Bio
Jordan F. Masakuna is a postdoctoral fellow in the department of computer science, University of Sherbrooke. He holds undergraduate and honors degrees in computer science from the University of Kinshasa, DR of Congo in 2010 and 2012 respectively, and holds masters and doctorate degrees in mathematical sciences and computer science from Stellenbosch University, South Africa, in 2015 and 2020 respectively. His research interests lie in artificial intelligence, cybersecurity, network analysis, distributed robotics systems and data fusion.
Research project
Many deep learning methods achieve high performances (in terms of accuracy, precision or any similar measures) and have demonstrated great tools for decision making support in various areas of applications including medicine, industry and telecommunications. However, many solutions do not indicate how trustworthy each prediction is. This means, with many current deep learning based solutions, it is not clearly indicated when to consider the underlying intelligent system producing predictions as fully autonomous or when to request human supervision on the interpretation of a prediction. The reliability problem on deep learning models remains fundamental and there is still room for improvement. Jordan F. Masakuna is interested in providing quantitative indications to assess the reliability of a deep learning model prediction under the context of cybersecurity/anomaly detection. He is interested in quantifying ignorance/uncertainty in deep learning models predictions (auto-encoders in particular) using Bayesian theory and/or Dempster-Shafer theory.