Konvolucijske nevronske mreže in njihova uporaba
delo diplomskega seminarja
Abstract
Konvolucijske nevronske mreže so vrsta nevronskih mrež, te pa spadajo pod metode strojnega učenja. Strojno učenje je vrsta umetne inteligence, kamor uvrščamo modele in algoritme za napovedovanje in analizo podatkov. Primarni namen konvolucijskih nevronskih mrež je analiziranje vizualnih podob, kot so na primer slike in video podatki. Z njihovo uporabo lahko prepoznavamo poljubne lastnosti in vzorce. Zgradba konvolucijskih nevronskih mrež je podobna splošnim nevronskim mrežam, z razliko od ekaterih specifičnih plasti. Vsako konvolucijsko nevronsko mrežo sestavljajo konvolucijska plast, ReLU plast (ali katera druga aktivacijska plast), združevalna plast, popolno povezana plast in običajno tudi softmax plast. V delu je predstavljena matematična izpeljava delovanja in struktura ob učenju modela. Konvolucijske nevronske mreže so čedalje bolj popularne in se jih uporablja na številnih področjih. Eno od področij, so finančnih trgi. V delu je podan primer uporabe konvolucijskih nevronskih mrež za napovedovanje gibanja finančnega indeksa.
Keywords
matematika;nevronske mreže;konvolucijske nevronske mreže;strojno učenje;konvolucijska plast;ReLU plast;združevalna plast;gradientni spust;
Data
| Language: | Slovenian |
|---|---|
| Year of publishing: | 2022 |
| Typology: | 2.11 - Undergraduate Thesis |
| Organization: | UL FMF - Faculty of Mathematics and Physics |
| Publisher: | [V. Rozman] |
| UDC: | 004.8 |
| COBISS: |
122087171
|
| Views: | 1100 |
| Downloads: | 145 |
| Average score: | 0 (0 votes) |
| Metadata: |
|
Other data
| Secondary language: | English |
|---|---|
| Secondary title: | Convolutional neural networks and their applications |
| Secondary abstract: | Convolutional neural networks are a type of neural networks, which fall under the category of machine learning methods. Machine learning is a type of artificial intelligence that includes models and algorithms for prediction and data analysis. The primary purpose of convolutional neural networks is to analyze visual images such as pictures and videos. With their help, we can recognize arbitrary features and patterns. Their structure is similar to general neural networks, with the difference of some specific layers. Each convolutional neural network consists of a convolutional layer, a ReLU layer (or some other activation layer), a pooling layer, a fully connected layer, and usually a softmax layer. This work presents the mathematical derivation of the operation and structure during model learning. Convolutional neural networks are becoming more and more popular and are being used in many areas. One of them is a financial market. For this purpuse an example of financial time-series data analysis using convolutional neural networks is described. |
| Secondary keywords: | mathematics;neural networks;convolutional neural networks;machine learning;convolutional layer;ReLU layer;pooling layer;gradient descent; |
| Type (COBISS): | Final seminar paper |
| Study programme: | 0 |
| Embargo end date (OpenAIRE): | 1970-01-01 |
| Thesis comment: | Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Finančna matematika - 1. stopnja |
| Pages: | 29 str. |
| ID: | 16479197 |
Recommended works: