|
| | | |
| Apache Spark’s scalable machine learning library containing common learning algorithms, featurization tools, and pipeline utilities. | Executing massive-scale ML model training across a distributed cluster without relying on single-node libraries.
| |
| A core feature transformer in Spark MLlib that merges multiple columns (continuous numbers, one-hot encoded categories, etc.) into a single vector column. | Preparing raw tabular data into the specific dense or sparse vector format expected by Spark MLlib machine learning algorithms. | As a component within Spark MLlib, its execution contributes to the overall compute costs of running your Spark workloads on Managed Service for Apache Spark. |
| A high-level API built on top of DataFrames that helps users chain multiple feature engineering steps and models together. | Grouping Transformers and Estimators into a unified pipeline to ensure identical data processing during training and inference. | Similar to other Spark MLlib components, costs are part of the execution charges for Managed Service for Apache Spark. |
| A service that lets you create and execute machine learning models in BigQuery using standard SQL queries. | Training models directly where your data lives before moving highly complex, iterative distributed jobs to Spark.
| BigQuery ML has its own pricing model. You are charged for - Model training (based on data processed during CREATE MODEL queries),
- Model prediction (standard BigQuery query pricing)
- Model storage (billed at standard BigQuery storage rates).
|
