Upcoming Student Presentations
CS Seminar Series: Graduate Student Presentations
Date: Thursday, November 27
Time: 11:00 am – 12:30 pm
Room: SHA 247
Number of Presentations: 3
Presentation 1:
Title of Presentation: Federated Learning in Vehicular Edge Computing
Presenter Name: Fateme Mazloomi (PhD Candidate, Computer Science, Ontario Tech University)
Supervisor: Dr. Shahram Heydari (Faculty of Business and Information Technology, Ontario Tech University), and Dr. Khalil El-Khatib (Faculty of Business and Information Technology, Ontario Tech University)
Abstract: Vehicular edge computing (VEC) enables intelligent transportation by processing rich sensor data close to where it is generated. Federated learning (FL) supports this vision by training models across vehicles without collecting raw data centrally, but real-world VEC introduces challenges: client mobility and intermittent connectivity, highly non-IID and noisy data, heterogeneous compute and bandwidth, and vulnerability to unreliable or malicious updates. This work presents scalable FL frameworks tailored to VEC,centered on a multi-server architecture in which roadside units coordinate local training while supporting seamless client handover. I develop (i) a performance-aware client selection mechanism using Thompson sampling to balance exploration and reliable participation; (ii) robust server-side evaluation and aggregation with statistical checks to detect, down-weight, or exclude outlier and potentially poisoned updates; and (iii) an incentive design that accounts for clients’ costs and leverages generative AI–based synthetic augmentation to rebalance classes and improve generalization. Simulation studies on traffic-sign recognition under non-IID and label-noise scenarios demonstrate improved convergence stability and accuracy, as well as resilience to mobility and adversarial conditions, compared with common baselines.
Presentation 2:
Title of Presentation: Transformation-Invariance Properties of Grad-CAM in Convolutional Neural Networks
Presenter Name: Emon Roy (MSc Candidate, Computer Science, Ontario Tech University)
Supervisor: Mehran Ebrahimi (Faculty of Science, Ontario Tech University), and Kourosh Davoudi (Faculty of Science, Ontario Tech University)
Abstract: The widespread adoption of Convolutional Neural Networks (CNNs) in image classification tasks has led to increasing interest in interpreting their decisions. Gradient-weighted Class Activation Mapping (Grad-CAM) is a post-hoc explanation technique that generates visual heatmaps to highlight the class-relevant regions in an input image. However, in real-world scenarios, input images often undergo transformations such as rotation, zoom, and shifts along the horizontal and vertical axis, which may compromise the reliability of Grad-CAM explanations. This research systematically investigates the transformation-invariance properties of Grad-CAM across popular CNN architectures—ResNet152, DenseNet201, and Xception. An experimental pipeline was designed and implemented to apply controlled transformations to input images, generate corresponding Grad-CAM heatmaps, and quantitatively assess their consistency using metrics like the Euclidean (L2) difference and AUC-ROC analysis. Through extensive experiments, we demonstrate that Grad-CAM heatmaps exhibit varying degrees of inconsistency under transformations. To mitigate this variability, we propose and validate heatmap averaging methods, producing robust, transformation-invariant heatmaps. This investigation provides insights into Grad-CAM’s robustness limitations and presents techniques to enhance the reliability of visual explanations in CNN-based image classification tasks.
Presentation 3:
Title of Presentation: Development and Validation of a tangible hybrid MR and VR simulator for Forceps-Assisted vaginal delivery simulator
Presenter Name: Hector Gualdron Hernandez (MSc Candidate, Computer Science, Ontario Tech University)
Supervisor: Dr. Alvaro Quevedo (Faculty of Business and Information Technology, Ontario Tech University), and Dr. David Rojas (Temerty Faculty of Medicine, University of Toronto, Toronto, Canada)
Abstract: Forceps-Assisted Vaginal Delivery (FAVD) is a safe and effective procedure that can help reduce maternal morbidity by lowering cesarean section rates in cases where assisted delivery is feasible. However, its use has declined in recent years, partly due to the lack of standardized training protocols. Simulation-based education using Virtual and Augmented Reality (VR & AR) offers new opportunities for experiential learning and skill development. This work presents the design and development of a hybrid FAVD simulator integrating a custom 3D-printed user interface (UI) compatible with mixed and virtual reality (MR & VR). A user study was conducted with residents and content experts in obstetrics and gynaecology to evaluate usability, cognitive load, presence and educational value. Results showed that the MR simulator achieved higher usability, presence, and educational ratings than a fully virtual version, underscoring the importance of haptic interactions in psychomotor skill development and supporting the simulator’s potential as an effective training tool for FAVD.