Mahima Sharma

Title: Mahima Sharma and Adaptive Training for Super-Resolution Models.

Introduction

Mahima Sharma is an inventor associated with Seattle, Washington, and Microsoft Technology Licensing, LLC. Her patent work addresses machine-learning methods for training super-resolution models.

Latest Patents

Sharma is named on 1 patent: “Training Processes for Machine Learning Super-Resolution Models.” The invention describes systems and methods that dynamically adjust a model’s training process. Training data pairs low-resolution images with corresponding high-resolution images and is divided into training and validation groups. After each training epoch, the model is evaluated on the validation group to calculate a loss value. The system tracks how that loss changes between epochs and ends training when the rate of change meets or falls below a predetermined threshold.

Career Highlights

The available information identifies Sharma as an inventor on this patent and connects her work with Microsoft Technology Licensing, LLC. The invention focuses on using validation results to determine when training should stop.

Collaborations

The patent’s listed company association is Microsoft Technology Licensing, LLC. The provided information does not specify additional collaborators.

Conclusion

Mahima Sharma’s patent describes an adaptive approach to training super-resolution models, using changes in validation loss to guide when the training process should conclude.