Parameters Of Computational Model Research Articles

Does the reliability of computational models truly improve with hierarchical modeling? Some recommendations and considerations for the assessment of model parameter reliability : Reliability of computational model parameters.

Computational modeling of behavior is increasingly being adopted as a standard methodology in psychology, cognitive neuroscience, and computational psychiatry. This approach involves estimating parameters in a computational (or cognitive) model that represents the computational processes of the underlying behavior. In this approach, the reliability of the parameter estimates is an important issue. The use of hierarchical (Bayesian) approaches, which place a prior on each model parameter of the individual participants, is thought to improve the reliability of the parameters. However, the characteristics of reliability in parameter estimates, especially when individual-level priors are assumed, as in hierarchical models, have not yet been fully discussed. Furthermore, the suitability of different reliability measures for assessing parameter reliability is not thoroughly understood. In this study, we conduct a systematic examination of these issues through theoretical analysis and numerical simulations, focusing specifically on reinforcement learning models. We note that the heterogeneity in the estimation precision of individual parameters, particularly with priors, can skew reliability measures toward individuals with higher precision. We further note that there are two factors that reduce reliability, namely estimation error and intersession variation in the true parameters, and we discuss how to evaluate these factors separately. Based on the considerations of this study, we present several recommendations and cautions for assessing the reliability of the model parameters.

Simulation of cancer progression in bone by a virtual thermal flux with a case study on lumbar vertebrae with multiple myeloma

BackgroundTwo main problems examining the mechanism of cancer progression in the tissues using the computational models are lack of enough knowledge on the effective factors for such events in vivo environments and lack of specific parameters in the available computational models to simulate such complicated reactions. MethodsIn this study, it was tried to simulate the progression of cancerous lesions in the bone tissues by an independent parameter from the anatomical and physiological characteristics of the tissues, so to degrade the orthotropic mechanical properties of the bone tissues, a virtual temperature was determined to be used by a well-known framework for simulation of damages in the composite materials. First, the reliability of the FE model to simulate hyperelastic response in the intervertebral discs (IVDs) and progressive failure in the bony components were verified by simulation of some In-Vitro tests, available in the literature. Then, the progression of the osteolytic damage was simulated in a clinical case with multiple myeloma in the lumbar vertebrae. ResultsThe FE model could simulate stress-shielding and diffusion of lesion in the posterior elements of the damaged vertebra which led to spinal stenosis. The load carrying shares associated with the anterior half and the posterior half of the examined vertebral body and the posterior elements were estimated equal to 41 %, 47 % and 12 %, respectively for the intact condition, that changed to 14 %, 16 % and 70 %, when lesion occupied one third of the vertebral body. ConclusionCorrelation of the FE results with the deformation shapes, observed in the MRIs for the clinical case study, indicated appropriateness of the procedure, proposed for simulation of the progressive osteolytic damage in the vertebral segments. The future studies may follow simulation of tumor growth for various metastatic tissues using the method, established here.

Efficient learning of Scale-Adaptive Nearly Affine Invariant Networks

Recent research has demonstrated the significance of incorporating invariance into neural networks. However, existing methods require direct sampling over the entire transformation set, notably computationally taxing for large groups like the affine group. In this study, we propose a more efficient approach by addressing the invariances of the subgroups within a larger group. For tackling affine invariance, we split it into the Euclidean group E(n) and uni-axial scaling group US(n), handling invariance individually. We employ an E(n)-invariant model for E(n)-invariance and average model outputs over data augmented from a US(n) distribution for US(n)-invariance. Our method maintains a favorable computational complexity of O(N2) in 2D and O(N4) in 3D scenarios, in contrast to the O(N6) (2D) and O(N12) (3D) complexities of averaged models. Crucially, the scale range for augmentation adapts during training to avoid excessive scale invariance. This is the first time nearly exact affine invariance is incorporated into neural networks without directly sampling the entire group. Extensive experiments unequivocally confirm its superiority, achieving new state-of-the-art results in affNIST and SIM2MNIST classifications while consuming less than 15% of inference time and fewer computational resources and model parameters compared to averaged models.

