Brain-Inspired Computing. 4th International Workshop, BrainComp 2019 Cetraro, Italy, July 15–19, 2019 Revised Selected Papers

Description

Preface Organization Contents Machine Learning and Deep Learning Approaches in Human Brain Mapping A High-Resolution Model of the Human Entorhinal Cortex in the ‘BigBrain’ – Use Case for Machine Learning and 3D Analyses 1 Introduction 2 Material and Methods 2.1 Histological Processing and 3D-Reconstruction of ‘BigBrain’ 2.2 Border Definition and Annotation of the EC in A3D 2.3 Segmentation of Pre-α Islands in Ilastik 2.4 Analysis of Pre-α Islands in ImageJ 2.5 Visualization of the EC and the Included Pre-α Islands in the Context of the Entire ‘BigBrain’ Data Set 3 Results 3.1 Overview of the Layers of the EC 3.2 Cytoarchitecture of Layer 2 (Pre-α Islands) and Modifications Along the Rostrocaudal Extent 3.3 Surface and Morphological Features of Pre-α Islands in EC 3.4 Number and Distribution of Pre-α Islands 4 Discussion References Deep Learning-Supported Cytoarchitectonic Mapping of the Human Lateral Geniculate Body in the BigBrain 1 Introduction 2 Materials and Methods 2.1 Histology 2.2 Manual Analysis and Reference Mapping of Histological Sections 2.3 Training of the Deep-Learning Algorithm to Predict Missing Delineations 3 Results 3.1 Cytoarchitectonic Mapping Based on Expert Annotations and Deep Learning 3.2 High-Resolution 3D Reconstruction 3.3 Volumes of Layers 4 Discussion and Conclusion References Brain Modelling and Simulation Computational Modelling of Cerebellar Magnetic Stimulation: The Effect of Washout 1 Cerebellar Transcranial Magnetic Stimulation 2 Experimental Protocols 3 Computational Modelling 4 Comparative Analysis 5 Discussion and Conclusions References Usage and Scaling of an Open-Source Spiking Multi-Area Model of Monkey Cortex 1 Introduction 2 Overview of the Multi-Area Model 3 The Multi-Area Model Workflow 4 Example Usage 5 Strong Scaling 6 Conclusions References Exascale Compute and Data Infrastructures for Neuroscience and Applications Modular Supercomputing for Neuroscience 1 Introduction 2 The Modular Supercomputing Architecture (MSA) 3 Current Hardware Platforms 3.1 JURECA Cluster-Booster 3.2 DEEP-EST Prototype 4 Software Environment 4.1 Scheduling 4.2 Programming Environment 5 Neuroscience Workflow on MSA 5.1 NEST 5.2 Arbor 6 Summary References Fenix: Distributed e-Infrastructure Services for EBRAINS 1 Introduction 2 Fenix Concept 3 Fenix Compute and Data Services 4 Selected EBRAINS Services 5 Resource Allocation 6 Summary and Outlook References Independent Component Analysis for Noise and Artifact Removal in Three-Dimensional Polarized Light Imaging 1 Introduction 2 Methods 2.1 Preparation of Brain Sections 2.2 Three-Dimensional Polarized Light Imaging (3D-PLI) 2.3 Segmentation of White and Gray Matter 2.4 Independent Component Analysis (ICA) 2.5 Automatic Noise Removal with ICA 3 Results 4 Discussion References Exascale Artificial and Natural Neural Architectures Brain-Inspired Algorithms for Processing of Visual Data 1 Introduction 2 Brain-Inspired Processing of Visual Data 2.1 Edge and Line Detection 2.2 Object(-part) Detection 2.3 Inhibition for Image Processing 3 Convolutional Networks for Visual Data Processing 3.1 Inhibition in Convolutional Networks 4 Conclusions References An Hybrid Attention-Based System for the Prediction of Facial Attributes 1 Introduction 2 Prediction of Facial Attributes 2.1 The Hierarchical HMAX Network 2.2 Local Texture Description Based on LBP 2.3 Binary Classification with Support Vector Machines 3 Experimental Results 4 Conclusion References The Statistical Physics of Learning Revisited: Typical Learning Curves in Model Scenarios 1 Introduction 2 Statistical Physics of Learning: Learning Curves 2.1 Learning a Linearly Separable Rule: Student and Teacher 2.2 The Density of Input Data 2.3 Generalization Error and the Perceptron Order Parameter 2.4 Training as a Stochastic Process and Thermal Equilibrium 2.5 Disorder Average and High-Temperature Limit 2.6 Two Concrete Examples 3 Summary and Conclusion References Emotion Mining: from Unimodal to Multimodal Approaches 1 Introduction 2 Emotion Theories 2.1 Discrete Theories of Emotions 2.2 Dimensional Emotional Models 3 Basic Unimodal Emotion Recognition Approaches 3.1 Emotion Recognition from Textual Sources 3.2 Affective Computing Methodologies 3.3 Emotion Recognition from Facial Expression 3.4 Emotion Recognition from Speech 4 Deep Learning Algorithms for Emotion Detection 5 Challenges and Tools for Multimodal Emotion Recognition 5.1 Existing Multimodal Dataset for Emotion Recognition 6 Conclusions References Author Index