17823148. FEDERATED AUTOMATIC MACHINE LEARNING simplified abstract (International Business Machines Corporation)
Contents
FEDERATED AUTOMATIC MACHINE LEARNING
Organization Name
International Business Machines Corporation
Inventor(s)
Lukasz G. Cmielowski of Krakow (PL)
Daniel Jakub Ryszka of Krakow (PL)
Oronde Jason Tucker of Ajax (CA)
Maksymilian Erazmus of Kraków (PL)
FEDERATED AUTOMATIC MACHINE LEARNING - A simplified explanation of the abstract
This abstract first appeared for US patent application 17823148 titled 'FEDERATED AUTOMATIC MACHINE LEARNING
Simplified Explanation
- Systems and methods for federated automatic machine learning - Define a search process for building an automatic machine learning pipeline definition - Distribute the search process across multiple parties - Each party retains federated data including training and holdout data - Receive evaluation results from each party on the model configuration against holdout data - Aggregate evaluation results to define aggregated parameters - Generate a new pipeline definition from the aggregated parameters - Aggregate trained local models from each party to define an aggregated model - Trained local models include the new pipeline definition
---
- Potential Applications
- Collaborative machine learning projects involving multiple parties with data privacy concerns - Federated learning applications in industries such as healthcare, finance, and telecommunications
- Problems Solved
- Protecting data privacy by keeping training data local to each party - Efficiently aggregating model configurations and parameters from multiple parties - Enabling collaborative machine learning without sharing sensitive data
- Benefits
- Enhanced data privacy and security - Improved model performance through collaboration - Scalability for large-scale machine learning projects involving multiple parties
Original Abstract Submitted
Aspects of the invention include systems and methods configured for federated automatic machine learning. A non-limiting example computer-implemented method includes defining a search process including a model configuration for building an automatic machine learning pipeline definition and distributing the search process across a plurality of parties. Each member of the plurality of parties retains federated data including training data and holdout data. The method includes receiving, from each member of the plurality of parties, an evaluation result of the model configuration against respective holdout data and aggregating the received evaluation results to define aggregated parameters. A new pipeline definition is generated from the aggregated parameters and trained local models received from each member of the plurality of parties are aggregated to define an aggregated model. Each trained local model includes the new pipeline definition.