SalDA: DeepConvNet Greets Attention for Visual Saliency Prediction

Predicting salient regions in images requires the capture of contextual information in the scene. Conventional saliency models typically use the encoder-decoder architecture and multi-scale feature fusion for modeling contextual features, which, however, possess huge computational cost and model parameters. In this paper, we address the saliency prediction task by capturing long-range dependencies based on the self-attention mechanism. Self-attention has been widely used in image recognition or other classification tasks, but is still rarely being considered in regression based saliency prediction task. Inspired by non-local block, we propose a new SALiency prediction network in which DeepConvNet is integrated with the Attention mechanism, namely SalDA. Considering each feature map may capture different salient regions, our spatial attention module firstly adaptively aggregates the feature at each position by a weighted sum of the features at all positions within each independent channel. Meanwhile, in order to capture interdependence between channels, we also introduce a channel attention module to integrate different features among different channels. We combine these two attention modules into a multi-attention module to further improve the saliency map prediction for the network. We show the effectiveness of SalDA on the largest saliency prediction dataset SALICON. Compared to other state-of-the-art methods in this area, we can yield comparable saliency prediction performance, but with substantially less model parameters and shorter inference time.

Efficient Deep Spiking Multilayer Perceptrons With Multiplication-Free Inference.

Advancements in adapting deep convolution architectures for spiking neural networks (SNNs) have significantly enhanced image classification performance and reduced computational burdens. However, the inability of multiplication-free inference (MFI) to align with attention and transformer mechanisms, which are critical to superior performance on high-resolution vision tasks, imposes limitations on these gains. To address this, our research explores a new pathway, drawing inspiration from the progress made in multilayer perceptrons (MLPs). We propose an innovative spiking MLP architecture that uses batch normalization (BN) to retain MFI compatibility and introduce a spiking patch encoding (SPE) layer to enhance local feature extraction capabilities. As a result, we establish an efficient multistage spiking MLP network that blends effectively global receptive fields with local feature extraction for comprehensive spike-based computation. Without relying on pretraining or sophisticated SNN training techniques, our network secures a top-one accuracy of 66.39% on the ImageNet-1K dataset, surpassing the directly trained spiking ResNet-34 by 2.67%. Furthermore, we curtail computational costs, model parameters, and simulation steps. An expanded version of our network compares with the performance of the spiking VGG-16 network with a 71.64% top-one accuracy, all while operating with a model capacity 2.1 times smaller. Our findings highlight the potential of our deep SNN architecture in effectively integrating global and local learning abilities. Interestingly, the trained receptive field in our network mirrors the activity patterns of cortical cells.

Методический подход к исследованию магнитного взаимодействия тонкодисперсных частиц в водной суспензии методом компьютерного моделирования

The need to increase the efficiency of mining operations determines the relevance of studying the properties of fine mineral particles for their extraction and reduction of environmental pollution. The development of approaches to studying magnetic interaction of fine particles is of scientific and practical importance and is associated with the complexity of calculating the superposition of their fields when solving the problem of controlling the magnetic properties of a suspension. A computer model of the dynamics of magnetic particle interaction was developed with account for their aggregation under the impact of the magnetic dipole-dipole interaction and the destruction of aggregates during thermal (Brownian) motion in order to predict the magnetic properties of the suspension using the example of magnetite deposits of the Zaimandrovsky iron ore region. The calculation shows that the electrostatic and dispersion interactions do not have a significant effect on the interaction dynamics of micron and submicron particles of magnetite with account of the experimentally measured zeta potential and the Hamaker's constant. A procedure has been developed for calibrating a computer model of the interaction dynamics of magnetic particles using the temperature dependence of the coefficient of translational diffusion of magnetite particles and the concentration dependence of the magnetic susceptibility of a suspension. An array of calculated values of the diffusion coefficient of the model particles and the initial magnetic susceptibility of their system is formed in a wide range of computer model parameters. A procedure has been developed for linking the calculated and experimental data by varying the normalization parameters of the particle size, viscosity of the medium, and thermal energy in order to minimize the maximum discrepancy between the values. The necessity is established of taking into account the change in magnetic properties with a decrease in the size of magnetite particles during the calibration of the magnetic susceptibility of a model system. The developed methodological approach ensures good convergence of the calculated and experimental data and makes it possible to visualize the aggregation of the model particles as the result of dipole-dipole interactions. The developed computer model of the interaction dynamics of magnetic particles can be used to study the effect of an external magnetic field on the aggregation ability of fine magnetite particles in order to control their extraction in separation processes.

