Travis Desell

Professor

Department of Software Engineering

Golisano College of Computing and Information Sciences

Graduate Program Director, Data Science

tjdvse@rit.edu

585-475-2991

Office Location

GOL-1559

Travis Desell

Professor

Department of Software Engineering

Golisano College of Computing and Information Sciences

Graduate Program Director, Data Science

Bio

I am an Associate Professor specializing in Data Science, housed in the Department of Software Engineering in the B. Thomas Golisano College of Computing and Information Sciences (GCCIS). My research focuses on the application of machine learning to large-scale, real world data sets using high performance and distributed computing, with an emphasis on developing systems for practical scientific use. I'm interested in the intersection of evolutionary algorithms and neural networks, or 'neuro-evolution', where evolutionary algorithms are used to automate and optimize the design of neural network architectures. I am actively developing the Evolutionary eXploration of Augmenting Convolutional Toplogies (EXACT) and Evolutionary eXploration of Augmenting Memory Models (EXAMM, formerly known as EXALT) algorithms, which are hosted on GitHub.

I am also active in the area of volunteer computing and citizen science, where I did the initial development of MilkyWay@Home, and more recently the Citizen Science Grid and NSF funded Wildlife@Home which has volunteer citizen scientists annotate hundreds of thousands of hours of video and millions of images to help in the development of computer vision algorithms. Recent work on Wildlife@Home has focused on the development of convolutional neural networks to detect various wildlife species in imagery collected from unmanned aerial systems.

My currently funded research projects include the National General Aviation Flight Information Database (NGAFID), used by general aviation institutions across the country to monitor and predict potential flight safety issues. We are actively developing an interface and methods to detect potential flight issues, trends and mine this massive database of over 800,000 hours of flight data. I am also working on Department of Energy Award #FE0031547, Improving Coal Fired Plant Performance through Integrated Predictive and Condition-Based Monitoring Tools, where we are developing neuro-evolution algorithms to evolve recurrent neural networks to predict coal fired power plant data.

I have also been a main contributor in the development of both the compiler and runtime of SALSA and SALSA Lite, a programming language based on the actor model of computation. SALSA enables easy development of concurrent and transparently distributed applications by following actor semantics.

tjdvse@rit.edu

585-475-2991

Personal Links

Personal Website

Areas of Expertise

Artificial Intelligence

Data Science

Select Scholarship

Published Conference Proceedings

ElSaid, AbdElRahman, Alexander Ororbia, and Travis Desell. "Ant-based Neural Topology Search (ANTS) for Optimizing Recurrent Networks." Proceedings of the 23nd International Conference on the Applications of Evolutionary Computation (EvoStar: EvoApps 2020). Ed. Unknown. N/A, N/A: n.p., Print.

ElSaid, AbdElRahman, et al. "Neuro-Evolutionary Transfer Learning through Structural Adaptation." Proceedings of the 23nd International Conference on the Applications of Evolutionary Computation (EvoStar: EvoApps 2020). Ed. Unknown. N/A, N/A: n.p., Print.

Desell, Travis, AbdElRahman ElSaid, and Alexander Ororbia. "An Empirical Exploration of Deep Recurrent Connections Using Neuro-Evolution." Proceedings of the 23nd International Conference on the Applications of Evolutionary Computation (EvoStar: EvoApps 2020. Ed. Unknown. n.p., Print.

Ororbia, Alexander, AbdElRahman ElSaid, and Travis Desell. "Investigating Recurrent Neural Network Memory Structures using Neuro-Evolution." Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2019). Ed. Unknown. n.p., Print.

ElSaid, AbdElRahman, et al. "Evolving Recurrent Neural Networks for Time Series Data Prediction of Coal Plant Parameters." Proceedings of the 22nd International Conference on the Applications of Evolutionary Computation (EvoStar: EvoApps 2019). Ed. Unknown. n.p., Print.

Journal Paper

Bowley, Connor, et al. "An Analysis of Altitude, Citizen Science and a Convolutional Neural Network Feedback Loop on Object Detection in Unmanned Aerial Systems." Journal of Computational Science 34. 19 (2019): 102-116. Print.

