Advances in Neural Computation, Machine Learning, and Cognitive Research VI: Selected Papers from the XXIV International Conference on ... (Studies in Computational Intelligence, 1064)

This book describes new theories and applications of artificial neural networks, with a special focus on answering questions in neuroscience, biology and biophysics and cognitive research. It covers a wide range of methods and technologies, including deep neural networks, large-scale neural models, brain–computer interface, signal processing methods, as well as models of perception, studies on emotion recognition, self-organization and many more. The book includes both selected and invited papers presented at the XXIV International Conference on Neuroinformatics, held on October 17–21, 2022, in Moscow, Russia. Preface Organization Editorial Board Advisory Board Program Committee of the XXIV International Conference “Neuroinformatics-2022” General Chair Co-chairs Program Committee Contents Neuroinformatics and Artificial Intelligence Tree Inventory with LiDAR Data 1 Introduction 2 Methods 2.1 Preprocessing Data 2.2 Cluster Analysis 2.3 Circle Fitting and Convex Hull 3 Experiment 4 Conclusion References Towards Reliable Solar Atmospheric Parameters Neural-Based Inference 1 Introduction 2 Data and Methods 3 Results 4 Conclusion and Discussion References Addressing Task Prioritization in Model-based Reinforcement Learning 1 Introduction 2 Background 2.1 Reinforcement Learning 2.2 Model-Based RL via World Models 3 Task Addressing for Prioritized Implicit Generalization 3.1 Addressing Mechanism 3.2 Expected Reward Optimization 4 Related Work 5 Experiments 6 Discussion and Conclusion References Automatic Generation of Conversational Skills from Dialog Datasets 1 Introduction 1.1 Related Work 2 Methodology 2.1 Pre-trained Models 2.2 Data 2.3 Evaluation 3 Experiments 4 Discussion 5 Conclusion References Neural Networks and Cognitive Sciences. Adaptive Behavior and Evolutionary Simulation Individual Topology Structure of Eye Movement Trajectories 1 Introduction 2 Materials and Methods 2.1 Statistics of Fixations and Saccades 2.2 Topological Data Analysis (TDA) 2.3 Dataset 2.4 Experiment 3 Results A Additional Definitions and Notation References Neural Network Providing the Involvement of Voluntary Attention into the Processing and Conscious Perception of Sensory Information 1 Introduction 2 Neural Mechanism for the Involvement of Attention in the Processing of Sensory Information in C-BG-Th-C Loops 3 Mechanism for Inclusion of Voluntary Attention into the Processing and Awareness of Sensory Information in the Neural Network Consisting of the Neocortex, Hippocampus, Thalamus, Basal ganglia and Cerebellum 4 Functioning of the Network in the Absence of Voluntary Attention Leading to the Inattentional Blindness 5 Comparison of Consequences of the Proposed Mechanism of Inattentional Blindness with Experimental Data 6 Conclusion References Alpha Rhythm Dynamics During Spoken Word Recognition 1 Introduction 2 Methods 3 Results: Alpha Frequency Jumps 4 Discussion References Robotic Devices Control Based on Neuromorphic Classifiers of Imaginary Motor Commands 1 Introduction 2 Contactless Control of Devices 3 Neuromorphic Classifiers of Imaginary Commands 4 Application of the Neuromorphic Classifier for Robot Control 5 Conclusion References A Software System for Training Motor Imagery in Virtual Reality 1 Introduction 2 Methods 2.1 General Pipeline 2.2 Virtual Reality Application 2.3 Classifier Program 3 Discussions 4 Conclusion References On the Importance of Diversity 1 Introduction 2 The Base Model of Segregated Groups 3 Examples of the Resource Distribution Principle 3.1 Resource Distribution According to Labor 3.2 Development-Oriented Resource Distribution 