A comparative analysis of 2D and 3D experimental data for the identification of the parameters of computational models

Computational models are becoming an increasingly valuable tool in biomedical research. Their accuracy and effectiveness, however, rely on the identification of suitable parameters and on appropriate validation of the in-silico framework. Both these steps are highly dependent on the experimental model used as a reference to acquire the data. Selecting the most appropriate experimental framework thus becomes key, together with the analysis of the effect of combining results from different experimental models, a common practice often necessary due to limited data availability. In this work, the same in-silico model of ovarian cancer cell growth and metastasis, was calibrated with datasets acquired from traditional 2D monolayers, 3D cell culture models or a combination of the two. The comparison between the parameters sets obtained in the different conditions, together with the corresponding simulated behaviours, is presented. It provides a framework for the study of the effect of the different experimental models on the development of computational systems. This work also provides a set of general guidelines for the comparative testing and selection of experimental models and protocols to be used for parameter optimization in computational models.

Inverse identification of region-specific hyperelastic material parameters for human brain tissue

The identification of material parameters accurately describing the region-dependent mechanical behavior of human brain tissue is crucial for computational models used to assist, e.g., the development of safety equipment like helmets or the planning and execution of brain surgery. While the division of the human brain into different anatomical regions is well established, knowledge about regions with distinct mechanical properties remains limited. Here, we establish an inverse parameter identification scheme using a hyperelastic Ogden model and experimental data from multi-modal testing of tissue from 19 anatomical human brain regions to identify mechanically distinct regions and provide the corresponding material parameters. We assign the 19 anatomical regions to nine governing regions based on similar parameters and microstructures. Statistical analyses confirm differences between the regions and indicate that at least the corpus callosum and the corona radiata should be assigned different material parameters in computational models of the human brain. We provide a total of four parameter sets based on the two initial Poisson’s ratios of 0.45 and 0.49 as well as the pre- and unconditioned experimental responses, respectively. Our results highlight the close interrelation between the Poisson’s ratio and the remaining model parameters. The identified parameters will contribute to more precise computational models enabling spatially resolved predictions of the stress and strain states in human brains under complex mechanical loading conditions.

Bayesian model calibration for diblock copolymer thin film self-assembly using power spectrum of microscopy data and machine learning surrogate

Identifying parameters of computational models from experimental data, or model calibration, is fundamental for assessing and improving the predictability and reliability of computer simulations. In this work, we propose a method for Bayesian calibration of models that predict morphological patterns of diblock copolymer (Di-BCP) thin film self-assembly while accounting for various sources of uncertainties in pattern formation and data acquisition. This method extracts the azimuthally-averaged power spectrum (AAPS) of the top-down microscopy characterization of Di-BCP thin film patterns as summary statistics for Bayesian inference of model parameters via the pseudo-marginal method. We derive the analytical and approximate form of a conditional likelihood for the AAPS of image data. We demonstrate that AAPS-based image data reduction retains the mutual information, particularly on important length scales, between image data and model parameters while being relatively agnostic to the aleatoric uncertainties associated with the random long-range disorder of Di-BCP patterns. Additionally, we propose a phase-informed prior distribution for Bayesian model calibration. Furthermore, reducing image data to AAPS enables us to efficiently build surrogate models to accelerate the proposed Bayesian model calibration procedure. We present the formulation and training of two multi-layer perceptrons for approximating the parameter-to-spectrum map, which enables fast integrated likelihood evaluations. We validate the proposed Bayesian model calibration method through numerical examples, for which the neural network surrogate delivers a fivefold reduction of the number of model simulations performed for a single calibration task.

