BRINGING AI TO ENTERPRISE
Driverless AI: Fast, Accurate, Interpretable AI
Driverless AI is an expert system designed to mimic Kaggle Grandmasters. Just like a chess Grandmaster, there are
only a few of these in the world. Driverless AI brings you the intelligence of a Kaggle Grandmaster in a box.
It is designed to take a raw dataset and automatically visualize
the most interesting patterns for data exploration. It then
applies automatic feature engineering to increase accuracy
by using Kaggle Grandmaster recipes for solving a wide
variety of use-cases. Next, it auto-tunes model parameters
and provides the user with the model that yields the best
results. Lastly, it gives plain English explanations of model
results. Driverless AI enables users of all backgrounds to draw
the most value from their data.
Automatic Visualization
• AutoViz allows users to visualize large datasets in the form
of various graphs and charts without having to write code
• Takes huge datasets and displays outliers and trends in an
interpretable way
• Uses statistics to automatically decide which visualizations
to present to the user
• Exploratory tool that presents an overview of the
distribution of data
Automatic Feature Engineering
• AutoDL employs a library of algorithms and feature
transformations to automatically engineer new features for
a given dataset
• A variable importance chart shows the significance of
newly engineered features
• Data scientists can spend less time exploring data and
engineering features; as a result, they can work on multiple
projects simultaneously and increase their efficiency
Model Interpretability
• The Machine Learning Interpretability feature provides
users with clear and concise explanations of model results
• Four dynamic graphs are generated automatically: K-LIME,
Variable Importance, Decision Tree Chart, and Partial
Dependence Plot; each one helps explore the model
output more closely
• Global interpretability techniques allow the user to broadly
understand models at the level of the full dataset
• Local interpretability techniques explain predictions
for individual data points, and produce plain English
explanations with reason codes
Python Pipeline
• A Jupyter notebook with the Python code that contains
the feature engineering and model building that
Driverless AI performed automatically
• Export Python pipelines to operate on new data in
production
• Conveniently usable through a Jupyter notebook
Personas & Benefits
Business Users SPECIFICATIONS
• Users with domain knowledge will understand the new For CPU Installations For GPU Installations
features generated by AutoDL and their impact on
Minimum Hardware Minimum Hardware
predictions
• x86-64 CPU • NVIDIA K80
• Analysts who have used visualization tools in the past
• 64 GB RAM
can easily transition into using the exploratory AutoViz
tool as well as the Machine Learning Interpretability tool Recommended Recommended
to understand the impact of a particular variable on their Hardware Hardware
model • Dual socket Intel Xeon • NVIDIA P100 or higher
x86-64 or IBM Power • NVLink and NCL
Data Analysts 8 CPU’s venabled
• Using basic data science knowledge, users can better
understand features generated by Driverless AI and use Software Software
outcomes to draw valuable insights for their businesses • Docker CE • Ubuntu 16.04
• AutoViz makes data exploration extremely simple so • CUDA 8
users can identify trends in data through visualizations • NVIDIA Docker 1.0.1
Data Scientists
• AutoViz simplifies the data exploration process For Cloud Installations
• Increase productivity of data scientists by enabling them EC2 Instances Azure Instances
to work on multiple projects simultaneously • P2 or G3 • NC or NV
• The visualizations produced by the Machine Learning • Ubuntu 16.04 • Ubuntu 16.04
Interpretability tool streamline the process of
understanding the impact of a model
Licensing
Driverless AI is a commercially licensed product. For more
information, contact [email protected].
About H2O.ai
H2O.ai is focused on bringing AI to businesses through software. Its flagship product is H2O, the leading open source platform that makes it
easy for financial services, insurance and healthcare companies to deploy machine learning and predictive analytics to solve complex problems.
More than 10,000 organizations and 100,000+ data scientists depend on H2O for critical applications like predictive maintenance and operational
intelligence. The company accelerates business transformation for 169 Fortune 500 enterprises, 8 of the world’s 12 largest banks, 7 of the 10
largest auto insurance companies and all 5 major telecommunications providers.
Follow us on Twitter @h2oai. To learn more about H2O customer use cases, please visit http://www.h2o.ai/customers/. Join the Movement.
Tel: +1.650.227.4572 [email protected] http://www.github.com/h2oai @h2oai
