Machine Learning, Optimization, and Data Science

Name: Machine Learning, Optimization, and Data Science | 6th International Conference, LOD 2020, Siena, Italy, July 19-23, 2020, Revised Selected Papers, Part II
Brand: Springer
Price: 117.69 EUR
Availability: OnlineOnly

6th International Conference, LOD 2020, Siena, Italy, July 19-23, 2020, Revised Selected Papers, Part II

Giuseppe Nicosia Varun Ojha Emanuele La Malfa Giorgio Jansen Vincenzo Sciacca Panos Pardalos Giovanni Giuffrida Renato Umeton(Herausgeber*in)

Springer (Verlag)

Erschienen am 6. Januar 2021

XXXVIII, 666 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-030-64580-9 (ISBN)

117,69 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
Preface
Organization
Contents - Part II
Contents - Part I
Multi-kernel Covariance Terms in Multi-output Support Vector Machines
1 Introduction
2 Related Work
3 The Proposed Model
4 Results and Discussion
4.1 Jura Dataset
4.2 Preliminary Results
4.3 Discussion
5 Conclusions
References
Generative Fourier-Based Auto-encoders: Preliminary Results
1 Introduction
2 Problem Formulation
3 Our Framework
4 Experiments
References
Parameterized Structured Pruning for Deep Neural Networks
1 Introduction
2 Parameterized Pruning
2.1 Parameterization
2.2 Regularization
2.3 Pruning
2.4 Hardware-Friendly Structures in CNNs
3 Experiments
3.1 Pruning Different Structures
3.2 CIFAR10/100 and ImageNet
3.3 Ablation Experiments
4 Related Work
5 Conclusion
References
FoodViz: Visualization of Food Entities Linked Across Different Standards
1 Introduction
2 Food Named-Entity Recognition
3 Food Data Normalization
4 FoodViz
5 Conclusion
References
Sparse Perturbations for Improved Convergence in Stochastic Zeroth-Order Optimization
1 Introduction
2 Related Work
3 Sparse Perturbations in SZO Optimization for Nonconvex Objectives
4 Convergence Analysis for Sparse SZO Optimization
5 Algorithms for Sparse SZO Optimization
5.1 Masking Strategies
5.2 Sparse Perturbations with Pruning or Freezing
6 Experiments
6.1 Task and Datasets
6.2 Architectures and Experimental Settings
6.3 Experimental Results
7 Conclusion
A Notation
B Proof for Lemma1
C Proof for Lemma 2
D Proof for Lemma 3
E Proof for Theorem 1
F Miscellaneous
G Empirical Sparsity
References
Learning Controllers for Adaptive Spreading of Carbon Fiber Tows
1 Introduction to Spreading of Carbon Fiber Tows
1.1 Related Work
2 The Process Model - A Supervised System Predictor
2.1 Data Acquisition
2.2 Models
3 A Process Control Model
4 Evaluation
4.1 Process Model
4.2 Process Control Model
5 Conclusion
References
Ensemble Kalman Filter Optimizing Deep Neural Networks: An Alternative Approach to Non-performing Gradient Descent
1 Introduction
2 Ensemble Kalman Filter Optimizing Neural Networks
2.1 Description and Properties of the Ensemble Kalman Filter
2.2 Experimental Setup
3 Numerical Results
3.1 Non Evolving Gradients and Activation Values with SGD
3.2 Slowly Evolving Gradients and Activation Values with ADAM
3.3 Performance of the Ensemble Kalman Filter
4 Conclusions and Outlook
References
Effects of Random Seeds on the Accuracy of Convolutional Neural Networks
1 Introduction
2 Background
3 Experimental Setup
4 Results and Discussion
5 Conclusion
6 Future Work
References
Benchmarking Deep Learning Models for Driver Distraction Detection
1 Introduction
2 Literature Review
2.1 Traditional Machine Learning Approaches
2.2 Deep Learning
3 Methodology
3.1 An Overview of CNNs and RNNs
3.2 Evaluation Metrics
4 Experiments
4.1 The American University in Cairo Distracted Driver Dataset
4.2 Hyper-parameter Configuration
4.3 Evaluation Protocol
5 Results and Discussion
