Potencial uporabe samoojačitvenega učenja za pametni nabiralnik Direct4.me

diplomsko delo

Eva Smolak (Author), Milan Zorman (Mentor), Ambrož Stropnik (Co-mentor)

Abstract

V diplomskem delu, izdelanem pod mentorstvom podjetja Kivi Com d.o.o., smo preučili samoojačitveno učenje, metode samoojačitvenega učenja, globoko samoojačitveno učenje in nevronske mreže. Prav tako smo na kratko opisali priljubljena ogrodja samoojačitvenega učenja in izdelali simulator sistema Direct4.me, kjer smo implementirali postopek dostavljanja in prevzemanja paketov oziroma odpiranja paketnikov. Simulator smo ustvarili v programskem jeziku C#, za izdelavo, učenje in uporabo nevronske mreže pa smo uporabili Python in knjižnico Scikit-learn. Na podlagi simulatorja in nevronske mreže smo preučili potencial uporabe samoojačitvenega učenja v sistemu Direct4.me.

Keywords

samoojačitveno učenje;pametni nabiralnik Direct4.me;umetna inteligenca;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2020
Typology:	2.11 - Undergraduate Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[E. Smolak]
UDC:	004.85(043.2)
COBISS:	39239683
Views:	486
Downloads:	64
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	The potential of reinforcement learning in the Direct4.me smart box application
Secondary abstract:	In this thesis, prepared under the mentorship of Kivi Com d.o.o., we examined reinforcement learning, methods of reinforcement learning, deep reinforcement learning and neural networks. We also shortly described most used reinforcement learning frameworks and created a simulator of the Direct4.me system, where we implemented the procedure of delivering and receiving packages or opening smart boxes. We wrote the simulator in the C# programming language and used Python with Scikit-learn library to build, train, and use the neural network. By combining a simulator and a neural network, we examined the potential of using reinforcement learning in the Direct4.me smart box application.
Secondary keywords:	reinforcement learning;Direct4.me;artificial intelligence;
Type (COBISS):	Bachelor thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Informatika in tehnologije komuniciranja
Pages:	IX, 45 f.
ID:	11929554