ATM Golam Bari recipient of an International Travel Grant to the IEEE World Congress on Computational Intelligence conference
May 29, 2018
Under the supervision of Associate Professor Alessio Gasper, ATM Golam Bari, a doctoral student in the Department of Computer Science and Engineering, is a recipient of an International Travel Grant from the Office of Graduate Studies. Bari will represent USF and travel to the IEEE World Congress on Computational Intelligence conference in Rio de Janeiro, Brazil where he will present his two papers.
Bari's research interests are in Evolutionary Algorithms (EAs) and their application to computer-aided learning. More specifically, he is interested in leveraging EAs to automatically design practice exercises for novice programmers. Simply looking at this problem from the perspective of the Coevolutionary dynamics that arise between two co-adapting populations (learners vs. practice problems) allows supporting algorithmic solutions to both exercises synthesis and characterization of concept inventories specific to both the study material and the student population.
See below to read more about Bari's papers.
To learn more about IEEE Computational Intelligence Society, click here.
Paper 1 title: Selection Methods to Relax Strict Acceptance Condition in test-based Coevolution
Authors: ATM Golam Bari, Alessio Gaspar, R. Paul Wiegand, Anthony Bucci
Summary: The work presented in this paper is motivated within a higher-level goal. This paper investigates how theoretical concepts of underlying test-based coevolutionary computation can be used to improve educational outcomes in situations in which student (candidate) performance is assessed through the application of exercise sets (tests). More specifically, a computer-aided teaching system allowing novice programmers to hone their skills against coevolved practice exercises is developed, deployed and analyzed. To this end, some new selection conditions that include both Pareto dominated and Pareto non-dominated tests, as well as other factors to help provide distinctions are proposed. The potential benefits of also considering Pareto non-dominated solutions are illustrated by a visualization of the underlying interaction space in terms of levels. In addition, some new performance metrics that allow one to compare our various selection methods in terms of ideal evaluation of coevolution are also developed.
Paper 2 title: DynTLBO - A Teaching Learning-based Dynamic Optimization Algorithm
Authors: ATM Golam Bari, Alessio Gaspar
Summary: The Teaching Learning Based Optimization algorithm simulates the knowledge-transfer process between teacher and learners as well as between peer learners. Although this algorithm has been already successfully applied to both constrained and unconstrained engineering optimization problems, it sometimes prematurely converges toward local optima, especially in high dimensional, multimodal, or deceptive fitness landscapes. This work is motivated to maintain diversity in converging populations; thus balancing the trade-off of search space explorations vs exploitation. To do so, we improve the peer selection policy in "learner phase" and introduce different types of immigrant solutions before starting of "teacher phase". Then we measure the performance of the algorithm both in static and especially in dynamic environment. We found that the resulting DynTLBO algorithm showed significant performance improvements on commonly used benchmarks.