izpis_h1_title_alt

Izboljšave metode globokih naključnih gozdov
ID Klemen, Matej (Author), ID Robnik Šikonja, Marko (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (479,65 KB)
MD5: 93A1CDF8ED337EE9523B2750BF2D1457

Abstract
Na področju globokega učenja je vodilna metoda globokih nevronskih mrež. Te za učenje potrebujejo veliko podatkov, njihova uspešnost pa je odvisna od uporabljenih parametrov. Ena izmed alternativ je model globokih naključnih gozdov. V delu s svojo implementacijo algoritma globokih naključnih gozdov preverimo ponovljivost rezultatov originalnega članka (Zhou in Feng, 2017). Raziščemo, ali lahko napovedno točnost modela izboljšamo z dodajanjem gozdov naključnih podprostorov ali z uporabo zlaganja za kombiniranje napovedi modelov zadnjega nivoja kaskade gozdov. Osnovni implementaciji ter implementacijo z našimi dodatki testiramo na petih podatkovnih množicah in predstavimo dosežene rezultate. Z zlaganjem na vseh podatkovnih množicah dosežemo enako oziroma boljšo povprečno napovedno točnost. Gozdovi naključnih podprostorov na treh podatkovnih množicah dosežejo slabše, na dveh pa boljše rezultate od osnovnega algoritma.

Language:Slovenian
Keywords:globoko učenje, globoki naključni gozdovi, ansambli, zrnato skeniranje, kaskada gozdov, zlaganje modelov
Work type:Bachelor thesis/paper
Organization:FRI - Faculty of Computer and Information Science
Year:2018
PID:20.500.12556/RUL-103415 This link opens in a new window
Publication date in RUL:17.09.2018
Views:1117
Downloads:365
Metadata:XML RDF-CHPDL DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Improving deep random forests
Abstract:
The most frequently used deep learning models are deep neural networks. Although they have been successfully applied to various problems, they require large training sets and careful tuning of parameters. An alternative to deep neural networks is the deep forest model, which we independently implemented to verify the replicability of results in (Zhou and Feng, 2017). We test if the accuracy of deep forest can be improved by including random subspace forests or by using stacking to combine predictions of cascade forest's last layer. We evaluate the original implementation and our improvements on five data sets. The algorithm with added stacking achieves equal or better results on all five data sets, whereas the addition of random subspace forests brings worse results on three data sets and better results on two data sets.

Keywords:deep learning, deep forest, ensemble methods, multi-grained scanning, cascade forest, stacking

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back