Pranav Dar — August 2, 2018
AVbytes

Overview

  • Auto-Keras is an open source package built for automated machine learning
  • It automatically searches for the hyperparameters of deep learning models
  • Plenty of examples are available to help you get started or assist you when you’re stuck

 

Introduction

Google has been championing AutoML (and with good reason) as the go-to tool for all your AI tasks. They have even released real-world scenarios where AutoML has proven to be more effective than traditional programming tools. But there’s a caveat – it’s a paid service (like so many similar services out there). And while these services may produce good results, there’s always a risk that an open source competitor might supersede it.

With Auto-Keras, that’s exactly what might happen. Auto-Keras is an open source library built for doing automated machine learning based on the popular Keras package. Automated machine learning (AutoML) has recently gained popularity because it makes ML techniques and usage available to non-data science folks. It reduces the technical barrier and allows domain experts to utilise the power of machine learning for their business.

The current version of Auto-Keras provides functions to automatically search for the hyperparameters of deep learning models. It is easy to install, understand, use and has a few examples to help you get started. And of course you can play around with the code and manually change the hyperparameters and other model settings (remember, it’s open source!).

To install Auto-Keras on your machine, paste the below command in your Terminal/Command Prompt window:

pip install autokeras

Check out the full code for Auto Keras on GitHub and also make sure you view all the examples here.

 

Our take on this

Try it out! It’s free so you have nothing to lose here. Automated ML will be a big thing in the future so it’s good, as a data scientist, to understand what all the current hype is about. Auto-Keras will help you build models with easy and unerring accuracy.

Note that this is not to condone the other AutoML tools out there. These tools were built by data scientists to help other data scientists so they should not be ignored. But when an open source opportunity like this comes along, it instantly becomes a hit in the community. If you do use Auto-Keras, share your experience with us in the comments section below.

 

Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!

 

About the Author

Pranav Dar

Senior Editor at Analytics Vidhya. Data visualization practitioner who loves reading and delving deeper into the data science and machine learning arts. Always looking for new ways to improve processes using ML and AI.

Our Top Authors

Download Analytics Vidhya App for the Latest blog/Article

5 thoughts on "Auto-Keras – A Must-Use Open Source Python Package for Automated Machine Learning!"

Hassine Saidane
Hassine Saidane says: August 03, 2018 at 2:33 pm
Hello Pranav, It looks like a great tool. I always believed that ML should be automated to make it more usable, and here it finally happened. What are other automated machine earning tools that are available now? Thanks for this useful information. Reply
Tony Holdroyd
Tony Holdroyd says: August 03, 2018 at 5:37 pm
I am currently testing Auto-Keras and will report back when it's finished running. However, testers should be aware that the install upgrades TensorFlow to v1.9 which could possibly break code that depends on an earlier version, or in my case my installation of CUDA in my virtual environment. Reply
Pranav Dar
Pranav Dar says: August 03, 2018 at 5:53 pm
Thanks for letting the community know, Tony. Reply
Sailesh
Sailesh says: August 11, 2018 at 10:19 am
A quick google would give you few more offerings, or you can dig up from AV archives as well. Feedzai AutoML is one such recent venture. Reply
Pranav Dar
Pranav Dar says: August 12, 2018 at 7:57 pm
There are plenty of tools available - Google's AutoML, IBM's Watson, Feedzai AutoML (as mentioned by Sailesh), Microsoft's Azure Studio, H2O.ai's Driverless AI, etc.. They are all paid services though, whereas Auto Keras is open source. Reply

Leave a Reply Your email address will not be published. Required fields are marked *