FPGA implementacija analize vremenskih serija metodama dubokog učenja

master's thesis

• Title (english): FPGA implementation of time series analysis using deep learning methods
• Author: Matija Sever
• Mentor: Vlado Sruk (mentor)
• Committee member: Vlado Sruk (predsjednik povjerenstva)
• Committee member: Dejan Škvorc (član povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-19, Croatia

Abstract

Ovaj rad istražuje FPGA implementaciju analize vremenskih serija korištenjem metoda dubokog učenja, konkretno povratnih neuronskih mreža (RNN) i dugih kratkotrajnih memorijskih mreža (LSTM). Modeli su trenirani koristeći TensorFlow 2.5.0 te su konvertirani u TensorFlow Lite (TFLite) format za izvođenje na rubnim uređajima poput FPGA platforme Xilinx ZCU102. Implementacija je ograničena na korištenje ARM Cortex-A53 procesora, a rezultati evaluacije pokazuju poboljšane performanse na FPGA platformi u usporedbi s klasičnim procesorima, s prijedlozima za buduća poboljšanja uporabom sklopovskih akceleratora i kvantizacije.

Abstract (english)

This thesis explores the FPGA implementation of time series analysis using deep learning methods, specifically Recurrent Neural Networks (RNN) and Long ShortTerm Memory (LSTM) models. The models were trained using TensorFlow 2.5.0 and converted into TensorFlow Lite (TFLite) format for execution on edge devices such as the Xilinx ZCU102 FPGA platform. The implementation was limited to the use of the ARM Cortex-A53 processor, and evaluation results show improved performance on the FPGA platform compared to conventional processors, with suggestions for future improvements using
hardware accelerators and quantization

• Keywords: FPGA; duboko učenje; vremenske serije; RNN; LSTM; analiza podataka; Xilinx Zynq-ZCU102
• Keywords (english): FPGA; deep learning; time series; RNN; LSTM; data analysis; Xilinx Zynq-ZCU102
• Language: croatian
• URN:NBN: urn:nbn:hr:168:415521
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2025-02-25 13:38:05