Modeling face recognition in the predictive coding framework: A combined computational modeling and functional imaging study

The learning of new facial identities and the recognition of familiar faces are crucial processes for social interactions. Recently, a combined computational modeling and functional magnetic resonance imaging (fMRI) study used predictive coding as a biologically plausible framework to model face identity learning and to relate specific model parameters with brain activity (Apps and Tsakiris, Nat Commun 4, 2698, 2013). On the one hand, it was shown that behavioral responses on a two-option face recognition task could be predicted by the level of contextual and facial familiarity in a computational model derived from predictive-coding principles. On the other hand, brain activity in specific brain regions was associated with these parameters. More specifically, brain activity in the superior temporal sulcus (STS) varied with contextual familiarity, whereas activity in the fusiform face area (FFA) covaried with the prediction error parameter that updated facial familiarity.Literature combining fMRI assessments and computational modeling in humans still needs to be expanded. Furthermore, prior results are largely not replicated. The present study was, therefore, specifically set up to replicate these previous findings. Our results support the original findings in two critical aspects. First, on a group level, the behavioral responses were modeled best by the same computational model reported by the original authors. Second, we showed that estimates of these model parameters covary with brain activity in specific, face-sensitive brain regions. Our results thus provide further evidence that the functional properties of the face perception network conform to central principles of predictive coding. However, our study yielded diverging findings on specific computational model parameters reflected in brain activity. On the one hand, we did not find any evidence of a computational involvement of the STS. On the other hand, our results showed that activity in the right FFA was associated with multiple computational model parameters. Our data do not provide evidence for functional segregation between particular face-sensitive brain regions, as previously proposed.

Sensitivity of TBL Wall-Pressure over the Flat Plate on Numerical Turbulence Model Parameter Variations

A two-fold sensitivity of the zero-pressure gradient (ZPG) turbulent boundary layer (TBL) wall-pressure spectrum to different RANS model parameters is investigated for a flat plate case, which is a close approximation to the aircraft fuselage or wing. The alteration in the mean square pressure fluctuations due choice of semi-empirical pressure model and the choice of computational model parameters like solver, near wall grid clustering, measuring location, and flow velocity are separately studied. The underlying effect of different TBL parameters in the said sensitivity has been studied while numerically replicating wind tunnel experiments and in-flight tests considering different RANS configurations. Initially, the best-predicting pressure spectrum models are selected by comparing them with available in-flight and wind tunnel test data. Subsequently, the accuracy of all the individual model parameters in predicting mean TBL flow quantities like wall shear stress, boundary layer thickness, displacement thickness, momentum thickness, etc., and eventually mean square pressure (MSP) is estimated. The sensitivity of the mean square pressure fluctuations value to the TBL flow quantities and the near-wall grid clustering is observed to be significant. In general, family of models is found to be best in terms of numerical convergence and closeness when compared to the experimental MSP values. family of models is suggested to be avoided while estimating MSP in flat plate TBL case

A computational musculoskeletal arm model for assessing muscle dysfunction in chronic obstructive pulmonary disease.

Computational models have been used extensively to assess diseases and disabilities effects on musculoskeletal system dysfunction. In the current study, we developed a two degree-of-freedom subject-specific second-order task-specific arm model for characterizing upper-extremity function (UEF) to assess muscle dysfunction due to chronic obstructive pulmonary disease (COPD). Older adults (65years or older) with and without COPD and healthy young control participants (18 to 30years) were recruited. First, we evaluated the musculoskeletal arm model using electromyography (EMG) data. Second, we compared the computational musculoskeletal arm model parameters along with EMG-based time lag and kinematics parameters (such as elbow angular velocity) between participants. The developed model showed strong cross-correlation with EMG data for biceps (0.905, 0.915) and moderate cross-correlation for triceps (0.717, 0.672) within both fast and normal pace tasks among older adults with COPD. We also showed that parameters obtained from the musculoskeletal model were significantly different between COPD and healthy participants. On average, higher effect sizes were achieved for parameters obtained from the musculoskeletal model, especially for co-contraction measures (effect size = 1.650 ± 0.606, p < 0.001), which was the only parameter that showed significant differences between all pairwise comparisons across the three groups. These findings suggest that studying the muscle performance and co-contraction, may provide better information regarding neuromuscular deficiencies compared to kinematics data. The presented model has potential for assessing functional capacity and studying longitudinal outcomes in COPD.