3.3 Equal Resource Distribution 4 The Intergroup Transition Model 5 Conclusion References “MYO-chat” – A New Computer Control System for People with Disabilities 1 Introduction 2 Overview of Similar Systems and Analysis of the Subject Area 3 Architecture of the MYO-Chat System 4 Hardware Part of the “MYO-chat” System 5 Software Part of the “MYO-chat” System 6 Discussion 7 Conclusion References Low-Bit Quantization of Transformer for Audio Speech Recognition 1 Introduction 2 Related Work 3 Model Architecture 4 Quantization Methodology 4.1 Quantization Scheme 4.2 Offsets 4.3 Initialization 4.4 Quantization Configuration 5 Problem Layers Handling 5.1 Problem Layers Definition 5.2 Problem Quantization Configuration 5.3 Blockwise Knowledge Distillation 6 Experiments 6.1 Offsets 6.2 Initialization 6.3 Problem Layers Handling 7 Conclusion References A Model of Predicting and Using Regularities by an Autonomous Agent 1 Introduction 2 Model Description 2.1 Agent and Environment 2.2 Description of the Simplified Variant of the Model 2.3 Description of the Neural Network Variant of the Model of Prediction and Use of Regularities by an Autonomous Agent 2.4 An Evolving Population 3 Results of Computer Simulation 3.1 Analysis of Processes of Learning of One Agent 3.2 Analysis of Processes of Learning and Evolution in a Population of Agents 4 Conclusions References A Review of One-Shot Neural Architecture Search Methods 1 Introduction 2 Search Space Design 3 One-stage NAS 3.1 Differentiable Search Space 3.2 Non-gradient Based Methods 4 Two-stage NAS 4.1 Low-level Sampling Techniques 4.2 High-level Sampling Techniques 5 Conclusion 6 Appendix References Does a Recurrent Neural Network Use Reflection During a Reflexive Game? 1 Introduction 2 Materials and Methods 3 Results and Discussion 4 Conclusion References A Gender Genetic Algorithm and Its Comparison with Conventional Genetic Algorithm 1 Introduction 2 Computational Experiment 2.1 Description of Test Tasks 2.2 Evaluation Metrics 2.3 GA and GGA Implementation 2.4 Experiment 1: Comparison of GA and GGA 2.5 Experiment 2: GA with Different Restrictions on the Number of Times a Female May Be Selected 3 Future Work 4 Conclusions References Associations of Morphometric Changes of the Brain with the Levels of IGF1, a Multifunctional Growth Factor, and with Systemic Immune Parameters Reflect the Disturbances of Neuroimmune Interactions in Patients with Schizophrenia 1 Introduction 2 Materials and Methods 3 Results 3.1 Morphometric Parameters of the Brain, Systemic Inflammation and Adaptive Immunity in Schizophrenic Patients 3.2 Immunity Indicators and Clinical Characteristics in Schizophrenic Patients Depending on IGF1 Levels 4 Conclusion References Modern Methods and Technologies in Neurobiology Dynamics of Background and Evoked Activity of Neurons in the Auditory Cortex of the Unanaesthetized Cat 1 Introduction 2 Methods 3 Results 4 Discussion References Search for Markers of Moderate Cognitive Disorders Through Phase Synchronization Between Rhythmic Photostimulus and EEG Pattern 1 Introduction 2 Materials and Methods 3 Results 4 Conclusions References Astrocytes Enhance Image Representation Encoded in Spiking Neural Network 1 Introduction 2 Materials and Methods 3 Discussion References Classification of Neuron Type Based on Average Activity 1 Introduction 2 Materials and Methods 2.1 Data 3 Results 3.1 Problem Statement 3.2 Data Preprocessing 3.3 Selection of Features 3.4 Data Balancing 3.5 Determining the Best Classification Algorithm References Comparative Analysis of Statistical and Neural Network Classification Methods on the Example of Synthetized Data in the Stimulus-Independent Brain-Computer Interface Paradigm 1 Introduction 2 Methods 2.1 Motor Imagery EEG Modelling 2.2 Frequency