Adaptive Tabu Dropout for Regularization of Deep Neural Networks

Md. Tarek Hasan 1, Arifa Akter 1, Mohammad Nazmush Shamael 1, Md Al Emran Hossain 1, H. M. Mutasim Billah 1, Sumayra Islam 1, and Swakkhar Shatabda 2, 1 Department of Computer Science and Engineering, United International University, 2 Department of Computer Science and Engineering, Brac University, · 2025

Read Adaptive Tabu Dropout for Regularization of Deep Neural Networks by Md. Tarek Hasan 1, Arifa Akter 1, Mohammad Nazmush Shamael 1, Md Al Emran Hossain 1, H. M. Mutasim Billah 1, Sumayra Islam 1, and Swakkhar Shatabda 2, 1 Department of Computer Science and Engineering, United International University, 2 Department of Computer Science and Engineering, Brac University, (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-00538.txt.

Description

A dropout trick that spares fresh units is set out by the authors in an arXiv paper whose head is dated 2024. Dropout trains deep nets by muting random units, which stops overfit. Tabu dropout bars units dropped in the last round from being dropped again, which spreads the load. The team adds a tenure, a set count of rounds a unit stays safe after a drop.

An adaptive form picks that tenure on its own from how training runs. Tests on standard data sets put both new forms ahead of plain and basic tabu dropout by clear edges. Spread in which units train seems to drive the gains.

Catalog record

CREATORMd. Tarek Hasan 1, Arifa Akter 1, Mohammad Nazmush Shamael 1, Md Al Emran Hossain 1, H. M. Mutasim Billah 1, Sumayra Islam 1, and Swakkhar Shatabda 2, 1 Department of Computer Science and Engineering, United International University, 2 Department of Computer Science and Engineering, Brac University,
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-00538

Browse the full library · About the Tower of Babel