54mUnderstand & Where to use AI & Machine Learning 101
01h 00mPostgreSQL and Artificial intelligence
45mForget SQL! Create your own AI Data Analyst
51mDiscover BigQuery ML, build your own CREATE MODEL stmt
37mIn database Machine Learning with Python in SQL Server