Parallel Information Research Articles

Coarse-to-Fine Feedback Guidance Based Stereo Image Quality Assessment Considering Dominant Eye Fusion

Considering that the human brain always follows a coarse-to-fine (low-to-high spatial frequency) visual processing and fusion mechanism, we propose a coarse-to-fine feedback guidance based stereo image quality assessment (SIQA) network which considers a coarse-to-fine feedback guidance and adaptive dominant eye mechanism. The proposed network consists of two main sub-network streams, each of which has three branches to extract low, middle and high spatial frequency information in parallel. To better realize the guidance of the high-level features in the low spatial frequency branch to the low-level features in the high spatial frequency branch, an information feedback guidance module (IFGM) is proposed, which realizes a top-down guidance mechanism in each sub-network stream. Simultaneously, according to the theory of ocular dominance in human visual system (HVS), we design an adaptive bi-directional parallax-based binocular fusion module (BPBFM), which synthesizes two types of fusion feature by taking the left and right view features as dominant eye input. Furthermore, in order to obtain the better perceptual quality of stereo images, we design a weighted fusion strategy to weigh the quality scores from the two types of fusion features obtained by using an ensemble model with two multi-layer perceptrons (MLPs). The experimental results on four public stereo image datasets show that the proposed method is superior to the mainstream metrics and achieves an excellent performance.

Probing the Dual-Route Model of the SNARC Effect by Orthogonalizing Processing Speed and Depth.

The dual-route model explains the SNARC (Spatial-Numerical Association of Response Codes) effect assuming two routes of parallel information processing: the unconditional route (automatic activation of pre-existing links) and the conditional route (activation of task-specific links). To test predictions derived from this model, we evaluated whether response latency in superficial number processing modulates the SNARC effect in a color task (participants judged the color of a number). In Experiment 1, participants performed a parity task, an easy color task (short RTs), and a difficult color task (RTs similar to those of the parity task). A SNARC effect emerged only in the parity task. In Experiment 2, participants performed a color task and a secondary task under four conditions chosen to orthogonally manipulate response latency (short vs. long) and processing depth (semantic vs. perceptual). Only the long-latency perceptual-processing condition elicited a SNARC effect. To explain these results, we suggest that the cognitive resources required by a secondary task might dilute the SNARC effect. Our results indicate that the dual-route model should be modified to take into account additional factors (e.g., working memory load) that influence the level of activation of the unconditional route.

Моделювання процесу митної переробки вантажів при проходженні транспортних засобів через державний кордон України

Developing a model of the customs cargo handling process that allows making sound technological, design, and management decisions, changing input parameters, and carrying out long-term planning based on information on the occupancy of control points, customs zones of the border customs office, customs teams, throughput capacity of checkpoints, technical means of customs control and communication, types of customs regimes, and standardisation of the duration of all components of customs cargo handling. To develop a model of the customs cargo processing process, it is advisable to use systems of parallel information processing and parallel operation of objects, which include, for example, modelling with the help of Petri nets. The model developed on the basis of Petri nets for describing the customs cargo handling process shows that, in addition to being clear and easy to use, it makes it possible to take into account various probabilistic factors. Having developed a mathematical model of customs processing of cargo when vehicles cross the state border, it can be concluded that along with technological operations (time for customs clearance at the internal customs, time for customs procedures at the checkpoint), there are non-productive operations - downtime (time for waiting for customs clearance at the internal customs, time for waiting for customs procedures at the checkpoint), which significantly affect the time of customs border crossing. The mathematical apparatus of Petri nets will allow to study the dynamics of the system under study and its behaviour under different initial conditions. It has been found that the number of customs control and customs clearance officers has the greatest impact on the time of customs processing of goods; By building a mathematical model based on Petri nets, an effect of 1.11% was achieved. As a result of applying the developed mathematical model of the customs cargo processing process, an effect of 42 minutes was obtained.

