diplomsko delo
Abstract
Veliko današnjih informacijskih sistemov uporablja v zalednem sistemu algoritme strojnega učenja, katerih rezultati so lahko težko razumljivi računski modeli, npr. umetne nevronske mreže. Eden izmed primerov uporabnosti teh modelov so sistemi za podporo odločanju, katerih cilj je, da ljudje z njihovo pomočjo pridejo do boljše odločitve. V tem delu raziščemo, kako lahko lokalne naknadne razlage vplivajo na odločanje uporabnikov in njihovo uporabnost v tovrstnih sistemih. V ta namen smo izvedli eksperiment z ljudmi, kjer smo v treh različnih sejah spremljali obnašanja uporabnikov in jih med seboj primerjali.
Keywords
interpretabilnost;LIME;konvolucijske nevronske mreže;računalništvo in informatika;univerzitetni študij;diplomske naloge;
Data
Language: |
Slovenian |
Year of publishing: |
2020 |
Typology: |
2.11 - Undergraduate Thesis |
Organization: |
UL FRI - Faculty of Computer and Information Science |
Publisher: |
[Ž. Jonke] |
UDC: |
004.85(043.2) |
COBISS: |
19155715
|
Views: |
1033 |
Downloads: |
178 |
Average score: |
0 (0 votes) |
Metadata: |
|
Other data
Secondary language: |
English |
Secondary title: |
Decision support system with explanations of artificial neural networks predictions |
Secondary abstract: |
A lot of modern day applications use machine learning algorithms, which may produce computer models, that are hard to understand. These models are applicable in decision support systems, where they help users make better decisions. Here we investigate how can local post-hoc explanations affect user decisions and their overall utility. We have conducted experiments with people where we simulated three environments and monitored user behaviour in each of them, which we later analysed. |
Secondary keywords: |
interpretability;LIME;convolutional neural networks;computer and information science;diploma; |
Type (COBISS): |
Bachelor thesis/paper |
Study programme: |
1000468 |
Embargo end date (OpenAIRE): |
1970-01-01 |
Thesis comment: |
Univ. v Ljubljani, Fak. za računalništvo in informatiko |
Pages: |
49 str. |
ID: |
11811516 |