Pattern Recognition Applications and Methods: 8th International Conference, ICPRAM 2019, Prague, Czech Republic, February 19-21, 2019, Revised ... Vision, Pattern Recognition, and Graphics)

✍ Scribed by Maria De Marsico (editor), Gabriella Sanniti di Baja (editor), Ana Fred (editor)

• Publisher: Springer
• Year: 2020
• Tongue: English
• Leaves: 170
• Category: Library

✦ Synopsis

This book contains revised and extended versions of selected papers from the 8th International Conference on Pattern Recognition, ICPRAM 2019, held in Prague, Czech Republic, in February 2019. The 25 full papers presented together 52 short papers and 32 poster sessions were carefully reviewed and selected from 138 initial submissions.
Contributions describing applications of Pattern Recognition techniques to real-world problems, interdisciplinary research, experimental and/or theoretical studies yielding new insights that advance Pattern Recognition methods are especially encouraged.

✦ Table of Contents

Preface
Organization
Contents
Theory and Methods
Fourier Spectral Domain Functional Principal Component Analysis of EEG Signals
1 Introduction
2 Methods
2.1 Fourier Power Spectra
2.2 Functional Representation of Signal Power Spectrum
2.3 Functional Probes
2.4 Classical Principal Component Analysis
2.5 Functional Principal Component Analysis
2.6 B-Spline as Functional Basis
2.7 Feature Extraction by Functional Principal Component Analysis
3 Results
4 Concluding Remarks
References
Online Budgeted Stochastic Coordinate Ascent for Large-Scale Kernelized Dual Support Vector Machine Training
1 Introduction
2 Support Vector Machine Training
2.1 Non-linear SVM Solvers
2.2 Budgeted Stochastic Gradient Descent
3 Budgeted Stochastic Coordinate Ascent
4 Analysis of BSCA
4.1 Optimization Progress of BSCA
4.2 Budget Maintenance Rate
5 Adaptive Coordinate Frequency on a Budget
6 Fast Randomized Merging
7 Stopping Criterion
8 Experiments
8.1 Primal Vs Dual Training
8.2 The Effect of Adaptive Coordinate Frequencies
8.3 The Effect of Random-59 Sampling and Lookup-WD
8.4 Adaptive Stopping Criterion
8.5 Overall Speed-Up
9 Conclusion
References
Attributes for Understanding Groups of Binary Data
1 Introduction
2 Logical Methods for Binary Data Characterization Problems
2.1 Basic Principles: Boolean Functions
2.2 Formulation of the Problem
2.3 Minimal Sets of Attributes
2.4 Patterns
3 Computation of Prime Patterns and Group Covers
3.1 Boros's Algorithm
3.2 Accelerated Algorithm
4 Find a Cover for a Group with the Minimal Number of Patterns
5 Correlation Based Feature Selection (CFS)
6 Experimental Study
6.1 Data Instances
6.2 Comparision Between Algorithm of Boros and Accelerated Algorithm
6.3 Experiments on Sets of Attributes
7 Conclusion
References
Annealing by Increasing Resampling
1 Introduction
2 Annealing by Increasing Resampling as an Approximation of Simulated Annealing
2.1 Acceptance Criterion in Simulated Annealing
2.2 Acceptance Criterion in Annealing by Increasing Resampling
2.3 Generalized Annealing Algorithm as a Unified View
2.4 Compatibility of Annealing Schedules in SA and AIR
3 Experimental Comparison of Acceptance Criteria
4 Experiments in Sparse Pivot Selection
4.1 Dimension Reduction Simple-Map
4.2 Setting of Experiments
4.3 Speed up by Improved Evaluation of Objective Function
4.4 Reuse of Resampling in AIR
4.5 Annealing Schedules of SA and AIR
4.6 Comparison in Different Number of State Transition Trials
4.7 Comparison in Samples of Different Size
5 Experiments in Annealing-Based Clustering
6 Conclusions
References
Applications
Implications of Z-Normalization in the Matrix Profile
1 Introduction
2 Related Work
3 Properties of the Z-Normalized Euclidean Distance
3.1 Definition
3.2 Link with Pearson Correlation Coefficient
3.3 Distance Bounds
3.4 Best and Worst Matches
3.5 Effects of Noise on Self-similarity
4 Flat Subsequences in the Matrix Profile
4.1 Running Example
4.2 Eliminating the Effect of Noise
5 Use Case: Anomaly Detection
6 Use Case: Semantic Segmentation for Time Series
7 Use Case: Data Visualization
8 Conclusion
References
Enforcing the General Planar Motion Model: Bundle Adjustment for Planar Scenes
1 Introduction
2 Related Work
3 Theory
3.1 Problem Geometry
3.2 Camera Parameterisation
4 Prerequisites
4.1 Geometric Reprojection Error
4.2 The Levenberg–Marquardt Algorithm
4.3 Obtaining an Initial Solution for the Camera Parameters
4.4 Obtaining an Initial Solution for the Scene Points
5 Planar Motion Bundle Adjustment
5.1 Block Structure of the Jacobian
5.2 Utilising the Sparse Structure
6 Experiments
6.1 Initial Solution
6.2 Impact of Pre-processing Steps
6.3 Bundle Adjustment Comparison
7 Conclusion
References
Deep Multi-biometric Fusion for Audio-Visual User Re-Identification and Verification
1 Introduction
2 Related Work
2.1 Deep Face Recognition
2.2 Deep Voice Recognition
2.3 Deep Audio-Visual Recognition
3 The Proposed Intermediate Fusion Approach
4 Experimental Evaluation
4.1 Training and Testing Datasets
4.2 Evaluation Setup and Protocols
4.3 Re-Identification Results
4.4 Verification Results
5 Conclusions, Open Challenges, and Future Directions
References
Author Index