DMRAE: discriminative manifold regularized auto-encoder for sparse and robust feature learning
- PDF / 630,108 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 52 Downloads / 199 Views
REGULAR PAPER
DMRAE: discriminative manifold regularized auto-encoder for sparse and robust feature learning Nima Farajian1 · Peyman Adibi2 Received: 31 July 2019 / Accepted: 13 June 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Although the regularized over-complete auto-encoders have shown great ability to extract meaningful representation from data and reveal the underlying manifold of them, their unsupervised learning nature prevents the consideration of class distinction in the representations. The present study aimed to learn sparse, robust, and discriminative features through supervised manifold regularized auto-encoders by preserving locality on the manifold directions around each data and enhancing between-class discrimination. The combination of triplet loss manifold regularization with a novel denoising regularizer is injected to the objective function to generate features which are robust against perpendicular perturbation around data manifold and are sensitive enough to variation along the manifold. Also, the sparsity ratio of the obtained representation is adaptive based on the data distribution. The experimental results on 12 real-world classification problems show that the proposed method has better classification performance in comparison with several recently proposed relevant models. Keywords Supervised regularized auto-encoder · Manifold regularization · Robust feature learning · Discriminative sparse representation
1 Introduction In recent decades, various machine learning algorithms have been proposed to analyze different types of data with significant performances. With increasing the effect of computers and digital equipment in different parts of life and industry, the volume, size, and types of data used in these systems like photographs, videos, and text documents have been increased significantly. Features extracted from the data strongly affect the machine learning algorithm performance. Data usually have high dimensions, and in the initial form they have the least information for discriminating and classifying data. Thus, many methods have been proposed to preprocess data and produce and extract important features from raw data which their result is a new representation of data that increases the performance of machine learning algorithms.
B
Peyman Adibi [email protected]
1
Department of Computer Engineering, Faculty of Computer and Electrical Engineering, University of Kashan, Kashan, Iran
2
Artificial Intelligence Department, Faculty of Computer Engineering, University of Isfahan, Isfahan, Iran
The main problem of these methods is the need to have prior knowledge of data. As a result, producing features based on prior knowledge of data should have been done separately and differently for each dataset [1]. Moreover, good features usually are not produced in many problems due to high complexity and lack of enough prior knowledge. These issues motivated researchers to focus on methods that can produce good features by examining data automatically a
Data Loading...