Parallel Spectral–Spatial Attention Network with Feature Redistribution Loss for Hyperspectral Change Detection

Change detection methods using hyperspectral remote sensing can precisely identify differences of the same area at different observing times. However, due to massive spectral bands, current change detection methods are vulnerable to unrelatedspectral and spatial information in hyperspectral images with the stagewise calculation of attention maps. Besides, current change methods arrange hidden change features in a random distribution form, which cannot express a class-oriented discrimination in advance. Moreover, existent deep change methods have not fully considered the hierarchical features’ reuse and the fusion of the encoder–decoder framework. To better handle the mentioned existent problems, the parallel spectral–spatial attention network with feature redistribution loss (TFR-PS2ANet) is proposed. The contributions of this article are summarized as follows: (1) a parallel spectral–spatial attention module (PS2A) is introduced to enhance relevant information and suppress irrelevant information in parallel using spectral and spatial attention maps extracted from the original hyperspectral image patches; (2) the feature redistribution loss function (FRL) is introduced to construct the class-oriented feature distribution, which organizes the change features in advance and improves the discriminative abilities; (3) a two-branch encoder–decoder framework is developed to optimize the hierarchical transfer and change features’ fusion; Extensive experiments were carried out on several real datasets. The results show that the proposed PS2A can enhance significant information effectively and the FRL can optimize the class-oriented feature distribution. The proposed method outperforms most existent change detection methods.

Convolutional Neural Network (CNN): The architecture and applications

The human brain is made up of several hundreds of billions of interconnected neurons that process information in parallel. Researchers in the field of artificial intelligence have successfully demonstrated a considerable level of intelligence on chips and this has been termed Neural Networks (NNs). Neural networks, also known as artificial neural networks (ANNs) or simulated neural networks (SNNs), are a subset of machine learning (ML) and they are at the heart of deep learning algorithms. These subsets of ML have their names and structures derived from the human brain and the way that biological neurons signal to one another. A class of NNs that are often used in processing digital data images is the Convolutional Neural Network (CNN or ConvNet). The human brain processes a huge amount of information with each neuron having its own receptive field connected to other neurons in a way that they cover the entire visual field. Mimicking the biological technique, where the neurons only respond to stimuli in the restricted region of the visual field referred to as the receptive field, each neuron in the CNN processes data only in its receptive field. In this review paper, the architecture and application of CNN are presented. Its evolution, concepts, and approaches to solving problems related to digital images, computer vision and are also examined.

Persistent Frustration-Induced Reconfigurations of Brain Networks Predict Individual Differences in Irritability

Aberrant responses to frustration are central mechanisms of pediatric irritability, which is a common reason for psychiatric consultation and a risk factor for affective disorders and suicidality. This pilot study aimed to characterize brain network configuration during and after frustration and test whether characteristics of networks formed during or after frustration relate to irritability. During functional magnetic resonance imaging, a transdiagnostic sample enriched for irritability (N= 66, mean age= 14.0 years, 50% female participants) completed a frustration-induction task flanked by pretask and posttask resting-state scans. We first tested whether and how the organization of brain regions (ie, nodes) into networks (ie, modules) changes during and after frustration. Then, using a train/test/held-out procedure, we aimed to predict past-week irritability from global efficiency (Eglob) (ie, capacity for parallel information processing) of these modules. Two modules present in the baseline pretask resting-state scan (one encompassing anterior default mode and temporolimbic regions and one consisting of frontoparietal regions) contributed most to brain circuit reorganization during and after frustration. Only Eglob of modules in the posttask resting-state scans (ie, after frustration) predicted irritability symptoms. Self-reported irritability was predicted by Eglob of a frontotemporal-limbic module. Parent-reported irritability was predicted by Eglob of ventral-prefrontal-subcortical and somatomotor-parietal modules. These pilot results suggest the importance of the postfrustration recovery period in the pathophysiology of irritability. Eglob in 3 specific posttask modules, involved in emotion processing, reward processing, or motor function, predicted irritability. These findings, if replicated, could represent specific intervention targets for irritability.

