17644077. FEDERATED LEARNING OF MACHINE LEARNING MODEL FEATURES simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)
FEDERATED LEARNING OF MACHINE LEARNING MODEL FEATURES
Organization Name
INTERNATIONAL BUSINESS MACHINES CORPORATION
Inventor(s)
FEDERATED LEARNING OF MACHINE LEARNING MODEL FEATURES - A simplified explanation of the abstract
This abstract first appeared for US patent application 17644077 titled 'FEDERATED LEARNING OF MACHINE LEARNING MODEL FEATURES
Simplified Explanation
The abstract of this patent application describes a method for improving the robustness of machine learning models using certification for federated learning. The method involves receiving machine learning model updates, a dataset, and hyperparameters. It then generates certification parameters and filtered machine learning model updates by certifying each data point using abstract representations and filtering the model updates.
- The method aims to enhance the adversarial robustness of machine learning models.
- It utilizes certification for federated learning, which allows for collaborative model training across multiple devices or servers.
- The method involves receiving updates to the machine learning model, a dataset, and hyperparameters.
- It generates certification parameters and filtered model updates by certifying each data point using abstract representations.
- The abstract representations help in identifying and filtering out potentially adversarial or unreliable model updates.
- The method improves the overall reliability and accuracy of the machine learning model by ensuring only trustworthy updates are incorporated.
Potential Applications
This technology has potential applications in various domains where machine learning models are used, including:
- Cybersecurity: Enhancing the robustness of intrusion detection systems and malware classifiers.
- Finance: Improving fraud detection algorithms and risk assessment models.
- Healthcare: Enhancing the accuracy and reliability of disease diagnosis and prediction models.
- Autonomous vehicles: Improving the safety and reliability of AI systems used in self-driving cars.
- Natural language processing: Enhancing sentiment analysis and text classification models.
Problems Solved
This technology addresses the following problems:
- Adversarial attacks: It helps protect machine learning models from malicious attempts to manipulate or deceive them.
- Unreliable updates: It filters out potentially harmful or unreliable model updates, ensuring only trustworthy updates are incorporated.
- Robustness: It improves the overall robustness of machine learning models, making them more resistant to adversarial inputs and noise.
Benefits
The use of certification for federated learning and abstract representations provides several benefits:
- Enhanced adversarial robustness: The method improves the ability of machine learning models to withstand adversarial attacks.
- Improved reliability: By filtering out unreliable updates, the method ensures the model remains accurate and trustworthy.
- Collaborative training: The use of federated learning allows for collaborative model training across multiple devices or servers.
- Efficient model updates: The filtering process helps optimize the incorporation of model updates, reducing computational overhead.
Original Abstract Submitted
Embodiments for providing enhanced adversarial robustness of machine learning models using certification for federated learning in a computing environment by a processor. Machine learning model updates, a dataset, and a set of hyperparameters may be received. One or more certification parameters and one or more filtered machine learning model updates for a machine learning model may be generated by certifying each of plurality of data points using one or more abstract representations in a machine learning operation and filtering the plurality of machine learning model updates.
