Datamation content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
See the full list of Machine Learning Solutions
Bottom Line
Google is an acknowledged leader in machine learning as well as deep learning and AI. Its Google Cloud ML Engine is designed to deliver highly advanced training and prediction services, either individually or together. The machine learning framework, which is designed for data scientists and developers, supports scikit-learn, XGBoost, Keras, and TensorFlow. The latter is used in Google products ranging from Google Photos to Google Cloud Speech.
The solution addresses online prediction through a serverless, fully managed hosting model that is designed to respond in real time with high availability and automatic scaling. Batch Prediction works with asynchronous applications. It scales to perform inference on large volumes of production data.
For large enterprise customers who need top performance, Google Cloud ML Engine is a leading choice.
Product Description
Google Cloud ML offers a robust environment for developing and managing machine learning. The platform runs machine learning training and predictions at scale through independent processes. It incorporates automatic resource provisioning and monitoring so that data scientists can manage CPUs, GPUs and TPUs at maximum efficiency.
A proprietary technology called HyperTune allows data scientists to manage thousands of tuning experiments in the cloud by automatically detecting and tuning deep learning hyperparameters. The platform uses a server-side preprocessing framework that allows user to send raw data to models in production in order to reduce local computation and prevent data skew.
Cloud ML allows data scientists and developers to import existing models and move them into production on Google Cloud without Docker containers and other tools.
Overview and Features
User Base
Data scientists and developers.
Interface
Web UI with command line.
Scripting Languages/Formats Supported
Uses Jupyter Notebooks and Python-based toolsets.
Formats Supported
Most major data formats can be used but datasets must be converted for use in the ML engine. This may require services such as BigQuery, Cloud DataProc, Cloud Dataflow, and Cloud Dataprep.
Integration
Works with Scikit-learn, XGBoost, Chainer, Keras, and TensorFlow. Uses REST API for managing projects.
Reporting and Visualization
Extensive audit logging, including admin activity logs and data access logs. Google offers internal tools and connects to third party tools that can provide rich visualizations. These include Java and Python.
Pricing
Google has established a pricing model that is based on resources used and training hours. It uses categories that range from less than 1 cent per hour to more than $31 per hour.
Google ML Engine Overview and Features at a Glance:
| Vendor and features | Google Machine Learning Engine |
|---|---|
| ML Focus | Hosted platform that runs ML training jobs and predictions at scale. |
| Key Features and capabilities | Automatic resource provisioning and monitoring for CPUs, GPUs and TPUs. Strong open source support for tools and scripting languages. |
| User comments | High ratings for functionality and integration. |
| Pricing and licensing | Based on resources and training hours used. Ranges from 1 cent per hour to $31 per hour for various services. |
