Neural Networks and Deep Learning: Hardware Acceleration

Overview

 The course is a module of the Neural Networks and Deep Learning course. The objective of this course is to present practical and implementation issues useful to deploy neural networks on a variety of embedded platforms using different languages and development environments. Topics covered in the course include common development frameworks, networks optimization for embedded platforms, acceleration of deep networks on GPGPUs and FPGA platforms.

Program:

  1. Programming frameworks for deep learning
  2. Modeling DNNs in Tensorflow and PyTorch
  3. DNN optimization for embedded platforms
  4. The Nvidia TensorRT inference framework
  5. Accelerating deep networks on FPGA
  6. GPU programming in CUDA
  7. Accelerating deep networks on GPGPUs
  8. The AMD/Xilinx Deep Learning Processing Unit

Format and exam
  1. Lectures (20 hours): Lectures will be given online over Microsoft Teams.
  2. Exam (2 CFU): Project work and oral discussion.

Schedule

 TBD

Course material

 TBD