College of Computing and Digital Media Dissertations

Date of Award

Spring 5-21-2021

Degree Type


Degree Name

Master of Science (MS)


School of Computing

First Advisor

Jacob Furst, PhD

Second Advisor

Daniela Stan Raicu, PhD

Third Advisor

Ilyas Ustun, PhD


The efficiency and generalizability of a deep learning model is based on the amount and diversity of training data. Although huge amounts of data are being collected, these data are not stored in centralized servers for further data processing. It is often infeasible to collect and share data in centralized servers due to various medical data regulations. This need for diversely distributed data and infeasible storage solutions calls for Federated Learning (FL). FL is a clever way of utilizing privately stored data in model building without the need for data sharing. The idea is to train several different models locally with same architecture, share the model weights between the collaborators, aggregate the model weights and use the resulting global weights in furthering model building. FL is an iterative algorithm which repeats the above steps over defined number of rounds. By doing so, we negate the need for centralized data sharing and avoid several regulations tied to it. In this work, federated learning is applied to gaze recognition, a task to identify where the doctor’s gaze at. A global model is built by repeatedly aggregating local models built from 8 local institutional data using the FL algorithm for 4 federated rounds. The results show increase in the performance of the global model over federated rounds. The study also shows that the global model can be trained one more time locally at the end of FL on each institutional level to fine-tune the model to local data.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.