Assessing Quality Attributes of Deep Learning Applications

A diagram depicting the quality attributes of deep learning applications

Left to right are three different implementations of RNN (recurrent neural network): LSTM, GRU, and IndRNN. Which of these implementations are more robust in handling missing data? How do we know? Join the Software Engineering Research Lab to engage in this cutting-edge research project

Deep learning has revolutionized many applications, one of which concerns with combatting combined sewer overflows (CSOs). CSOs represent significant risks to human health as untreated water is discharged to the environment. However, for municipalities that have collected large amounts of time-series data, assessing quality attributes (e.g., robustness) of deep learning solutions (e.g., RNN implementations of LSTM, GRU, and IndRNN) is challenging. In this research project, the Protege student will join a team of graduate students in the Software Engineering Research Lab to develop novel and scalable ways to automatically test deep learning solution’s quality aspects in the context of CSOs.

Headshot of Nan Niu

Nan Niu

Assoc Professor, CEAS - Computing Sciences & Informatics

832 Rhodes Hall

513-556-0051

My current research interests focus on the information seeking strategies that developers use in software engineering. I take an ecological-evolutionary, foraging-theoretic approach to understanding and improving developers' search for relevant information in their daily activities, such as debugging, refactoring, and reuse. My research group investigates how the task environment and the information environment re-shape developers' behaviors, or more accurately, how the developers' behaviors and their environments co-evolve, each shaping the other in important ways. My CAREER project (see http://ceas.uc.edu/news-1415/niu-receives-nsf-career-award.html) links software developers' rational behaviors together with their social information foraging, learning, and co-creation.