Product Overview: H2O Driverless AI
H2O Driverless AI is an advanced artificial intelligence (AI) platform designed to automate and streamline the most complex data science and machine learning workflows. This award-winning automatic machine learning (AutoML) platform is engineered to help both expert and novice data scientists develop highly accurate models quickly and efficiently.
What H2O Driverless AI Does
H2O Driverless AI automates a wide range of time-consuming data science tasks, including:
- Advanced Feature Engineering: Automatically generates new, high-value features and feature combinations that would be challenging to identify manually. This process is driven by a unique evolutionary competition that optimizes feature selection for maximum predictive accuracy.
- Model Selection and Tuning: Automatically selects the best algorithms and tunes hyperparameters to achieve the highest predictive accuracy. This includes model stacking and ensembling to enhance model performance.
- Model Validation: Prevents overfitting and data leakage through robust validation techniques, ensuring models are reliable and generalizable.
- Model Deployment: Creates easy-to-deploy, low-latency scoring pipelines for rapid model deployment. This includes standalone batch scoring pipelines and client/server scoring via HTTP or TCP protocols.
Key Features and Functionality
Automation and Efficiency
- Automates data cleaning, feature selection, feature engineering, model selection, and model tuning, allowing data scientists to work on projects much faster, often in minutes or hours instead of months.
Machine Learning and Deep Learning
- Supports a variety of machine learning tasks, including regression, binary classification, and multinomial classification. It also handles time-series problems with advanced time-causal feature engineering and validation schemes.
Interpretability and Transparency
- Provides industry-leading capabilities for understanding, debugging, and sharing model results. This includes Machine Learning Interpretability (MLI), fairness dashboards, automated model documentation, and reason codes for each prediction. These features are crucial for building trust with business stakeholders and regulators.
Data Handling and Visualization
- Supports a wide range of data types, including tabular structured data, numeric, categorical, and textual fields, images, and time-series data. It also allows for custom recipes to handle video, audio, and graph data.
- Automatically generates visualizations and data plots that are most relevant from a statistical perspective, helping users quickly understand their data before starting the model building process.
Natural Language Processing (NLP)
- Converts text strings into features using advanced techniques like TFIDF, CNN, GRU, and state-of-the-art PyTorch BERT transformers. This enables the platform to solve business problems such as sentiment analysis, document classification, and content tagging.
Technical Capabilities
- GPU and Multi-CPU Support: Optimized to take advantage of graphical processing units (GPUs), including multi-GPU workstations and servers like the NVIDIA DGX-1, for order-of-magnitude faster training.
- Client/Server Application: Offers a user-friendly GUI and APIs in Python and R for rapid experimentation and deployment of state-of-the-art supervised machine learning models.
- Multi-User Support and Backward Compatibility: Supports multiple users and ensures backward compatibility, making it a versatile and reliable tool for various use cases.
Conclusion
H2O Driverless AI is a powerful and comprehensive platform that empowers data scientists of all backgrounds to derive maximum value from their data. By automating complex machine learning tasks, providing robust interpretability features, and supporting a wide range of data types and sources, H2O Driverless AI accelerates AI strategies and enables organizations to adopt AI at scale efficiently.