Google is hoping to push machine learning as a mainstream business tool, with today's release of a set of preview products that take advantage of its cloud platform.
Cloud Machine Learning relies on the open-source TensorFlow library, released late last year.
Google said developers can use the company's tools such as Cloud Dataflow, BigQuery, Cloud Dataproc, Cloud Storage and Datalab, to train its machine learning.
However, it is also offering pre-trained machine learning models with application programming interfaces.
The company has been able to leverage large amounts of user and customer data stored in the applications it offers for the pre-trained models.
Applications that use Google's new Cloud Machine Learning include image search for Google Photos, voice search for the Google App and the Translate application programming interface, as well as the smart reply feature for the online giant's web-based Inbox email client.
Although parts of Google's Cloud Machine Learning offering are open source, such as the TensorFlow tool, the full offering requires a paid subscription.
Google also made its managed Hadoop Mapreduce, Spark, Pig and Hive service Dataproc generally available for customer analytics, charging one US cent per hour and virtual processor.
Cloud-based machine learning aims to give business customers a way to sift through large amounts of data so as to predict outcomes and spot trends.
Despite having access to large amounts of user data, Google has been a latecomer in the machine learning field, with Microsoft having established a lead two years ago with its Azure-based Cortana Analytics Suite.
Amazon Web Services also made a foray into machine learning last year with a fully-fledged offering.