Currently Teaching

DSCI-601

Applied Data Science I

3 Credits

This is the first of a two course applied data science seminar series. Students will be introduced to the data science masters program along with potential projects which they will develop over the course of this series in con-junction with the applied data science directed studies. Students will select a project along with an advisor and sponsor, develop a written proposal for their work, and investigate and write a related work survey to refine this proposal with their findings. Students will begin preliminary design and implementation of their project. Work will be presented in class for peer review with an emphasis on developing data science communication skills. This course will keep students up to date with the broad range of data science applications.

DSCI-602

Applied Data Science II

3 Credits

This is the second of a three course applied data science seminar series. Students will design an implementation plan and preliminary documentation for their selected applied data science project, along with an in class presentation of this work. At the end of the semester students will present preliminary demos of their project and write a preliminary project report. Writing and presentations will be peer reviewed to further enhance data science communication skills. This course will keep students up to date with the broad range of data science applications.

DSCI-640

Neural Networks

3 Credits

This course will cover modern and deep neural networks with a focus on how they can be correctly implemented and applied to a wide range of data types. It will cover the backpropagation algorithm and how it is used and extended for deep feedforward, recurrent and convolutional neural networks. An emphasis will be placed on the implementation, design, testing and training of neural networks. The course will also include an introduction to using a modern neural network framework.

DSCI-644

Software Engineering for Data Science

3 Credits

This course focuses on the software engineering challenges of building scalable and highly available big data software systems. Software design and development methodologies and available technologies addressing the major software aspects of a big data system including software architectures, application design patterns, different types of data models and data management, and deployment architectures will be covered in this course.

DSCI-650

High Performance Data Science

3 Credits

This course will cover concurrent, parallel and distributed programming paradigms and methodologies with a focus on implementing them for use in applied data science or scientific computing tasks. In particular, the course will focus on developing software using graphical processing units (GPUs) and the message passing interface (MPI); with an emphasis on properly handling large-scale, real-world data as part of these applications. The course will also teach scalability and load balancing techniques for developing efficient distributed systems. Programming assignments are required.

DSCI-700

Part-Time Graduate Co-op Experience

0 Credits

The main goal of this course is to provide a mechanism for graduate students to continue participating in co-op education after already completing a full-time co-op at the same company. This will consist of part-time paid employment in the discipline of data science. Co-op education enriches the graduate experience for many students, especially those who are transitioning to data science form another discipline or another domain. Part time co-op hours contribute to curricular practical training (CPT). Completion of all bridge courses and 17 semester hours of graduate courses are required for enrollment.

DSCI-770

Data Science Masters Thesis

3 Credits

This course provides the student with an opportunity to develop a thesis project, and analyze and document the project in thesis document form. An in-depth study of a data science topic will be research focused, having built upon the thesis proposal developed prior to this course. The student is advised by their primary faculty advisor and committee. The thesis and thesis defense is presented for approval by the thesis advisor and committee.

DSCI-771

Continuation of Data Science Masters Thesis

0 - 1 Credits

This course provides the student with an opportunity to complete their thesis project after having enrolled in the data science thesis course (DSCI-770), if extra time if needed. The student continues to work closely with his/her advisor and thesis committee.

DSCI-781

Continuation of Capstone

0 - 1 Credits

This course provides the student with an opportunity to complete their capstone project, if extra time is needed after enrollment in the on campus capstone courses DSCI-601 and DSCI-602 (Applied Data Science I and II) or the online capstone course DSCI-799 (Graduate Capstone). The student continues to work closely with his/her advisor to complete their project.

DSCI-789

Advanced Topics In Data Science

1 - 3 Credits

This course will cover advanced specialized topics data science. Such topics are may be emerging and advanced. Specific prerequisites will be noted for each specific special topic.

DSCI-790

Independent Study

1 - 3 Credits

This course provides the graduate student an opportunity to explore an aspect of data science independently and in depth, under the direction of an advisor. The student selects a topic and then works with a faculty member to describe the value of the work and the deliverables.

SWEN-790

Thesis

6 Credits

This course provides the student with an opportunity to execute a thesis project, analyze and document the project in thesis document form. An in-depth study of a software engineering topic will be research focused, having built upon the thesis proposal developed prior to this course. The student is advised by their primary faculty adviser and committee. The thesis and thesis defense is presented for approval by the thesis adviser and committee.