Google Adds AutoML to Kaggle
Google has integrated its automated machine learning tools into Kaggle, the data science development site it acquired in 2017, with the goal of expanding access to the site’s more than 3.5 million users.
Google has been introducing its AutoML on the Kaggle competition site, then benchmarking the results. During a four-week competition, AutoML bested more than 90 percent of daily submissions until the end of the competition. Ultimately users were able to achieve improved results more quickly and without domain expertise of supervision, Google said.
Competitors using the Google’s AutoML tools “spent very little time on data prep, and virtually no time on feature engineering, model selection, and hyperparameter tuning,” Devvret Rishi, Google Cloud’s product manager for Kaggle, noted Monday (Nov. 4) in a blog post.
AutoML tools tend to have a defined set of features, including: acquiring and prepping data; engineering features from the data; selecting the best algorithm; tuning the algorithm; and deployment and monitoring of production models.
Google’s Cloud AutoML assists users in building custom machine learning models for a variety of data science tasks ranging from machine vision to natural language processing to analyzing structured data.
Google said integration of AutoML with Kaggle is similar to its addition earlier this year of its BigQuery analytics data warehouse with Kaggle Notebooks. This week’s release allows ingesting of data via either a web user interface or a software development kit. The SDK can be used with Kaggle notebooks, Rishi added.