magistrsko delo
Abstract
Na področju strojnega učenja se pogosto pojavljajo problemi z množicami visokih razsežnosti, ki pa so zaradi “prekletstva razsežnosti” zahtevni za reševanje. Pri reševanju takih problemov pogosto uporabljamo metode za manjšanje razsežnosti množic. Popularna metoda za manjšanje razsežnosti so samokodirniki, ki so ponavadi zgrajeni iz nevronskih mrež. Slabost nevronskih mrež je, da zahtevajo veliko procesorskega časa in da ima uporabnik zaradi njihove kompleksnosti zelo slab vpogled v njihovo delovanje. Zato želimo v magistrskem delu razviti samokodirnik na osnovi naključnega gozda, ki teh slabosti ne bi imel. Za konstrukcijo samokodirnika iz naključnega gozda izberemo nabor listov, ki skupaj čim bolje opišejo podatkovno množico, in jih združimo v kodirni vektor. Smokodirnik nato primer zakodira na osnovi njegove pripadnosti listom konstruiranega kodirnega vektorja. Za postopek dekodiranja imamo na razpolago dve informaciji: poti v odločitvenih drevesih, ki vodijo do listov v kodirnem vektorju in shranjene napovedi naključnega gozda. Da poiščemo čim boljšo rekonstrukcijo zakodiranih primerov, uporabimo oba podatka. Naš samokodirnik testiramo, da določimo čim boljše nastavitve parametrov, in njegovo natančnost primerjamo s samokodirniki iz nevronskih mrež. Ugotovimo, da je zaenkrat manj natančen od standardnega pristopa, in premislimo možnosti, kako ga lahko v prihodnosti izboljšamo.
Keywords
strojno učenje;manjšanje razsežnosti podatkov;samokodirniki;naključni gozdovi;umetne nevronske mreže;
Data
Language: |
Slovenian |
Year of publishing: |
2021 |
Typology: |
2.09 - Master's Thesis |
Organization: |
UL FU - Faculty of Administration |
Publisher: |
[T. Makovecki] |
UDC: |
004.42 |
COBISS: |
69733379
|
Views: |
2086 |
Downloads: |
117 |
Average score: |
0 (0 votes) |
Metadata: |
|
Other data
Secondary language: |
English |
Secondary title: |
Autoencoder via random forest |
Secondary abstract: |
In the field of machine learning, problems with high-dimensional data sets are common, and difficult to solve due to the “curse of dimensionality”. To solve these problems, we usually apply methods for dimensionality reduction. A popular method for this are autoencoders, which are usually built with neural networks. However, the downside of neural networks is high computation costs of training and their complexity which obscures the user insight into how they work. To address these issues, we aim at developing an autoencoder that is based on random forests and does not have such problems. To construct an autoencoder from a random forest, we select a set of forest leaves, which describe the data set well, and save them into an encoding vector. We use the encoding vector to encode data samples. There are two types of information we can use to decode the data: the decision tree paths leading to leaves in the encoding vector and the saved predictions form the random forest. We combine the two to get the best possible reconstruction of encoded data. We test the constructed autoencoder to tune the parameter settings and evaluate its performance in comparison to neural network autoencoders. We establish that at this point our autoencoder is significantly less accurate compared to common autoencoders and consider the possibilities for upgrading it in the future. |
Secondary keywords: |
machine learning;dimensionality reduction;autoencoders;random forests;artificial neural networks; |
Type (COBISS): |
Master's thesis/paper |
Study programme: |
0 |
Embargo end date (OpenAIRE): |
1970-01-01 |
Thesis comment: |
Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Matematika - 2. stopnja |
Pages: |
IX, 40 str. |
ID: |
13123209 |