Spectrum Features 2.3 Time-Frequency Features 2.4 Classification Scenarios 2.5 CNN Model 3 Results and Discussion 3.1 Perfect Segments 3.2 Full 2-s Epochs 3.3 Segment Selection by FC 3.4 CNN Application to Full 2-s Epoch and Segments 4 Conclusions References Shunting Effect of Synaptic Channels Located on Presynaptic Terminal 1 Introduction 2 Mathematical Model 3 Results 4 Conclusion References Analysis of Appearances, Formation and Evolution of Biological Functional Systems 1 Introduction 2 Modeling Negentropy Agents Analogous to Archbiosystems 3 Results 4 Discussion References The Reinforcement Learning Theory, Value Function, and the Nature of Value Function Calculation by the Insular Cortex 1 Introduction 2 What Analogs of the Characteristics of the RL Theory Should Be Found in the Brain 2.1 What is the State s 2.2 What Is Action a 2.3 Brain Control is Flexible Due to Distributed Control Net 2.4 Avoidance Reactions 2.5 How and Where is the Value of s Displayed 2.6 How Should the Choice of Action Be Implemented in the Brain? 2.7 Role of Dopamine in Action Choice 3 The Insular Cortex as a Key Element of the Behavioral Choice Network 4 Participation of Dorsal Anterior Insula in Cognitive Processing in Humans 4.1 The Instrumental Task 4.2 Solving a Cognitive Problem 5 Conclusion References To the Role of Inferior Olives in Cerebellar Neuromechanics 1 Introduction 2 Main Features of the Cerebellar Circuitry 3 Simple and Complex Spikes of Cerebellar Purkinje Cells 4 Mauk Hypothesis and Its Consequences 5 Theoretical Treatment of Convergence of ClFCs Interspike Interval to Equilibrium Value 6 Phase-based Cerebellar Learning of Dynamic Signals 7 Computational Modeling and Physiological Experiments 8 Applications of Cerebellar Equilibrium Model to Prognosis of Dynamic Signals 9 Conclusion References Sleep of Poor and Good Nappers Under the Afternoon Exposure to Very Weak Electromagnetic Fields 1 Introduction 2 Methods 2.1 Samples of Experiment 1, Experiment 2, and Experiment 3 2.2 Procedures 2.3 Data Analysis 3 Results 3.1 Experiment 1 3.2 Experiment 2 3.3 Experiment 3 4 Discussion References Applications of Neural Networks Classification of Light Microscopy Image Using Probabilistic Bayesian Neural Network 1 Introduction 2 Materials and Methods 3 Computational Experiment 4 Conclusions References SPICE Model of Analog Content-Addressable Memory Based on 2G FeFET Crossbar 1 Introduction 2 SPICE Model of 2G FeFET Based on Gate-Drain Current-Voltage Characteristic 3 SPICE Model of 2G FeFET Crossbar 4 Conclusions References IQ-GAN: Instance-Quantized Image Synthesis 1 Introduction 2 Instance-Quantized GAN 2.1 IQ-GAN Framework Overview 2.2 3D Semantic Noise 2.3 IQ-Generator 3 Experimental Results 3.1 Network Training and Baselines 3.2 Network Training 3.3 Qualitative Evaluation 3.4 Quantitative Evaluation 4 Conclusion References Specifics of Crossbar Resistor Arrays 1 Introduction 2 Crossbar Resistor Array 3 Reading the Conductivity Matrix 4 Generating the Conductivity Matrix 5 Conclusions References Recurrent and Graph Neural Networks for Particle Tracking at the BM@N Experiment 1 Introduction 2 Related Work 3 Artificial Neural Networks for the Tracking 3.1 Building Track-Candidates with TrackNETv3 3.2 Graph Neural Network Classifier 4 Experiments 4.1 Dataset 4.2 Local Tracking with TrackNETv3 4.3 Graph Neural Network Results 4.4 Discussion 5 Conclusion References Modeling of a Neural Network Algorithm for Suppressing Non-stationary Interference in an Adaptive Antenna Array 1 Introduction 2 Adaptive Antenna Array 3 Neural Network Algorithm 4 Numerical Simulation 5 Conclusions References Learning Various Locomotion Skills from Scratch with Deep Reinforcement Learning 1 Introduction 2 Related Work 3 Background 4 Our Method 5 Results 