VStyclone: Real-time Chinese voice style clone

This paper proposes a novel Chinese speech cloning model named VStyclone, which consists of three stages: multi-speaker training, target speaker encoding, and target speaker synthesis. In this work, we design an efficient tone extractor, which can reallocate resources to the sequences of log-mel spectrogram frames obtained from multiple speakers’ speech, thus allowing the network to learn multiple speakers’ features differently. This approach allows the network to focus more on the voice features of the target speaker and extract the target features accurately. To cluster the voices of the same speaker and sparse the voices of different speakers, we build an optimal softmax loss to optimize the model. Then, we develop a style synthesizer, which adopts the idea of transformer instead of recurrent neural network, so that the model can not only process text information in parallel, but also improve the model's ability to process long-distance contextual information. Meanwhile, we embed a style extraction module in the style synthesizer to dynamically capture style ranges in an unsupervised manner. In addition, the VStyclone model uses generative adversarial networks as the base generation model of the vocoder to improve the generation speed, which runs 1.2 times faster than the real-time generation speed on CPU, and finally the VStyclone model obtains the SOTA effect.

Image-Based Pain Intensity Estimation Using Parallel CNNs with Regional Attention.

Automatic pain estimation plays an important role in the field of medicine and health. In the previous studies, most of the entire image frame was directly imported into the model. This operation can allow background differences to negatively affect the experimental results. To tackle this issue, we propose the parallel CNNs framework with regional attention for automatic pain intensity estimation at the frame level. This modified convolution neural network structure combines BlurPool methods to enhance translation invariance in network learning. The improved networks can focus on learning core regions while supplementing global information, thereby obtaining parallel feature information. The core regions are mainly based on the tradeoff between the weights of the channel attention modules and the spatial attention modules. Meanwhile, the background information of the non-core regions is shielded by the DropBlock algorithm. These steps enable the model to learn facial pain features adaptively, not limited to a single image pattern. The experimental result of our proposed model outperforms many state-of-the-art methods on the RMSE and PCC metrics when evaluated on the diverse pain levels of over 12,000 images provided by the publicly available UNBC dataset. The model accuracy rate has reached 95.11%. The experimental results show that the proposed method is highly efficient at extracting the facial features of pain and predicts pain levels with high accuracy.

Neuroscience of consciousness: cognition, physics and philosophy of decoding the human brain

The biophysical roots of consciousness have been the subject of an ongoing debate for centuries. In order to understand the data, create novel experimental methodologies, and increase our ability to investigate this phenomenon of interest, the proposed theories must lead to empirical, repeatable, and testifiable studies. Contemporary theories of consciousness often do not relate to one another, and none of them has been distinguished as complete or proven empirically so far. The aim of this study is an investigation into some of the possible approaches that could merge neuronal brain activity with the laws of physics and some philosophical principles that may be associated with the emergence of consciousness in the first place. As a result, the relationship between consciousness and attention, working memory, access consciousness and phenomenal consciousness is evaluated. The contrast between conscious and unconscious perception, perceived visual inputs and subliminal ones is investigated to facilitate a discussion about the neural correlates of self-awareness. Consciousness as a global broadcast of information to integrated brain modules is being considered, as well as viewing a brain as a parallel information processor linked to attention inputs. Relationship between consciousness and attention is explored, as well as attention without consciousness and vice versa. Implications and shortcomings of the proposed approaches based on brain science, philosophy and quantum physics are also covered to shed some more light on this ever present experience of being conscious that everyone seems to self-witness but no one manages to adequately explain.

MODENN: A Shallow Broad Neural Network Model Based on Multi-Order Descartes Expansion.

Deep neural networks have achieved great success in almost every field of artificial intelligence. However, several weaknesses keep bothering researchers due to its hierarchical structure, particularly when large-scale parallelism, faster learning, better performance, and high reliability are required. Inspired by the parallel and large-scale information processing structures in the human brain, a shallow broad neural network model is proposed on a specially designed multi-order Descartes expansion operation. Such Descartes expansion acts as an efficient feature extraction method for the network, improve the separability of the original pattern by transforming the raw data pattern into a high-dimensional feature space, the multi-order Descartes expansion space. As a result, a single-layer perceptron network will be able to accomplish the classification task. The multi-order Descartes expansion neural network (MODENN) is thus created by combining the multi-order Descartes expansion operation and the single-layer perceptron together, and its capacity is proved equivalent to the traditional multi-layer perceptron and the deep neural networks. Three kinds of experiments were implemented, the results showed that the proposed MODENN model retains great potentiality in many aspects, including implementability, parallelizability, performance, robustness, and interpretability, indicating MODENN would be an excellent alternative to mainstream neural networks.

