Название: Advances in Data Science
Автор: Ilke Demir, Yifei Lou, Xu Wang
Издательство: Springer
Год: 2021
Страниц: 374
Язык: английский
Формат: pdf (true)
Размер: 11.2 MB

This volume highlights recent advances in Data Science, including image processing and enhancement on large data, shape analysis and geometry processing in 2D/3D, exploration and understanding of neural networks, and extensions to atypical data types such as social and biological signals.

Edge enhancement is an essential problem in image processing, computer graphics, computer vision. It aims to make image edges salient so that users can better visualize images and videos. In this study, we mainly focus on edge sharpening, a subset of edge enhancement that can be applied to many important problems, including image super-resolution (SR), deblurring, etc. Recently, many methods for edge sharpening have been proposed. R.C. Gonzalez and R.E. Woods develops adaptive image filters to operate pixel intensities in the spatial domain or pixel frequency in the frequency domain. The unsharp filter in MATLAB boosts high-contrast regions for edge sharpening, but its success depends on the accuracy of users’ parameter settings. Adobe Photoshop is a prevalent commercial software for image processing that also provides an edge sharpening filter for users and gets competitive edge sharpening results. Xie et al. propose a gradient-domain edge sharpening filter that effectively enhances the sharpness of an image. Their approach includes three gradient-domain operations, i.e., sharpness saliency representation, affinity-based gradient transformation and gradient-domain reconstruction, and can be applied to noisy images.

Many scenarios of interest in data mining are naturally represented as networks (or graphs). Common examples are social networks, biological networks, and web graphs. Interesting problems have arisen with the proliferation of network data, such as community detection and link prediction. Cluster discovery in graphs has many uses such as finding communities and topics in social networks, and protein complexes and functional modules in protein interaction networks. Role discovery has also been applied to many networks. It first arose in sociology, to analyze people’s roles in society (e.g. mother, son, friend, student, etc.). More recently, it has been applied in online social networks, biological and technological networks. Roles are typically assigned to nodes based on their structural and connectivity patterns (e.g. bridge nodes). Roles can be used for descriptive and predictive modeling of dynamic networks. They can also be used to visualize and capture the relevant differences and important patterns in big graph data.

Contents:
Part I. Image Processing
Two-stage Geometric Information Guided Image Reconstruction
Image Edge Sharpening via Heaviside Substitution and Structure Recovery
Two-Step Blind Deconvolution of UPC-A Barcode Images
Part II. Shape and Geometry
An Anisotropic Local Method for Boundary Detection in Images
Towards Learning Geometric Shape Parts
Machine Learning in LiDAR 3D Point Clouds
Part III. Machine Learning
Fitting Small Piece-Wise Linear Neural Network Models to Interpolate Data Sets
On Large-Scale Dynamic Topic Modeling with Nonnegative CP Tensor Decomposition
A Simple Recovery Framework for Signals with Time-Varying Sparse Support
Part IV. Data Analysis
Role Detection and Prediction in Dynamic Political Networks
Classifying Sleep States Using Persistent Homology and Markov Chains: A Pilot Study
A Survey of Statistical Learning Techniques as Applied to Inexpensive Pediatric Obstructive Sleep Apnea Data
Nonparametric Estimation of Blood Alcohol Concentration from Transdermal Alcohol Measurements Using Alcohol Biosensor Devices