Discover BigQuery ML, build your own CREATE MODEL stmt
summary
Demo shows how to train models inside BigQuery using SQL: issue a CREATE MODEL statement to train multiclass logistic regression, k-means clustering and Matrix factorization (product recommendations), or ARIMA forecasting; models train in minutes (example run ~7 minutes), BigQuery streams ingestion at ~3 GB/s, supports importing TensorFlow models up to 250 MB and running inference in-SQL, and calls external APIs via remote functions with batched requests to reduce cold-start cost.
AI-written summary · about
chapters (11)
- 0:00Start
- 2:10E-commerce Workloads and data models
- 5:34Where to use BigQuery? A few standard use cases.
- 6:18A Petabyte Query execution
- 12:37Type of models in BigQuery ML
- 14:14Model selection guide
- 16:27Create a binary logistic regression model
- 22:33Matrix Factorization: Problem definition
- 25:46Creating a TABLE for access to Cloud Storage FILES
- 29:08Run inference by using a classification model
- 35:16Analyze object tables by using remote functions
related talks
32mMachine Learning with Google Cloud Platform
25mSQL and TensorFlow: Enabling smart queries on a data warehouse
27mScaling ML to the next level: Hosting thousands of models on SageMaker
06mGoogle BigQuery for enterprise
26mManaging ML Models @ Scale
52mGenerate ML predictions without writing any code
21mWhat If We Could Use Machine Learning Models as Tables?
31mIn-database machine learning: Automate, scale, deploy