Propostas para Dissertação

Mestrados no Departamento de Informática

Consultar ficha completa de uma proposta de dissertação

proponente: António Esteves
instituição/empresa: Universidade do Minho - Departamento de Informática
tema/título: How do different architectures support deep learning
área científica: Engenharia de Computadores
local: DI-UM
curso de mestrado: Mestrado em Engenharia Informática
The objective of the proposed thesis is to study the viability of different architectures (CPUs,
GPUs and FPGAs) for deep learning. To keep work feasible it is suggested to base the comparison on the widely used deep Convolutional
neural networks (CNNs). A language that can be used to target all 3 types of architectures is OpenCL. Python is another
challenging alternative, especially for FPGAs. At least, the comparison must include metrics such as performance, portability of code, development
time, code reusability, availability of libraries and other tools. References: [1] What’s a Convolutional Neural Network? [2] Fundamentals of Deep Learning - Starting with Artificial Neural Network. Aarshay Jain, 2016.

s:// [3] Deep Learning for Computer Vision - Introduction to Convolution Neural Networks. Aarshay Jain,

 [4] Convolutional Networks for Computer Vision Applications. Andrea Vedaldi, 2016. [5] Flexible, High Performance Convolutional Neural Networks for Image Classification, D. Cireşan,
U. Meier, et al., 2011. [6] NVIDIA GPUs - The Engine of Deep Learning. [7] Intel Math Kernel Library functions for Deep Neural Networks. [8] An OpenCL Deep Learning Accelerator on Arria 10. U. Aydonat, S. O’Connell, D. Capalija, et
al., 2017. [9] Conversor de Python para linguagens de descrição de hardware.