Chang Wan Ryu

Title: Chang Wan Ryu: Inventor in Machine Learning

Introduction

Chang Wan Ryu is an inventor based in Mountain View, California, and is associated with Google Inc. His patent work addresses machine learning, reinforcement learning, and the use of information retrieval to provide feedback for training an agent model.

Latest Patents

Ryu is named on 1 patent: “Reinforcement Learning with Information Retrieval Feedback.” The invention describes a computer-implemented method for generating feedback signals to train a machine-learned agent model. A machine-learned reward model evaluates an agent’s next-state feature in the context of preceding states, using reference data to help determine the feature’s quality. The resulting quality indicator can then be provided to a model trainer to update the agent model.

Career Highlights

Ryu’s listed patent reflects work on improving how machine-learned agents are evaluated and trained. By incorporating information retrieved from reference data, the disclosed approach aims to provide feedback about an agent’s generated output in relation to its prior states.

Collaborations

Ryu is associated with Google Inc. and is listed as a coworker of Dongseong Hwang. The patent’s described method brings together an agent model, a reward model, reference data, and a model trainer.

Conclusion

Chang Wan Ryu’s patent highlights an approach to reinforcement learning that uses information retrieval to assess generated features and provide training feedback.