The Machine Learning Shortcut – AutoML

Machine learning is in high demand! If you run an engineering organization, you’re likely struggling to find qualified employees with Machine Learning experience.

And if you’re a developer, it can be a real challenge to master the depths of Machine Learning in the margins while you manage your existing career.

But what if there was a Machine Learning shortcut that could help us all?

What if you could begin to train your own Machine Learning models without having experience in Machine Learning programming?

Enter Cloud AutoML from Google

Cloud AutoML is a suite of Machine Learning products that enables developers with limited machine learning expertise to train high quality models by leveraging Google’s state of the art transfer learning, and Neural Architecture Search technology.

Using AutoML’s drag-and-drop interface, you can literally deploy trained Machine Learning models in minutes.

How to use AutoML

The AutoML application process is broken up into three main steps:

  1. Train
  2. Deploy
  3. Serve

First, you upload your data set and then click train.

From there, AutoML begins it’s behind-the-scenes analysis and learning work with the Neural Network Technology. This is where you’re leveraging Google’s proprietary systems.

Upon completion of the train phase, you’ll receive complete statistics and data results.

From there, you’re ready to have the model predict new results via the AutoML interface or through the REST API integrating with your own application. This allows you to deploy the application and serve results whether on the AutoML website, your own web application, or on an enterprise-grade solution.

Limitations of AutoML

Google’s AutoML is in an Alpha “invite only” status, but you can request AutoML access here.

Additionally, the machine learning models are currently limited to “vision based” processing and recognition. This means your models will focus on image recognition, categorization, and prediction. But you can’t drop in other types of data sets.

At BETSOL, for example, we have a Machine Learning application used to better predict IT project time and budget. Based on past and current data sets, our model improves and delivers highly-accurate estimates for IT projects. Unfortunately with the current state of AutoML, we cannot dump a spreadsheet full of project data in.

Learn more about our machine learning application for IT project estimation here or download the whitepaper here.

Worry not though, more learning models will be released shortly.

Last, your data will be processed on the Google Cloud and not in your own environment, which has its own obvious sets of pros and cons.

Exciting and accessible future of Machine Learning

AutoML is an exciting glimpse of advances in machine learning yet to come. Even in its alpha state, enterprises are flocking to AutoML. It eliminates specialized roles, the grunt work of development and processing, and makes an evolving technology readily available to us all.

Have you used it? Let us know your thoughts below.

[embedit snippet=”after-article-getresponse”] 


Explore other topics of interest