- BigQuery ML and TensorFlow: Data Warehouse + Machine Learning enables smart analytics
- BigQuery MLとTensorFlow: データウェアハウスと機械学習で実現する「賢い」データ分析
45 min
Data analysts, data scientists and database engineers. Intermediate level (requires no ML expertise)
BigQuery is Google's petabyte scale data warehouse. It's new feature BigQuery ML realizes smart data analytics with the power of machine learning, from simple logistic regression to deep learning prediction with TensorFlow In this session we will see the technology enables a powerful "data warehouse + ML" solution.
BigQuery is Google's fully managed, petabyte scale, low cost enterprise data warehouse. By leveraging the petabit network and tens of thousands of servers in Google datacenter, BigQuery allow you to execute your SQL with as a Massively Parallel Processing query with hundreds of CPU cores and disk storages, scanning ang aggregating terabytes of data in seconds.
BigQuery has new feature BigQuery ML that let you create and use a simple Machine Learning (ML) model as well as deep learning prediction with TensorFlow model. This is the key technology to integrate the scalable data warehouse with the power of ML. The solution enables variety of smart data analytics, such as logistic regression on large dataset, similarity search and recommendation on images, documents, products or users, by processing feature vectors of the contents. Or you can even run TensorFlow model prediction inside BigQuery.
In this session, we will discuss how BigQuery ML and TensorFlow can be used to build a powerful solution with the cloud data warehouse and ML.
- What is BigQuery (10 min)
- Smart analytics with the signatures (15 min)
- ML Engine for smarter analytics (10 min)
- Large scale demand forecast with BigQuery (10 min)
BigQueryは、Googleが提供するフルマネージドの大規模データウェアハウスサービスです。BigQueryに備わるユニークな機能、User Defined Function (UDF)を用いると、JavaScriptコードを記述してSQL関数を新たに定義が可能です。このUDFを用いてコンテンツの特徴ベクトル処理を実装することで、画像やドキュメントの類似検索やリコメンド等、機械学習の力を活かした様々な「賢い」検索が実現します。BigQuery内部でニューラルネットワークを動作させることも可能です。このセッションでは、BigQueryとTensorFlow、ML Engineにより、データウェアハウスと機械学習を組み合わせた強力なソリューションを作る方法を紹介します。