Blending multiple algorithmic granular components: a recipe for clustering

Emerging trends in algorithm design have shown that hybrid algorithms, which combine or merge multiple algorithms, can create synergies to overcome the inherent limitations of the underlying individual algorithms. There are two broad types of hybridization: collaborative—individual algorithms tackle an instance of the problem sequentially or in parallel and exchange information accordingly while solving the problem; integrative—individual algorithms are dedicated to tackling different aspect(s) of the problem-solving process. In this research, we propose a schema for an enhanced form of integrative hybridization that blends granular algorithmic components from multiple algorithms to derive a new singular clustering algorithm. As a case study, we examine the ant clustering algorithm (a swarm intelligence algorithm that is based on the natural phenomenon of brood sorting in some species of ants); highlight the strengths and weaknesses of the algorithm; and present a blend of algorithmic components from Tabu search into the algorithm to improve its exploration strategy and solution quality. Empirical results from applying the blended algorithm to clustering benchmark datasets show improved clustering validation measures for the proposed blended hybrid algorithm compared to other hybridization of the same underlying individual algorithms. Besides, the quality of clusters uncovered by this hybrid algorithm competes favorably with those uncovered using popular clustering algorithms such as DBSCAN and mean shift.

Research on Co-Channel Interference Cancellation for Underwater Acoustic MIMO Communications

Multiple-input–multiple-output (MIMO) communication systems utilize multiple transmitters to send different pieces of information in parallel. This offers a promising way to communicate at a high data rate over bandwidth-limited underwater acoustic channels. However, underwater acoustic MIMO communication not only suffers from serious inter-symbol interference, but also critical co-channel interference (CoI), both of which degrade the communication performance. In this paper, we propose a new framework for underwater acoustic MIMO communications. The proposed framework consists of a CoI-cancellation-based channel estimation method and channel-estimation-based decision feedback equalizer (CE-DFE) with CoI cancellation functionalities for underwater acoustic MIMO communication. We introduce a new channel estimation model that projects the received signal to a specific subspace where the interference is free; therefore, the CoI is cancelled. We also introduce a CE-DFE with CoI cancellation by appending some filters from traditional CE-DFE. In addition, the traditional direct adaptive decision feedback equalization (DA-DFE) method and the proposed method are compared in terms of communication performance and computational complexity. Finally, the sea trial experiment demonstrates the effectiveness and merits of the proposed method. The proposed method achieves a more than 1 dB of output SNR over traditional DA-DFE, and is less sensitive to parameters. The proposed method provides a new approach to the design of robust underwater acoustic MODEM.

High quality monocular depth estimation with parallel decoder

Monocular depth estimation aims to recover the depth information in three-dimensional (3D) space from a single image efficiently, but it is an ill-posed problem. Recently, Transformer-based architectures have achieved excellent accuracy in monocular depth estimation. However, due to the characteristics of Transformer, the model parameters are huge and the inference speed is slow. In traditional convolutional neural network–based architectures, many encoder-decoders perform serial fusion of the multi-scale features of each stage of the encoder and then output predictions. However, in these approaches it may be difficult to recover the spatial information lost by the encoder during pooling and convolution. To enhance this serial structure, we propose a structure from the decoder perspective, which first predicts global and local depth information in parallel and then fuses them. Results show that this structure is an effective improvement over traditional methods and has accuracy comparable with that of state-of-the-art methods in both indoor and outdoor scenes, but with fewer parameters and computations. Moreover, results of ablation studies verify the effectiveness of the proposed decoder.

Visual Information Computing and Processing Model Based on Artificial Neural Network.

This paper analyzes the parallel and serial information processing structure of visual system and proposes a visual information processing model with three layers: visual receptor layer, visual information conduction and relay layer, and information processing layer of visual information computing and processing area. Based on the analysis, abstraction, and simplification of the biological prototype of each layer in the visual system, a framework model of an artificial neural system corresponding to the visual system is proposed. An artificial neural network model is proposed to simulate the mechanism of visual attention. A network model is formed by introducing the saliency mask map as additional information on the benchmark network, and the selective enhancement operation is performed on the extracted features in different regions according to the mask map. The experimental results show that the visual computing processing network model can effectively improve the classification performance of the network when the appropriate saliency mask is used. The visual information computing and processing model network can work effectively for different data sets and different structures of the benchmark network, which is a universal network model. The complexity of visual information computing and processing model network is very small, and the improvement of network performance is not at the cost of increasing model complexity, but in the way of improving network efficiency. The performance of artificial neural network visual information computation and processing model is directly related to the performance of saliency map used as mask map.

