He joined LTU as an adjunct professor in 2017 and in 2019 became a full-time professor teaching such courses as fundamentals of programming for business, database systems, web design, data mining, and human-computer interaction.
His current research projects involve finding solutions to a workflow task scheduling problem and, additionally, detecting COVID-19 infection from chest x-ray images. While the two projects are seemingly unrelated, he found the answers to both using AI.
For his PhD dissertation, Mousavi Mojab was, and still is, interested in how we can determine a near-optimal solution to the workflow scheduling problem. Involved are the constraints of time and money. Let’s say someone needs to complete a specific task within a specific timeframe, and in a cost-effective manner, but they don’t want to buy a computer. Mousavi Mojab applied evolutionary algorithms to search for the best solution within a reasonable planning time to allocate tasks to appropriate cloud resources for execution.
Specifically, he and his coresearchers introduced an innovative fitness function that combines the time and monetary cost of a workflow to schedule them. The results of this research were presented and published at the 2019 IEEE Fifth International Conference on Big Data Computing Service and Applications (BigDataService). Right now, Mousavi Mojab is working on the optimization of workflow scheduling in a cloud environment using his recently proposed algorithm Epistocracy. The algorithm is a novel evolutionary method for solving complex optimization problems.
Mousavi Mojab has also applied his Epistocracy algorithm to build and optimize an AI model called “EpistoNet” to detect Covid-19 infection from chest X-ray images. He explained the importance of his research this way: “The diagnosis of COVID-19 from radiograph images is a massive challenge that requires high expertise and dedicated knowledge. We do have a computer that can take an x-ray of the lungs, and with 95 percent accuracy, identify the presence of COVID-19 infection.”
“EpistoNet” has been trained on 2,500 images from Henry Ford Hospital. “We ‘trained’ a predictive model that was composed of sub-models that have been effectively optimized for maximum performance,” he said, adding: “In all evolutionary algorithms, we’re generating different random solutions to a problem, then using an objective function, we sort the solutions based on their performance. We then take the optimal solution and apply it.”
He’s in the process of testing whether Epistocracy performs better than other algorithms. This research was peer-reviewed and published in volume 11 of the prestigious Scientific Reports, one of the respected Springer Nature publications. “Like any other scientific discipline, computer science starts with a hypothesis and seeks to prove it. I’m excited with our results so far,” Mousavi Mojab said.