Feature Learning and Understanding
Algorithms and Applications
Book
Chapter
Feature learning is the process of using domain knowledge and special techniques to transform raw data into features. Feature learning can build derived features, eliminate irrelevant, redundant, or noisy data...
Chapter
Principal component analysis (PCA) tries to find an orthogonal linear projection that projects the data into a novel coordinate system, in which the greatest variance by some scalar projection of the data lies...
Chapter
Linear discriminant analysis (LDA) is widely studied in statistics, machine learning, and pattern recognition, which can be considered as a generalization of Fisher’s linear discriminant (FLD). LDA is designed...
Chapter
The traditional linear feature extraction methods focus ℓ2, 1on data global structure information or data local structure information. Although these learning methods perform well in some real applications to som...
Chapter
Recently, tensor-based feature learning has attracted great interest in the emergence of multi-linear data. In this chapter, we briefly introduce some Tucker-based feature learning methods. Different from vect...
Chapter
Convolutional neural network (CNNs) are a kind of feedforward neural network with convolutional computation and deep structure. In recent years, the application of CNN is very extensive, such as visual images,...
Chapter
Latent semantic feature extraction (LSFE) is a feature extraction framework to obtain meaningful features from large volumes of data. In this chapter, we give a brief introduction to LSFE and mainly focus on o...
Chapter
Manifold-learning-based algorithms have widely studied in the last two decades and have been considered as powerful tools for feature learning. The theory of differential geometry shows that the intrinsic geom...
Chapter
Kernel-based nonlinear feature learning plays an important role in pattern recognition. Before deep learning, the combination of kernels and classical feature learning methods such as principal component analy...
Chapter
Low-rank representation (LRR), which constructs a robust low rank representation for data processing, has attracted much attention in the past decades. It is assumed that the data points lie on a low-dimension...
Chapter
In recent years, deep learning has garnered tremendous success in a variety of application domains. As a representative of unsupervised learning in deep learning, auto-encoder (AE) is favored by many researche...
Chapter
Recurrent Neural Networks (RNNs) are a class of artificial neural networks for the processing and predicting sequential data, which add recurrent connections feeding the hidden layers of the neural network bac...