Učinkovit iterativni algoritem učenja razložljivih značilnic za izboljšano klasifikacijo : doktorska disertacija

Vlahek, Dino

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Učinkovit iterativni algoritem učenja razložljivih značilnic za izboljšano klasifikacijo : doktorska disertacija
Authors:	ID Vlahek, Dino (Author) ID Mongus, Domen (Mentor) More about this mentor...
Files:	DOK_Vlahek_Dino_2024.pdf (1,22 MB) MD5: 27C6007CB133CDB724E20B8AE16A173C
Language:	Slovenian
Work type:	Doctoral dissertation
Typology:	2.08 - Doctoral Dissertation
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	V doktorski disertaciji opišemo nov postopek učenja razložljivih značilnic za klasifikacijske namene. Značilnice med vsako iteracijo rekombiniramo na osnovi vnaprej podanih aritmetičnih operacij, ocenimo pa jih glede na njihovo primernosti za klasifikacijo. Slednja temelji na prekrivanju porazdelitve verjetnosti med vrednostmi vzorcev, ki pripadajo različnim razredom. Za nadaljnji razvoj v naslednjo iteracijo izberemo podmnožico najbolj kakovostnih nekoreliranih značilnic z uporabo nove metode, ki temelji na rezu grafa. Pri tem se postopek opira na dva vhoda parametra, ki omogočata nadzor nad številom členov izhodnih značilnic. Prvi opisuje minimalno sprejemljivo kakovost značilnic, ki jih je treba vključiti v izhodni prostor značilnic, medtem ko drugi določa najvišjo dovoljeno stopnjo podobnosti med značilnicama. Rezultati pokažejo, da je metoda nizko občutljiva na oba vhodna parametra. Naučene značilnice pa statistično značilno izboljšajo klasifikacijsko točnost vseh testiranih klasifikatorjev, medtem ko najboljše točnosti dosežemo z uporabo klasifikatorja naključnih gozdov. Z rezultati primerjave pokažemo, da je predlagani postopek v vseh testnih primerih dosegal ali presegal klasifikacijske točnosti trenutnega stanje tehnike. Prav tako pokažemo tudi pravilnost razlage naučenih značilnic dobro preučene množice testnih podatkov.
Keywords:	klasifikacija podatkov, razložljiva umetna inteligenca, učenje značilnic, odkrivanje znanja
Place of publishing:	Maribor
Place of performance:	Maribor
Publisher:	[D. Vlahek]
Year of publishing:	2024
Number of pages:	IX, 87 str.
PID:	20.500.12556/DKUM-86582
UDC:	004.85.021(043.3)
COBISS.SI-ID:	194558723
Publication date in DKUM:	07.05.2024
Views:	649
Downloads:	162
Metadata:
Categories:	KTFMB - FERI
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	18.12.2023

Secondary language

Language:	English
Title:	An efficient iterative approach to explainable feature learning for improved classification
Abstract:	This dissertation considers a new procedure for learning interpretable features for classification purposes. The features are recombined during each iteration based on predefined arithmetic operations and evaluated according to their suitability for classification. The latter is based on overlapping probability distributions between feature's samples belonging to different classes. For further development in the next iteration, we select a subset of the best quality uncorrelated features using a new method based on a graph cut. Here, the procedure is based on two-parameter input, which allows control over the number of terms of the output features. The first describes the minimum acceptable quality of features that must be included in the output feature space, while the second defines the maximum allowed degree of similarity between features. The experiments display the method's low sensitivity to both input parameters. On the other hand, the learned features statistically significantly improve the classification accuracy of all tested classifiers, while the random forest classifier achieves the best accuracies. As confirmed by experiments, the proposed method achieved or exceeded the classification accuracies of the current state of the art in all test cases. We have also demonstrated the correctness of interpreting the learned features of a well-studied test data set.
Keywords:	data classification, explainable artificial intelligence, feature learning, knowledge discovery

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back