An alternative method for the separation and analysis of acidic biomarkers from crude oil samples

This study aimed to develop and/or optimize a method of separation and analysis of neutral and acidic constituents from oil samples from Brazilian sedimentary basins and make a broad characterization of these samples through the respective molecular fossils. Three oil samples (A, B and C) of different origins and biodegradation were analyzed in this work. The neutral fractions were subjected to classic separation processes using gradient elution and analyzed by gas chromatography–mass spectrometry (GC–MS). The acidic fraction was isolated using KOH modified silica as stationary, adapted in SPE, and analyzed by GC–MS as silylated derivatives. The molecular parameters for B oil suggest a marine evaporitic depositional environment, with less continental influence, low siliciclastic sediments input, low oxygenation, hypersaline, stratified water column, and persistent photic zone euxinic. For the A and C oils, the parameters indicate a significant similarity between the samples, differing essentially only in the absence of acyclic n-alkanes and isoprenoids because of different stages of biodegradation. The presence of TPP and β-carotane in samples A and C allows us to classify them as mixed origin, with characteristics of lacustrine freshwater and saline depositional environments. Biomarkers associated with dinoflagellates of marine origin in the saturated fraction (dinosteranes) and the aromatic fraction (triaromatic dinosteranes) were detected in all three samples and suggest dinoflagellates were important primary producers during deposition time. 25-Norhopanes were detected only at low concentrations, and biomarker-based geochemical parameters indicate the biodegradation did not change the distribution of steranes and hopanes. The 25-Norhopanoic acids detection also agree with the detection of 25-norhopanes in the A and C oils, supports the same order of biodegradation, C > A ≫ B. The low abundance of tricyclic terpanoic acids in oil B acidic fraction agrees with the low abundance of tricyclic terpanes into the corresponding neutral fraction. Unlike hopanes, hopanoic acids distribution shows the ββ-isomers, biologically and thermally less stable configuration. Especially in oil B, series of 3-carboxysteranes, 3-carboxymethylsteranes were observed. In summary, acidic fractions data from three samples indicate that acidic biomarkers provide parallel information and complementary to that obtained from the hydrocarbon biomarkers. The extraction methods, derivatization and analysis of carboxylic acids from oil samples proved to be efficient, leading to quantitative recoveries and good yields.

Design of Cross-Platform Information Retrieval System of Library Based on Digital Twins.

In order to improve the library's ability of cross-platform information retrieval and data scheduling and distribution, a library cross-platform information retrieval system based on digital twin technology is designed. Using data warehouse decision support and data source structured query methods, the spectral characteristics of Library cross-platform information resources are extracted. Using the method of Hadoop data parallel loading, the library cross-platform operation data is divided into decision-making data, computing resource pool data, and Hadoop parallel loading data. A library cross-platform information digital twin parallel retrieval and information fusion feature matching model is established, and the retrieval channels are allocated through multiple complex and balanced task scheduling sequences. According to the queue configuration model of Library cross-platform information retrieval, the optimization design of Library cross-platform information retrieval system is realized. The simulation test results show that the designed system has good recall ability of cross-platform information retrieval data, and improves the utilization rate of cross-platform resources and the dynamic scheduling ability of online resources.

Вычислительный эксперимент – обезразмеривание уравнений, вычислительная устойчивость и тестирование программ

From the stages of the computational experiment, the stage of non-dimensionalization of the initial equation (system of equations) of the problem is considered - the replacement of its variables by the product of the corresponding dimensionless quantities by their units of measurement with subsequent transformations. Such a transition from a physical model to a mathematical (dimensionless) one makes it possible to obtain software implementations for research. A critical evaluation of its complexity is carried out and possible errors in the results are evaluated. At the same time, new versions of software are formed. Object-oriented programming tools and version control systems (for example, git) allow you to create versions of software tools adapted to different conditions of their use and for different types of users. Parallelization of work on versions is carried out. At the same time, for further software implementation, the set-theoretic language of formulas with partially recursive functions is effective. To implement versions with large amounts of calculations and data, high-performance computing systems based on software and hardware acceleration, parallel information processing and cloud architectures are used. As a rule, a difference model of the problem and iterative methods for solving it are constructed for a program version. Computational stability conditions are usually stipulated in modern instructions for standard program libraries. For new algorithms, it is necessary to analyze the stability of difference schemes based on the refinement of their spectral properties and the use of functional analysis methods. For storage and subsequent application of the results of computational experiments, it is advisable to use modern databases. As a kind of computational experiment, testing of alpha and beta versions of programs and their releases is also considered.