Magnetic Resonance Imaging-Derived Microvascular Perfusion Modeling to Assess Peripheral Artery Disease.

Background Computational fluid dynamics has shown good agreement with contrast-enhanced magnetic resonance imaging measurements in cardiovascular disease applications. We have developed a biomechanical model of microvascular perfusion using contrast-enhanced magnetic resonance imaging signal intensities derived from skeletal calf muscles to study peripheral artery disease (PAD). Methods and Results The computational microvascular model was used to study skeletal calf muscle perfusion in 56 individuals (36 patients with PAD, 20 matched controls). The recruited participants underwent contrast-enhanced magnetic resonance imaging and ankle-brachial index testing at rest and after 6-minute treadmill walking. We have determined associations of microvascular model parameters including the transfer rate constant, a measure of vascular leakiness; the interstitial permeability to fluid flow which reflects the permeability of the microvasculature; porosity, a measure of the fraction of the extracellular space; the outflow filtration coefficient; and the microvascular pressure with known markers of patients with PAD. Transfer rate constant, interstitial permeability to fluid flow, and microvascular pressure were higher, whereas porosity and outflow filtration coefficient were lower in patients with PAD than those in matched controls (all P values ≤0.014). In pooled analyses of all participants, the model parameters (transfer rate constant, interstitial permeability to fluid flow, porosity, outflow filtration coefficient, microvascular pressure) were significantly associated with the resting and exercise ankle-brachial indexes, claudication onset time, and peak walking time (all P values ≤0.013). Among patients with PAD, interstitial permeability to fluid flow, and microvascular pressure were higher, while porosity and outflow filtration coefficient were lower in treadmill noncompleters compared with treadmill completers (all P values ≤0.001). Conclusions Computational microvascular model parameters differed significantly between patients with PAD and matched controls. Thus, computational microvascular modeling could be of interest in studying lower extremity ischemia.

Multi-FusNet of Cross Channel Network for Image Super-Resolution

Image Super-resolution (SR) has gained considerable attention in artificial intelligence (AI) research and image-based applications. Recent deep learning-based SR models have demonstrated remarkable accuracy and perceptual quality in the resulting images. However, the computational cost and model parameters are the most challenging limitations in real-world applications. Additionally, designing an efficient and lightweight SR algorithm to improve the perceptual quality of the SR images is a critical issue. According to these considerations, we propose a Multi-FusNet of Cross Channel Network (MFCC) network by modeling a multipath residual network, named multi-RG, with cross-filtering fusion. Additionally, a pixel shuffling fusion technique is used to fuse low-level features into the up-sampled features of the multi-RG. The experimental results show the comparison of the proposed MFCC to the state-of-the-art SR models. The proposed method significantly reduces the number of network parameters (8.4 times compared to RCAN) while preserving the visual quality of the result and achieving the best PSNR value compared to the other state-of-the-art methods.

Bayesian inference with correction of model bias for Thermo-Hydro-Mechanical models of large concrete structures

Drying and creep are mainly driving the continuous strain evolution in time of aging large prestressed concrete structures, and may jeopardize the safety of their environment. Therefore, it is crucial to dispose of an accurate assessment of structures’ long-term strain level so as to evaluate their integrity. In the framework of probabilistic modeling of the delayed mechanical analysis of aging concrete structures, Bayesian inference allows to update uncertainties related to input parameters of computational models through the assimilation of noisy monitoring data. In this paper, a Bayesian inference methodology aiming at updating uncertain parameters of computational models for large concrete structures is proposed. This methodology combines a powerful surrogate modeling technique named PC-PCE (Principal Component Polynomial Chaos Expansions) with the so-called BUS (Bayesian Updating with Structural reliability methods) framework, in order to efficiently draw samples from posterior distributions for a sensibly reduced computational cost. In particular, the methodology is based on a framework to account for model uncertainties and biases, which are usually disregarded in the existing literature related to large concrete structures, and it enables correction predictions through Bayesian inference. The proposed approach is illustrated through an application to a large aging prestressed concrete structure, namely a 1:3 scale mock-up of a Nuclear Containment Building. In this context, a thermo-hydro-mechanical computational model with uncertain parameters is adopted to model the time evolution of strains of the structure. Results emphasize that the proposed approach performs Bayesian updating for a reduced computational cost. The proposed Bayesian inference approach also enables the identification of model biases, and correction of strain predictions.

