Follow Datanami:
July 7, 2017

H2O.ai Boasts New AI Product Like ‘Kaggle Grandmaster in a Box’

Want to get started with data science and artificial intelligence, but lack the skilled personnel to do it? You could be a candidate for machine learning software company H2O.ai’s latest creation, Driverless AI, which it compares to “Kaggle Grandmaster in a box.”

The Mountain View, California company today announced the beginning of beta testing for Driverless AI. The new product combines H2O.ai‘s automated machine learning and deep learning products, AutoML and AutoDL respectively, which provide automatic training and tuning of models on GPU-accelerated hardware.

It’s all about reducing complexity and making the most of what data science skill sets users already have, the company says.

“Although H2O has made it easy for non-experts to experiment with machine learning, there is still a fair bit of knowledge and background in data science that is required to produce high-performing machine learning models,” H2O.ai’s Erin LeDell, Navdeep Gill, and Ray Peck wrote in a blog post last month.

“Deep Neural Networks in particular are notoriously difficult for a non-expert to tune properly,” they continue. “We have designed an easy-to-use interface which automates the process of training a large, diverse, selection of candidate models and training a stacked ensemble on the resulting models (which often leads to an even better model).”

Driverless AI takes AutoML and AutoDL to the next level, and automates many of the other aspects that go into building, running, and maintaining a predictive analytics pipeline. In short, it does quite a bit to take the human data scientist, i.e. the “Kaggle Grandmaster,” out of the loop.

Kaggle Grandmasters, like chess

Where Driverless AI fits into the big data analytics stack (Source: H2O.ai)

grandmasters, are few and far between, and represent the epitome of excellence in the profession of data science. Before it was acquired by Google earlier this year, Kaggle hosted thousands of competitions pitting data scientists against one another in a race to elegantly solve tough machine learning challenges. More than 800,000 people have participated in Kaggle challenges, making it by far the biggest data science competition site.

H2O.ai says Driverless AI “brings you the intelligence of a Kaggle Grandmaster in a box.” The software will incorporate multiple data science capabilities to “take care of the whole data pipeline,” a company spokesman says.

“Users do not have to worry about data preparation, finding the best algorithms to use, calibrating the parameters, creating a test/validation data set, creating and comparing models, and deploying models,” he says. “Everything is done by Driverless AI.”

What’s more, even organizations that have data scientists on staff — or more likely, a team of developers, analysts, engineers, and line-of-business experts working together as a data science team — can use Driverless AI and get a productivity boost, the company says.

“Even organizations with existing data science practices can simplify their workflow,” the spokesman says. “The ‘data-in, models-out’ paradigm reduces many collaboration points, communication gaps and efforts to go through the whole data pipeline. Businesses can devote their resources on business priorities to create more revenue.”

Driverless AI runs on GPUs, which H2O.ai says gives it up to a 40x speed boost compared to running on non-GPU hardware. It also includes the new XGBoost optimized distributed gradient boosting algorithm that H2O.ai delivered with its flagship H2O software late last month.

To sign up for the beta, go to www.h2o.ai/driverless-ai.

Related Items:

Google Mimics Human Brain with Unified Deep Learning Model

Machine Learning, Deep Learning, and AI: What’s the Difference?

Google Buys Data Science Competition Site Kaggle

 

Datanami