Elmnet: Feature learning using extreme learning machines

Representation learning with efficient extreme learning machine auto-encoders

“…Extreme Learning Machine Auto-Encoder (ELM-AE) [23,24] [31][32][33][34] were developed for supervised tasks. Meanwhile, it has also shown local receptive fields-based structure [25,26,[35][36][37] is effective for unsupervised feature extraction.…”

Enhanced extreme learning machines for image classification

“…Fourthly, target coding (coding for the labels of the samples) is an indispensable part of image classification. With years of research, there are lots of work that have been done on how to extract useful features using ELMs [42,43,44,45], but little work has been done on target coding for ELM. By far, most people adopt one-of-k target coding methods instantly when they use ELM classifiers, while others choose the numeric coding methods to simplify the computation.…”

“…Chapter 5 uses the material from References [45,55]. We propose a novel feature representation method named CFR-ELM by using ELM-AE for image classification.…”

“…), has been extensively studied in the past decade and attracted increasing attention from academia and industry [38,30,61,62,63]. ELM theories demonstrate that hidden neurons can be set randomly in many applications, including clustering, feature learning, regression, and classification [64,45,65,66]. Many variants of the ELM have been proposed in the last decade, and most of our work concentrate on the early ELM (called "basic ELM" here), explore and enhance its abilities of image classification and feature learning.…”

“…Target coding (or labels encoding) is an indispensable part of the supervised learning algorithms. With years of research, there are lots of work have been done on how to extract effective features using ELMs[42,43,44,45]. However, little work has been done on target coding for ELM.…”