Pacing Dynamics Determines the Arrhythmogenic Mechanism of the CPVT2-Causing CASQ2G112+5X Mutation in a Guinea Pig Ventricular Myocyte Computational Model.

Calsequestrin Type 2 (CASQ2) is a high-capacity, low-affinity, Ca2+-binding protein expressed in the sarcoplasmic reticulum (SR) of the cardiac myocyte. Mutations in CASQ2 have been linked to the arrhythmia catecholaminergic polymorphic ventricular tachycardia (CPVT2) that occurs with acute emotional stress or exercise can result in sudden cardiac death (SCD). CASQ2G112+5X is a 16 bp (339-354) deletion CASQ2 mutation that prevents the protein expression due to premature stop codon. Understanding the subcellular mechanisms of CPVT2 is experimentally challenging because the occurrence of arrhythmia is rare. To obtain an insight into the characteristics of this rare disease, simulation studies using a local control stochastic computational model of the Guinea pig ventricular myocyte investigated how the mutant CASQ2s may be responsible for the development of an arrhythmogenic episode under the condition of β-adrenergic stimulation or in the slowing of heart rate afterward once β-adrenergic stimulation ceases. Adjustment of the computational model parameters based upon recent experiments explore the functional changes caused by the CASQ2 mutation. In the simulation studies under rapid pacing (6 Hz), electromechanically concordant cellular alternans appeared under β-adrenergic stimulation in the CPVT mutant but not in the wild-type nor in the non-β-stimulated mutant. Similarly, the simulations of accelerating pacing from slow to rapid and back to the slow pacing did not display alternans but did generate early afterdepolarizations (EADs) during the period of second slow pacing subsequent acceleration of rapid pacing.

Optimization framework for patient-specific modeling under uncertainty.

Estimating a patient-specific computational model's parameters relies on data that is often unreliable and ill-suited for a deterministic approach. We develop an optimization-based uncertainty quantification framework for probabilistic model tuning that discovers model inputs distributions that generate target output distributions. Probabilistic sampling is performed using a surrogate model for computational efficiency, and a general distribution parameterization is used to describe each input. The approach is tested on seven patient-specific modeling examples using CircAdapt, a cardiovascular circulatory model. Six examples are synthetic, aiming to match the output distributions generated using known reference input data distributions, while the seventh example uses real-world patient data for the output distributions. Our results demonstrate the accurate reproduction of the target output distributions, with a correct recreation of the reference inputs for the six synthetic examples. Our proposed approach is suitable for determining the parameter distributions of patient-specific models with uncertain data and can be used to gain insights into the sensitivity of the model parameters to the measured data.

Multi-Source Domain Transfer Discriminative Dictionary Learning Modeling for Electroencephalogram-Based Emotion Recognition

Cognitive computing is dedicated to researching a computing principle and method that can simulate the intelligence ability of human brain. Human emotion is the basic component of human cognitive activities. Electroencephalogram (EEG) computer signals obtained from a brain computer interface are difficult to conceal, and using machine learning methods to analyze EEG emotion is a hot topic in artificial intelligence. However, the EEG signal is non-stationary, making it difficult to select sufficient data from the same person to train a classifier for a subject. To promote the performance of emotion recognition methods, a multi-source domain transfer discriminative dictionary learning modeling (MDTDDL) is proposed in this study. The method integrates transfer learning and dictionary learning in a learning model, including the concepts of subspace learning, manifold smoothness, margin-based discriminant embedding, and large margin. The domain-specific transformation matrix projects EEG signals from various domains into the transfer subspace. The domain-invariant dictionary can find potential connections between multiple source domains and target domain. The manifold smoothness and margin-based discriminant embedding term further improve the model’s learning ability. The alternating optimization technique is used in model solving to efficiently compute model parameters. Experiments on the SEED and DEAP datasets demonstrate the effectiveness of MDTDDL.