How to teach a blind person to hear colours? Multi-method training for a colour-to-sound sensory substitution device – design and evaluation

Sensory substitution is thought to be a promising non-invasive assistive technology for people with complete loss of sight because it provides inaccessible visual information via a preserved modality. However, Sensory Substitution Devices (SSDs) are still rarely used by visually impaired persons, possibly due to a lack of structured and supervised training that could be offered alongside these devices. Here, we developed and evaluated a training program that supports the usage of a recently developed colour-to-sound SSD – the Colorophone. Following our recently proposed theoretical model of SSD development, we propose that this training should help people with complete loss of sight to learn how to efficiently use the device by developing relationships between the components of the user-environment-technology system. We applied systematic case studies combined with a mixed-method approach to evaluate the efficacy of this SSD training program. Five blind users underwent ca. 22 h of training, divided into four main parts: identification of the users’ individual characteristics and adaptations; sensorimotor training with the device; semi-structured explorations with the device; and evaluation of the training. We demonstrated that this training allows users to successfully acquire a set of skills (i.e., master the sensorimotor contingencies required by the device, develop visual-like perceptual skills, as well as learn about colours) and progress along developmental trajectories (e.g., switch from serial to parallel information processing, recognize more complex colours, increase environment and task complexity). Importantly, we identified individual differences in learning strategies (i.e., sensorimotor vs. metacognitive strategy) that had an impact on the users’ training progress and required the training assistants (TAs) to apply different assistive strategies. Additionally, we described the crucial role of a (non-professional) training assistant in the training progress: this person facilitates the development of relationships between elements of the user-environment-technology system by supporting a metacognitive learning strategy, thereby reducing the risk of abandonment of the SSD. Our study shows the importance for SSD development of well-designed, tailored training, and it provides new insights into the process of SSD-related perceptual learning.

Research on Speech Coding Technology Based on Nonlinear System Theory in in AI-Based Disease Detection

In order to minimize the redundancy of speech decoding and realize the reliable transmission of speech data files, a speech coding method based on nonlinear system theory is proposed. Through pitch extraction, the residual harmonic spectrum condition of speech data information is calculated, and then combined with error correction coding technology, the speech feature analysis based on nonlinear system theory is realized. On this basis, the full search vector quantizer is set to suppress the noise parameters. According to the synthesis standard of speech coding information, the smooth application of speech coding technology based on nonlinear system theory is realized. The experimental results shows that, compared with MELP coding principle, the above coding technology can fully reduce the decoding redundancy level of voice information, realize the synchronous transmission of serial and parallel voice information, and better guarantee the reliable transmission environment of voice data files.

Artificial Tactile Sensing System with Photoelectric Output for High Accuracy Haptic Texture Recognition and Parallel Information Processing.

Developing multifunctional artificial sensory systems is an important task for constructing future artificial neural networks. A system with multisignal output capability is highly required by the rising demand for high-throughput data processing in the Internet of Things (IoT) society. Here, a novel dual-output artificial tactile sensing (DOATS) system with parallel output of photoelectric signals was proposed. Because of the ionic-electronic coupling mechanism in light-emitting synaptic (LES) devices in the DOATS system, modulating electric current and light emission can coexist through ion accumulation and electron-hole recombination. As a result, the DOATS system can realize the simulation of human tactile information, and the recognition of 16 kinds of fabrics was demonstrated with an accuracy rate of 94.1%. A photoelectric hybrid artificial neural network was proposed, which achieved efficient and accurate multitask operation. The DOATS system proposed in this work is promising for implementing photoelectric hybrid neural network and promoting the development of interactive artificial intelligence.