18383946. SYSTEMS AND METHODS TO IDENTIFY NEURAL NETWORK BRITTLENESS BASED ON SAMPLE DATA AND SEED GENERATION simplified abstract (Capital One Services, LLC)

SYSTEMS AND METHODS TO IDENTIFY NEURAL NETWORK BRITTLENESS BASED ON SAMPLE DATA AND SEED GENERATION

Organization Name

Capital One Services, LLC

Inventor(s)

Austin Walters of Savoy IL (US)

Vincent Pham of Champaign IL (US)

Galen Rafferty of Mahomet IL (US)

Anh Truong of Champaign IL (US)

Mark Watson of Urbana IL (US)

Jeremy Goodsitt of Champaign IL (US)

SYSTEMS AND METHODS TO IDENTIFY NEURAL NETWORK BRITTLENESS BASED ON SAMPLE DATA AND SEED GENERATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 18383946 titled 'SYSTEMS AND METHODS TO IDENTIFY NEURAL NETWORK BRITTLENESS BASED ON SAMPLE DATA AND SEED GENERATION

Simplified Explanation

The abstract describes systems and methods for determining neural network brittleness. This includes receiving a modeling request with a preliminary model and dataset, determining brittleness scores for the preliminary and reference models, comparing the scores, and generating a preferred model based on the comparison.

• The system includes memory units storing instructions and processors executing the instructions.
• Operations involve receiving a modeling request with a preliminary model and dataset.
• Determining brittleness scores for the preliminary and reference models.
• Comparing the scores and generating a preferred model based on the comparison.
• Providing the preferred model to the user.

Potential Applications

• Improving the robustness and reliability of neural networks.
• Enhancing the performance of machine learning models.
• Identifying and addressing vulnerabilities in AI systems.

Problems Solved

• Addressing brittleness issues in neural networks.
• Improving the overall accuracy and stability of machine learning models.
• Enhancing the trustworthiness of AI systems.

Benefits

• Increased reliability and robustness of neural networks.
• Improved performance and accuracy of machine learning models.
• Enhanced security and trust in AI systems.

Original Abstract Submitted

Systems and methods for determining neural network brittleness are disclosed. For example, the system may include one or more memory units storing instructions and one or more processors configured to execute the instructions to perform operations. The operations may include receiving a modeling request comprising a preliminary model and a dataset. The operations may include determining a preliminary brittleness score of the preliminary model. The operations may include identifying a reference model and determining a reference brittleness score of the reference model. The operations may include comparing the preliminary brittleness score to the reference brittleness score and generating a preferred model based on the comparison. The operations may include providing the preferred model.