6 Conclusions and Future Work
References
Chronologically Guided Deep Network for Remaining Useful Life Estimation
1 Introduction
2 Problem Setting
2.1 Chronological Loss
3 Model Architectures
3.1 Fully Convolutional Network
3.2 Recurrent Networks
4 Experiments
4.1 Comparative Baselines
4.2 Datasets
4.3 Experimental Setup
4.4 Predictive Performance
4.5 Loss Analysis
4.6 Degradation Analysis
4.7 Sensor Importance
5 Conclusion and Future Work
References
A Comparison of Machine Learning and Classical Demand Forecasting Methods: A Case Study of Ecuadorian Textile Industry
1 Introduction
2 Related Works
3 Case Study Description
4 Demand Forecasting Methods
5 Results and Discussion
6 Conclusions
References
Automatic Curriculum Recommendation for Employees
1 Introduction
2 Problem Setting
3 Design Considerations
4 Content-Based Recommendations
4.1 Cross-Domain Similarity
4.2 Similarity Scoring
5 Interaction-Based Personalization
5.1 Collaborative Filtering
5.2 Multi-interaction Model
5.3 Combining Scores of Content and Interest-Based Recommendations
5.4 Implementation
6 Conclusion and Future Work
References
Target-Aware Prediction of Tool Usage in Sequential Repair Tasks
1 Introduction
2 Related Work
3 Models
3.1 Problem Definition
3.2 Variable Order Markov Model
3.3 Recurrent Neural Network Model
3.4 Integrating the Task Target in a VMM
3.5 Integrating the Task Target in an RNN Model
4 Experiments
4.1 Dataset Description
4.2 Prediction Without Target-Awareness
4.3 Target-Aware Prediction
5 Conclusion
References
Safer Reinforcement Learning for Agents in Industrial Grid-Warehousing
1 Introduction
2 Related Work
2.1 Safe Reinforcement Learning
3 Background
3.1 Safer Policy Updates
3.2 Safe Exploration
4 Safer Dreaming Variational Autoencoder
4.1 Implementation
4.2 Exploration and Policy Update Constraints
5 Results
5.1 Predictive Model
5.2 Failure Rate
6 Conclusion and Future Work
References
Coking Coal Railway Transportation Forecasting Using Ensembles of ElasticNet, LightGBM, and Facebook Prophet
1 Introduction
2 Literature Review
3 Research Methods
3.1 Russian Railways Coking Coal Freight Transportation Dataset
3.2 Ensembling of ElasticNet, LightGBM, and Facebook Prophet
3.3 Forecasting Quality Measurement
4 Results
4.1 Coking Coal Export Transportation Forecasting
4.2 Coking Coal Domestic Transportation Forecasting
4.3 On Coking Coal Import and Transit Transportation Forecasting
5 Discussion
References
Multi-parameter Regression of Photovoltaic Systems using Selection of Variables with the Method: Recursive Feature Elimination for Ridge, Lasso and Bayes
1 Introduction
2 Method
3 Experimental Results
3.1 Description of the Data Set and Data Preprocessing
3.2 Experiments and Results
4 Validation of Results
4.1 Linearity
4.2 Normality of Error Terms
4.3 Non-multicollinearity Between Predictors - CORRELATION
4.4 No Autocorrelation of the Error Terms
4.5 Homocedasticity
5 Conclusion
References
High-Dimensional Constrained Discrete Multi-objective Optimization Using Surrogates
1 Introduction
2 An Algorithm for Constrained Discrete Multi-objective Optimization Using Surrogates
2.1 Algorithm Description
2.2 Radial Basis Function Interpolation
3 Numerical Results on the Large-Scale Mazda Benchmark
4 Summary
References
Exploring Gaps in DeepFool in Search of More Effective Adversarial Perturbations
1 Introduction
2 Related Work
2.1 Technical Background
3 Attack Algorithm
4 Exploiting Gaps in the Linear Approximation of the Decision Boundaries
5 Evaluating Gaps in the Linear Approximation of the Decision Boundary
5.1 Case of Study
5.2 Analysis of the Results
6 Conclusion
References
