Elizabeth Patterson Cohn

Title: Elizabeth Patterson Cohn: Inventor in Machine Learning

Introduction

Elizabeth Patterson Cohn is an inventor based in Raleigh, North Carolina. Her patent work concerns machine learning and reinforcement learning, with a focus on adapting confidence thresholds as models are retrained.

Latest Patents

Cohn is associated with 1 patent, titled “Reinforcement Learning for Machine Learning Models Using Dynamic Confidence Thresholds.” The invention describes generating training datasets by augmenting labeled data with synthetic labeled data, then creating retrained versions of a machine learning model. It evaluates each version by comparing its output with a validation dataset and uses a reward indicator to adjust the model’s confidence threshold.

Career Highlights

Cohn is associated with Optum, Inc. Her patent reflects work on methods for improving machine-learning model training and adapting model behavior based on validation results.

Collaborations

Cohn’s listed coworkers include Ayan Sengupta and Sudhanshu Sharma. The patent’s described approach combines dataset augmentation, model retraining, validation, and threshold adjustment.

Conclusion

Elizabeth Patterson Cohn’s patent highlights an approach to machine learning that uses validation-based rewards to dynamically adjust confidence thresholds.