Home

Project Title:	MLWPC: Machine Learning for Wireless Propagation Channels
Funding Agency:	National Science Foundation
Award Number:	2008443

Project summary:

The goal of this project is the development of new approaches to wireless channel prediction by means of machine learning. This prediction can be in time, space, or frequency, and serves to enhance the efficiency and reliability of wireless communications. The special structure and the physics of wireless propagation need to be taken into account in order to optimize performance of such channel predictions; consequently "standard" machine learning methods known, e.g., from image classification and processing, cannot be directly applied.

Particular topics of investigation are:

1. Development of novel data augmentation strategies
2. New methods for transfer learning from one environment to another
3. Description of the wireless channel as a non-Euclidean graph, and the development of geometric deep learning methods based on these graphs

The goal is not only to investigate the quality of the predictions, but also the impact of those predictions on system operation.