6 Conclusion References Center3dAugNet: Effect of Rotation Representation on One-Stage Joint Car Detection and 6D-Pose Estimation 1 Introduction 2 Related Work 3 Problem Definition 4 Training of Model for Joint Car Detection and Pose Estimation 4.1 Model Architecture 4.2 Data Augmentations 4.3 Rotation Representation 4.4 Loss Functions 5 Experiments 6 Conclusions References Global Memory Transformer for Processing Long Documents 1 Introduction 2 Global Slot Memory Augmented Transformer with Hierarchical Attention 3 Masked Language Modeling 3.1 Experiments Setup 3.2 Experimental MLM Task Results 4 Fine Tuning - Question Answering Task 5 Ablation Study 5.1 Ablation Study Results 6 Conclusion and Future Work References Development of Convolutional Neural Network for Defining a Renal Pathology Using Computed Tomography Images 1 Introduction 2 Convolutional Neural Networks in Analysis of CT Images 3 Aspects of the Application of Neural Networks 4 Dataset Description 5 Algorithm Explanation 6 Neural Network Architecture 7 Results 8 Conclusion References Possibility of Using Various Architectures of Convolutional Neural Networks in the Problem of Determining the Type of Rhythm 1 Introduction 2 Preparation of the Training Sample 3 ECG Augmentation 4 Quality Metrics 5 Neural Network Architectures 5.1 ResNet 5.2 DenseNet 5.3 XceptionNet 6 Experiments 6.1 Sensitivity and Specificity 6.2 Confusion Matrix 7 Conclusions References DeepPavlov Topics: Topic Classification Dataset for Conversational Domain in English*-6pt 1 Introduction 2 Related Work 3 Dataset Collection 3.1 Data Sources 3.2 Data Filtering 3.3 Dataset Characteristics 4 Baselines 5 Conclusion A Data Sources References Multi-input Convolutional Neural Networks in Real-Time Semantic Segmentation Tasks 1 Introduction 2 Multi-input Convolutional Neural Networks 3 Data Base for Learning 4 Generation of Artificial Neural Networks 5 Learning of CNN 6 Conclusions References Integration of Data and Algorithms in Solving Inverse Problems of Spectroscopy of Solutions by Machine Learning Methods 1 Introduction 2 Experiment 2.1 Preparation of Solutions 2.2 Raman Spectroscopy 2.3 Optical Absorption Spectroscopy 3 Application of Machine Learning Methods 3.1 Integration of Physical Methods 3.2 Dataset 3.3 Basic Models 3.4 Ensemble Learning 4 Results and Discussion 5 Conclusion References Investigation of Pareto Front of Neural Network Approximation of Solution of Laplace Equation in Two Statements: with Discontinuous Initial Conditions or with Measurement Data 1 Introduction 2 Materials and Methods 2.1 Problem Statement 2.2 Selection of Optimal Solutions 2.3 Evolution Algorithm Based on Pareto Front 3 Computational Experiments and Results 3.1 Problem (1)+(2) 3.2 Problem (1)+(3) 4 Conclusions and Discussion References Multitask Learning for Extensive Object Description to Improve Scene Understanding on Monocular Video 1 Introduction 2 Related Work 3 Problem Definition 4 Multitask Learning for Extension Model Description 4.1 Dataset Expansion via Segmentation Masks 4.2 Model Architecture 4.3 Data Augmentations 4.4 Loss Functions 5 Experiments 5.1 Metrics 5.2 Result Analysis 6 Conclusions References Use of Classification Algorithms to Predict the Grade of Geomagnetic Disturbance 1 Introduction 2 Data Description 3 Methods and Experiments 4 Results 5 Conclusions References Information Processing in Spiking Neuron-Astrocyte Network in Ageing 1 Introduction 2 Model and Architecture of Neuron-Astrocyte Network 2.1 Neural Network 2.2 Astrocytic Network 2.3 Bidirectional Neuron-Astrocyte Interaction 3 Stimulation Protocol 4 Memory Performance Metrics 5 Results 6 Conclusions References Multilingual Case-Insensitive Named Entity Recognition 1 Introduction 2 Related Work 2.1 Robustness to Capitalisation 