Lottery Ticket Hypothesis: Placing the k-orrect Bets
1 Introduction
1.1 Need for Over-Parametrization
1.2 Network Sparsifying Techniques
1.3 Lottery Tickets and Network Pruning
1.4 Rewind to k
2 Various Approaches to Pruning
2.1 Pruning at Initialization
2.2 Pruning with Rewinding to k
2.3 Optimal Value of k
2.4 Training in Constrained Setting
3 Our Work
4 Experimental Setup
4.1 Datasets and Architectures Used
4.2 Hyperparameter Selection
4.3 Pruning Method
5 Results
5.1 Initialization Experiment
5.2 Rewind to k Experiment
6 Conclusion
7 Future Work
References
A Double-Dictionary Approach Learns Component Means and Variances for V1 Encoding
1 Introduction
2 Sparse Coding with Mean and Variance Dictionaries
3 Encoding of Natural Image Patches
4 Conclusion
References
A Forecasting Model to Predict the Demand of Roses in an Ecuadorian Small Business Under Uncertain Scenarios
1 Introduction
1.1 Bibliographic Review
2 Materials and Methods
2.1 Phase 1: Macro-Forecast of Tons of Ecuadorian Roses to Be Exported
2.2 Phase 2: Micro-Forecast of Rose Varieties Offered
3 Results
3.1 Results of the Macro-Forecast (Phase 1)
3.2 Forecast of the Demand for the Different Varieties of Roses (Phase 2)
4 Conclusions
Annexes
References
Steplength and Mini-batch Size Selection in Stochastic Gradient Methods
1 Introduction
2 The Method: Steplength Selection via Ritz and Harmonic Ritz Values
3 Numerical Experiments
References
Black Box Algorithm Selection by Convolutional Neural Network
1 Introduction
2 Related Works
2.1 Portfolio Approaches and Ensemble Approaches
2.2 Exploratory Landscape Analysis Features and Algorithm Selection Approaches
2.3 Deep Learning and Its Application to Algorithm Selection Problem
3 Algorithm Selection by Convolutional Neural Network
3.1 Extract Information from Optimization Problems
3.2 Convolutional Neural Network
4 Experimental Results
4.1 Data Description
4.2 Training Details
4.3 Classification Results of Classifying Problem Instances to Different Classes
4.4 Classification Results of Algorithm Selection
4.5 Comparisons of Fitness Results
5 Conclusion and Future Works
References
A Unified Approach to Anomaly Detection
1 Introduction
2 Autoencoders for Anomaly Detection
3 Data
4 Methodology
5 Results and Discussion
6 Conclusion
References
Skin Lesion Diagnosis with Imbalanced ECOC Ensembles
1 Introduction
2 Skin Lesion Classification with ECOC Ensemble
2.1 Base Models
2.2 The ImbECOC Framework
3 Experiments and Results
3.1 Dataset and Problem Definition
3.2 Base Networks
3.3 ImbECOC Model
3.4 Data Augmentation
3.5 Results and Discussion
4 Conclusions
References
State Representation Learning from Demonstration
1 Introduction
2 Related Work
3 Learning a State Representation from Demonstration
3.1 Demonstrations
3.2 Imitation Learning from Demonstration
4 Experimental Setup
5 Results
6 Conclusion
References
A Deep Learning Based Fault Detection Method for Rocket Launcher Electrical System
1 Introduction
2 Related Work
2.1 Deep Learning
2.2 Deep Learning Based Fault Detection Method
3 Method
3.1 Problem Description
3.2 A DL-Based Fault Detection Method for Rocket Launcher Electrical System
4 Experimental Results
4.1 Dataset Setup
4.2 Results and Performance
5 Conclusion
References
Heuristic Search in LegalTech: Dynamic Allocation of Legal Cases to Legal Staff
1 Introduction
2 Background
3 The Matter Allocation Problem (MAP)
4 Solution Approach
4.1 Representation
4.2 Algorithms
5 Experimental Setup
5.1 Problem Generator
5.2 Algorithms and Problem Parameter Settings
5.3 Experimental Results
6 Conclusion and Future Work
References
Unsupervisedly Learned Representations - Should the Quest Be Over?
