Automating Machine Learning: Works Every Time
October 24, 2019
Automated machine learning, or AutoML, is the natural next step in the machine learning field. The technique automates the process of creating machine learning models, saving data scientists a lot of time and frustration. Now, InfoWorld reports, “A2ML Project Automates AutoML.” Automation upon automation, if you will.
An API and command-line tools make up the beta-stage open source project from Auger.AI. The company hopes the project will lead to a common API for cloud-based AutoML services. The API naturally works with Auger.AI’s own API, but also with Google Cloud AutoML and Azure AutoML. Writer Paul Krill tells us:
“Auger.AI said that the cloud AutoML vendors all have their own API to manage data sets and create predictive models. Although the cloud AutoML APIs are similar—involving common stages including importing data, training models, and reviewing performance—they are not identical. A2ML provides Python classes to implement this pipeline for various cloud AutoML providers and a CLI to invoke stages of the pipeline. The A2ML CLI provides a convenient way to start a new A2ML project, the company said. However, prior to using the Python API or the CLI for pipeline steps, projects must be configured, which involves storing general and vendor-specific options in YAML files. After a new A2ML application is created, the application configuration for all providers is stored in a single YAML file.”
Krill concludes his write-up by supplying this link for interested readers to download A2ML from GitHub for themselves.
Cynthia Murrell, October 24, 2019
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
