Roadmaps and Vision, Michal Malohlavav - H2O World San Francisco
0 likes209 views
The document outlines the future roadmap and vision for H2O.ai's products, including features and enhancements planned for Driverless AI and H2O-3 between 2018 and mid-2019. Key features mentioned involve data augmentation, various machine learning algorithms, and integrations with cloud platforms like AWS, Azure, and Google Cloud. It also highlights support for advanced capabilities in model reproducibility, interpretability, and lifecycle management.
Roadmaps and Vision, Michal Malohlavav - H2O World San Francisco
- 3. 2018 Q1-2’19 Future H2O-3 Sparkling Water AutoML § Isolation Forest, Cox PH § GLM Ordinal Regression § Monotonic Constraints in GBM § Custom metrics for early stopping § Tree models visualization § XGBoost integration § Target encoding § Checkpointing during AutoML/Grid § Driverless AI MOJO model support § Enterprise Steam integration § Job queues § Spark 2.4 support § Tree SHAP § GLM Negative Binomial § GBM Custom loss functions § Native Hive support § SVM § Factorization Machines § 1-Click Model Reproducibility § Holdout stacking in Stacked Ensembles § AutoML Benchmakrs § Automatic target encoding § Variable importance for Stacked Ensembles § Wide datasets support § AutoML report § Full support of Spark pipeline § K8S support § Pipelines monitoring § Spark 3 support § Limited topology execution § Spark Hydrogen support § H2O ML Algorithms on demand
- 4. AutoViz MLI Driverless AI 2018 Q1-2’19 Future § Multi node, multi user support § Project workspaces § Data augmentation § Model productionalization § Image recipe § Unsupervised learning § Model lifecycle management § DAI and H2O-3 interoperability § Data augmentation marketplace § Explanations and Fairness for DAI & Python models § Model debugging for DAI & Python models § What-if analysis § Model debugging and testing suite § Interpretability for unstructured data § Automated compliant models § Advanced model diagnostics § Interactivity § Actionability
- 5. Data Augmentation Partner Engineering 2018 2019 § Driverless AI, H2O-3 available in AWS, Azure and Google clouds § Driverless AI data connectors for AWS S3, Azure Blob Store Google BigQuery and cloud storage, Minio, Snowflake § Driveress AI, H2O-3 KubeFlow integration § Data augmentation with user or 3rd party data § A smart data exchange § Data augmentation marketplace