1 Introduction
2 The Reinforcement Learning Framework in a Nutshell
3 Learning Representations by Means of Reinforcement Learning
4 Unsupervised or Supervised Learning?
5 Discussion and Conclusions
References
Bayesian Optimization with Local Search
1 Introduction
2 Bayesian Optimization with Local Search
2.1 Generic Multi-start Algorithms
2.2 Bayesian Optimization
2.3 The BO with LS Algorithm
3 Numerical Examples
3.1 Mathematical Test Functions
3.2 Logistic Regression
4 Conclusions
A Construction of the GP Model
B The Mathematical Test Functions
References
Sparsity Meets Robustness: Channel Pruning for the Feynman-Kac Formalism Principled Robust Deep Neural Nets
1 Introduction
1.1 Notation
1.2 Organization
2 Related Work
3 Regularity and Sparsity of the Feynman-Kac Formalism Principled Robust DNNs' Weights
4 Unstructured and Channel Pruning with AT
4.1 Algorithms
4.2 Theoretical Guarantees
5 Numerical Results
5.1 Model Compression for at ResNet and EnResNets
5.2 RVSM/RGSM Versus ADMM
5.3 Beyond ResNet Ensemble and Beyond CIFAR10
5.4 Beyond ResNet Ensemble
5.5 Beyond CIFAR10
6 Concluding Remarks
A Proof of Theorem 1
B Adversarial Attacks Used in This Work
C More Visualizations of the DNNs' Weights
References
Limits of Transfer Learning
1 Introduction
2 Distinctions from Prior Work
3 Background
3.1 Transfer Learning
3.2 The Search Framework
3.3 Decomposable Probability-of-Success Metrics
3.4 Casting Transfer Learning into the Search Framework
4 Preliminaries
4.1 Affinity
5 Theoretical Results
6 Examples and Applications
6.1 Examples
6.2 Transferability Heuristic
7 Conclusion
References
Learning Objective Boundaries for Constraint Optimization Problems
1 Introduction
2 Related Work
3 Background
4 Learning to Estimate Boundaries
4.1 Instance Representation
4.2 Eliminating Inadmissible Estimations
4.3 Estimated Boundaries During Search
5 Experimental Evaluation
5.1 Avoiding Inadmissible Estimations
5.2 Estimating Tighter Domain Boundaries
5.3 Effects on Solver Performance
6 Conclusion
References
Hierarchical Representation and Graph Convolutional Networks for the Prediction of Protein-Protein Interaction Sites
1 Introduction
2 Materials and Methods
2.1 Representations
2.2 Input Features
2.3 Graph Convolutional Networks
3 Experiments
3.1 Dataset
3.2 Implementation and Results
4 Conclusions and Future Work
References
Brain-Inspired Spike Timing Model of Dynamic Visual Information Perception and Decision Making with STDP and Reinforcement Learning
1 Introduction
2 Model Structure
3 Simulation Results and Discussion
4 Conclusions
References
Automatic Classification of Low-Angle Fuze-Quick Craters Using Deep Learning
1 Introduction
2 Related Work
2.1 Deep Convolutional Neural Networks
3 Methodology
3.1 Data Collection
3.2 Data Preprocessing
3.3 Feature Extraction
3.4 Classification
4 Experiments and Results
5 Conclusions
References
Efficient Text Processing via Context Triggered Piecewise Hashing Algorithm for Spam Detection
1 Introduction
2 Related Work
3 Method
3.1 High Level Architecture
3.2 Text Preprocessing
3.3 CTPH and Preliminary Grouping
3.4 DBSCAN Clustering
3.5 Longest Common Substring Extraction and Validation
3.6 Performance
4 Experiments
