Dropout-based active learning for regression

Evgenii Tsymbalov, Maxim Panov, Alexander Shapeev

Результат исследований: Глава в книге, отчете, сборнике статейМатериалы для конференциирецензирование

14 Цитирования (Scopus)

Аннотация

Active learning is relevant and challenging for high-dimensional regression models when the annotation of the samples is expensive. Yet most of the existing sampling methods cannot be applied to large-scale problems, consuming too much time for data processing. In this paper, we propose a fast active learning algorithm for regression, tailored for neural network models. It is based on uncertainty estimation from stochastic dropout output of the network. Experiments on both synthetic and real-world datasets show comparable or better performance (depending on the accuracy metric) as compared to the baselines. This approach can be generalized to other deep learning architectures. It can be used to systematically improve a machine-learning model as it offers a computationally efficient way of sampling additional data.

Язык оригиналаАнглийский
Название основной публикацииAnalysis of Images, Social Networks and Texts - 7th International Conference, AIST 2018, Revised Selected Papers
РедакторыAlexander Panchenko, Wil M. van der Aalst, Michael Khachay, Panos M. Pardalos, Vladimir Batagelj, Natalia Loukachevitch, Goran Glavaš, Dmitry I. Ignatov, Sergei O. Kuznetsov, Olessia Koltsova, Irina A. Lomazova, Andrey V. Savchenko, Amedeo Napoli, Marcello Pelillo
ИздательSpringer Verlag
Страницы247-258
Число страниц12
ISBN (печатное издание)9783030110260
DOI
СостояниеОпубликовано - 2018
Событие7th International Conference on Analysis of Images, Social Networks and Texts, AIST 2018 - Moscow, Российская Федерация
Продолжительность: 5 июл. 20187 июл. 2018

Серия публикаций

НазваниеLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Том11179 LNCS
ISSN (печатное издание)0302-9743
ISSN (электронное издание)1611-3349

Конференция

Конференция7th International Conference on Analysis of Images, Social Networks and Texts, AIST 2018
Страна/TерриторияРоссийская Федерация
ГородMoscow
Период5/07/187/07/18

Fingerprint

Подробные сведения о темах исследования «Dropout-based active learning for regression». Вместе они формируют уникальный семантический отпечаток (fingerprint).

Цитировать