2.2 Multilinguality 3 Datasets 4 Experimental Setup 5 Results 6 Conclusion References Multilevel Separation Pipeline for Similar Structure Data 1 Introduction 2 Experimental 2.1 Task Description 2.2 Application of Machine Learning Methods (Stage 1) 2.3 Context-Structuring Data 2.4 Organization of Data Processing (Stage 2) 2.5 Multilevel Filtering (Stage 3) 3 Results and Discussion 4 Conclusion References Creating a Brief Review of Judicial Practice Using Clustering Methods 1 Introduction 2 Description of the Abstracts Created by the System 3 The Problem Description 4 A Brief Overview of Related Works 5 The Clustering Module “Court Clustering” 6 The Module for Creating a Brief Review of Judicial Practice 7 The Analytical Module 8 Conclusions References Neural Network Theory, Concepts and Architectures “GAS” Instead of “Liquid”: Which Liquid State Machine is Better? 1 Introduction 2 Materials and Methods 2.1 Liquid State Machine 2.2 Input Nodes 2.3 SNN Serving as “Liquid” 2.4 Finding LSM with Best Memory Using Genetic Algorithm and Coordinate Descent 3 Results and Their Analysis 4 Conclusion References Using a Resistor Array to Tackle Optimization Problems 1 Introduction 2 A Resistor Array Enabling Concurrent Writing and Reading 3 Conductivity Matrices Search Strategy 4 Elaborate Optimization 5 Conclusions References Generative Adversarial Networks as an Approach to Unsupervised Link Prediction Problem 1 Introduction 2 Formalization of the Algorithm 3 Implementation 4 Discussion of Computational Experiments 5 Conclusions References DGAC: Dialogue Graph Auto Construction Based on Data with a Regular Structure 1 Introduction 2 Related Work 3 Dialogue Graph Auto Construction (DGAC) 3.1 Dialogue Graph 3.2 Utterance Embedding 3.3 Utterance Clustering 3.4 Utterance Reclustering 3.5 Node Construction 3.6 Linking 4 Metrics 4.1 Nodes 4.2 Edges 5 Experiments 5.1 One-Stage Node Construction 5.2 Two-Stage Node Construction 5.3 Retrieval Analysis 6 Conclusion 7 Future Work 8 Limitations A Appendix A.1 Response Prediction Using SBERT-MAP A.2 Resources References Relay System of Differential Equations with Delay as a Perceptron Model 1 Introduction 2 Single Neuron Model 3 Perceptron Model 4 Result 4.1 Associative Neuron Coupling with N Sensory Neurons 4.2 Responsive Neuron Coupling with M Associative Neurons 4.3 Proof of Theorem 3 5 Conclusion References Analysis of Predictive Capabilities of Adaptive Multilayer Models with Physics-Based Architecture for Duffing Oscillator 1 Introduction 2 Problem Statement and Model Description 2.1 Duffing Equation with Dynamical Parameter 2.2 Duffing Equation Adaptive Multilayer Model 3 Computational Experiments and Results 4 Conclusions and Discussion References An Attempt to Formalize the Formulation of the Network Architecture Search Problem for Convolutional Neural Networks 1 Introduction 2 Related Works 2.1 Search Space 2.2 NAS Strategies/Algorithms 3 NAS Approach 4 Experiment Description 4.1 Description of the Data Set 4.2 Results 5 Conclusions References Use of Conditional Variational Autoencoders and Partial Least Squares in Solving an Inverse Problem of Spectroscopy 1 Introduction 2 Materials and Methods 2.1 Raman Spectroscopy 2.2 Data Preprocessing 2.3 Neural Networks 2.4 Partial Least Squares 2.5 Gaussian Mixture Module 3 Computational Experiments 4 Conclusions References On the Similarities Between Denoising Diffusion Models and Autoencoders 1 Introduction 2 Denoising Diffusion Models 2.1 Problem Formulation 2.2 Forward Process 2.3 Training Objective 2.4 Reverse Process 3 Connection with Autoencoders 3.1 Sampling 3.2 Training Objective 3.3 Modified DAE Objective 4 Experiments 4.1 Dataset 4.2 Architectures and Optimization 4.3 Evaluation 4.4 Results 5 Conclusion References Author Index