5 Results
References
Machine Learning for Big Data Analysis in Drug Design
1 Introduction
2 Statement of the Problem
3 Approaches and Methods
3.1 Multi-agent Systems
3.2 Artificial Immune Systems Approach
4 Multi-agent Technology for Computer Molecular Design
5 Experimental Researches
5.1 Data Sets Description
5.2 Simulation Results
6 Conclusion
References
Pareto-Weighted-Sum-Tuning: Learning-to-Rank for Pareto Optimization Problems
1 Introduction
2 Related Work
3 Method Developed
3.1 Theoretical Analysis
4 Model Application
5 Experimental Results
5.1 Low Tolerance Sample Decision-Maker
5.2 High Tolerance Sample Decision-Maker
5.3 Analysis of Experiments
6 Conclusions
References
Fast Hyperparameter Tuning for Support Vector Machines with Stochastic Gradient Descent
1 Problem
2 Solution - Main Contribution
3 Theoretical Analysis
3.1 Bounds for M
4 Method
4.1 Computational Performance
5 Experiments
6 Conclusion
References
PlattForm: Parallel Spoken Corpus of Middle West German Dialects with Web-Based Interface
1 Introduction
2 Related Works
3 Data Collection
4 Extraction of Relevant Data
5 Text-to-Speech Conversion for Audio Alignment
6 Searching
7 Getting Audios for Dialects
8 Results and Discussions
8.1 Results
8.2 Discussions
9 Conclusion and Future Works
References
Quantifying Local Energy Demand Through Pollution Analysis
1 Introduction
2 Related Work
2.1 Modeling Oil Spot Price Distributions
2.2 Learning from Air Pollution Data
3 Datasets
3.1 AirData
3.2 Petroleum Data
4 Description of Analytic, Algorithms
4.1 Time-Lagged Cross Correlation
4.2 TLCC Performance
4.3 TLCC Results
5 Application Design
5.1 Extract, Transform, Load Design
6 Area Scoring Analytic
7 Analysis and Visualization
8 Conclusion
9 Future Work
References
On Graph Learning with Neural Networks
1 Introduction
2 Preliminaries
2.1 Graph Learning
2.2 Neural Network
3 Graph Neural Network
3.1 Representation of Graph Learning
3.2 Graph Convolution Layer
3.3 GNN Categorizations
4 Task-Based Categorizations
4.1 Node-Level
4.2 Edge-Level
4.3 Graph-Level
5 Architecture-Based Categorization
5.1 Recurrent Graph Neural Network (RecGNNs)
5.2 Convolutional Graph Neural Network (ConvGNN)
5.3 Graph Autoencoder (GAEs)
5.4 Graph Reinforcement Learning
5.5 Attention-Based Models
6 Conclusions
References
On Bayesian Search for the Feasible Space Under Computationally Expensive Constraints
1 Introduction
2 Background
2.1 Modelling Constraints with Gaussian Processes
2.2 Classifying the Feasible Space
2.3 Bayesian Search Framework
3 Acquisition Functions
3.1 Probability of Being at the Boundary and Entropy (PBE)
4 Experiments
5 Conclusions
References
A Transfer Machine Learning Matching Algorithm for Source and Target (TL-MAST)
1 Introduction
2 Foundations and Related Work
3 Methodological Approach
4 An Algorithm for Optimizing the Source-Target Matching in a Sequential Transfer
4.1 Transferability
4.2 Sequential Transfer Approach
4.3 TL-MAST Algorithm
5 Evaluation and Results
5.1 Case 1: Predicting Sales for Restaurants
5.2 Case 2: Predicting Stock Values
5.3 TL-MAST Results
6 Conclusion
References
Machine Learning and Statistical Models for the Prevalence of Multiple Sclerosis