First coarse, fine afterward: A lightweight two-stage complex approach for monaural speech enhancement

Deep neural network-based speech enhancement systems have achieved promising results. However, the state-of-the-art (SOTA) models usually have too many parameters and require too much computational work to be used on devices for practical applications. In this paper, we propose a novel lightweight complex spectral mask-based neural network with a two-stage pipeline for monaural speech enhancement. The network utilizes the idea of decoupling a primary problem into several simple sub-problems, which reduces the computational burden and model parameters. Specifically, the network contains two mask-based sub-networks, i.e., CoarseNet, and FineNet, implemented in the complex domain to improve the enhancement performances progressively. The CoarseNet takes the coarse-grained compact features as input and estimates the corresponding full-band complex mask. The FineNet focuses on further removing residual noises in the low-frequency components of CoarseNet output by predicting a fine-grained mask. The transforms between coarse- and fine-scale are based on a novel learnable complex-valued rectangular bandwidth (LCRB) filter bank. Furthermore, we also propose a lightweight and general complex-valued attention mechanism to improve the modeling capability of convolutional encoder/decoder of the network and uses cross-stage skip connections (CSSC) between sub-networks to facilitate information flowing between sub-networks. Extensive experiments on two standard corpora demonstrate that our proposed approach achieves better performances over previous SOTA systems under various conditions while maintaining relatively small model sizes and low computational complexity.

The Effects of Tinnitus in Probabilistic Learning Tasks: Protocol for an Ecological Momentary Assessment Study.

Chronic tinnitus is an increasing worldwide health concern, causing a significant burden to the health care system each year. The COVID-19 pandemic has seen a further increase in reported cases. For people with tinnitus, symptoms are exacerbated because of social isolation and the elevated levels of anxiety and depression caused by quarantines and lockdowns. Although it has been reported that patients with tinnitus can experience changes in cognitive capabilities, changes in adaptive learning via decision-making tasks for people with tinnitus have not yet been investigated. In this study, we aim to assess state- and trait-related impairments in adaptive learning ability on probabilistic learning tasks among people with tinnitus. Given that performance in such tasks can be quantified through computational modeling methods using a small set of neural-informed model parameters, such approaches are promising in terms of the assessment of tinnitus severity. We will first examine baseline differences in the characterization of decision-making under uncertainty between healthy individuals and people with tinnitus in terms of differences in the parameters of computational models in a cross-sectional experiment. We will also investigate whether these computational markers, which capture characteristics of decision-making, can be used to understand the cognitive impact of tinnitus symptom fluctuations through a longitudinal experimental design. We have developed a mobile app, AthenaCX, to deliver e-consent and baseline tinnitus and psychological assessments as well as regular ecological momentary assessments (EMAs) of perceived tinnitus loudness and a web-based aversive version of a probabilistic decision-making task, which can be triggered based on the participants' responses to the EMA surveys. Computational models will be developed to fit participants' choice data in the task, and cognitive parameters will be estimated to characterize participants' current ability to adapt learning to the change of the simulated environment at each session when the task is triggered. Linear regression analysis will be conducted to evaluate the impacts of baseline tinnitus severity on adapting decision-making performance. Repeated measures linear regression analysis will be used to examine model-derived parameters of decision-making in measuring real-time perceived tinnitus loudness fluctuations. Ethics approval was received in December 2020 from Dublin City University (DCUREC/2021/070). The implementation of the experiments, including both the surveys and the web-based decision-making task, has been prepared. Recruitment flyers have been shared with audiologists, and a video instruction has been created to illustrate to the participants how to participate in the experiment. We expect to finish data collection over 12 months and complete data analysis 6 months after this. The results are expected to be published in December 2023. We believe that EMA with context-aware triggering can facilitate a deeper understanding of the effects of tinnitus symptom severity upon decision-making processes as measured outside of the laboratory. PRR1-10.2196/36583.