JLR Challenge # 3 Technical Workshop (2nd Offering) by: Farzaneh Kazemzadeh

School of Computer Science – JLR Challenge # 3 Technical Workshop

CUDA with CNN: Accelerating Deep Learning through Parallel Computing

(2nd Offering)

Presenter: Farzaneh Kazemzadeh

Date: Wednesday, November 12th, 2025

Time: 1:30 pm

Location: 4th Floor - 300 Ouellette Ave., School of Computer Science, Advanced Computing Hub

Abstract:

Convolutional Neural Networks (CNNs) form the backbone of most computer-vision applications, but their computational intensity makes training and inference challenging on traditional CPUs. This workshop introduces the fundamentals of GPU acceleration through NVIDIA’s CUDA platform, focusing on how parallel computing can optimize CNN operations. The session covers how convolution, pooling, and activation functions are executed in parallel using CUDA kernels and how frameworks such as cuDNN and TensorRT integrate with deep-learning libraries like PyTorch and TensorFlow. By the end of the workshop, participants will understand the principles of mapping CNN operations onto GPU architecture and the advantages of CUDA for accelerating model performance.

Workshop Outline:

- Overview of GPU computing and CUDA architecture
- CPU vs GPU computation for deep-learning workloads
- Thread and block hierarchy in CUDA
- Mapping convolution and pooling operations to CUDA kernels
- Role of cuDNN and TensorRT in CNN acceleration
- Performance-gain comparison between CPU and GPU execution

Prerequisites:

- Basic understanding of neural networks and deep-learning concepts
- Familiarity with Python and frameworks such as PyTorch or TensorFlow

Biography:

Farzaneh Kazemzadeh is a PhD student in Computer Science at the University of Windsor. Her research focuses on trustworthy AI, particularly on privacy-preserving machine learning, with applications in genomics and social networks. Her current work explores memorization and privacy risks in large language models.