1 Introduction and Motivation
2 Finding the Center of Population a Country
3 Converting Geographic to Cartesian Coordinates
4 Predictive Models for Multiple Sclerosis Prevalence
5 Linear Regression for Multiple Sclerosis Prevalence
6 Summary and Conclusions
References
Robust and Sparse Support Vector Machines via Mixed Integer Programming
1 Introduction
2 Linear Binary Classification
2.1 Introducing Binary Variables
2.2 Our Approach
2.3 A MIO Formulation
2.4 Solving the Problem Using Gurobi
2.5 Computational Cost
3 Experiments on Synthetic Data Sets
3.1 Experimental Setup
3.2 Results
4 Experiments on Real Data Sets
4.1 Experimental Setup
4.2 Results and Discussion
5 Conclusion
References
Univariate Time Series Anomaly Labelling Algorithm
1 Introduction
1.1 The Risk of Rejecting or Accepting Data as Anomalies
1.2 Anomaly Labelling Problem Specification
2 Univariate Time Series Anomaly Labelling (UTAL) Algorithm
2.1 Detection of Suspected Anomalies
2.2 Spatiotemporal Analysis, Scoring and Labelling of Anomalies
2.3 Experiments
3 Result
3.1 Experiment with a Small Modified District Heating Dataset
3.2 Experiment with Open Source Datasets from Yahoo and NAB
3.3 Experiment with District Heating Substation Energy Profiles
4 Related Work and Discussion
5 Conclusions
References
Challenges in Real-Life Face Recognition with Heavy Makeup and Occlusions Using Deep Learning Algorithms
1 Introduction
2 Artificial Neural Networks
3 Data Set
4 Experiments
5 Results
6 Discussion and Future Work
References
Who Accepts Information Measures?
1 Introduction
2 Behavioral Experiment
3 Results
3.1 Monotone Influence of Psychological Scores
3.2 Acceptability of Each Axiom, Separately
4 Conclusions
References
An Error-Based Addressing Architecture for Dynamic Model Learning
1 Introduction
2 Methods
2.1 Clustering of Training Data
2.2 Training
3 Results
3.1 Mountaincar Environment
3.2 Acrobot Environment
3.3 Prediction of Cosine and Sine Values - Weighted Angle
3.4 Phase-Functioned Neural Network Database
4 Discussion
A Appendix - Performance of the Architecture in the Acrobot Environment
References
Learn to Move Through a Combination of Policy Gradient Algorithms: DDPG, D4PG, and TD3
1 Introduction
2 Methods
2.1 Combination of Algorithms
2.2 Comparison of Deterministic Policy Gradient Algorithms
2.3 OpenSim Environment
3 Experiments
3.1 Results of the Experiments
4 Discussion
A Appendix
A.1 Deterministic Policy Gradient Algorithms
A.2 Twin-delayed Deep Deterministic Policy Gradient (TD3)
A.3 Distributed Distributional Deep Deterministic Policy Gradient (D4PG)
References
1 Regularized Robust and Sparse Linear Modeling Using Discrete Optimization
1 Introduction
2 Mixed Integer Optimization Formulation
2.1 Introducing Binary Variables
2.2 MIO Formulation of Problem (4)
2.3 MIO Formulation of Problem (5)
3 Discrete First Order Algorithms
3.1 Discrete First Order Algorithm for Problem (4)
3.2 Discrete First Order Algorithm for Problem (5)
4 Experiments on Synthetic Data Sets
4.1 Setup
4.2 Selecting Tuning Parameters
4.3 Experiment 2 on Synthetic Data
4.4 Computational Time
4.5 Results
5 Experiments on Real Data Sets
6 Conclusion
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken