Deep Learning Neural Network Approach Research Articles

Deep learning neural network approach to thermal-wave imaging of damage in solids with application to diffusivity measurements of a green (unsintered) metal powder compact slab

A method is proposed to measure the thermal diffusivity of a green (unsintered) metal powder compact slab by combining numerical solution and deep learning. The thermal diffusion and thermal-wave equation is first discretized in space and time domains. Square wave excitation signals at different frequencies are considered, and the corresponding thermal wave responses are obtained. The space and frequency dependencies of the thermal-wave amplitude and phase are obtained by using lock-in thermography. Then, deep learning candidate neural network sets composed of spatial coordinates, excitation frequency, amplitude, and phase values are quickly and massively obtained for training the deep learning network. The validity of the deep learning network is verified by predicting the known thermal diffusivity, and the parameters and robustness of the deep learning network are discussed. Experimentally, a square wave modulated laser beam is used to illuminate one side of a green metal powder compact slab and the spatial distribution of amplitude and phase at three excitation frequencies is obtained. The predicted slab thermal diffusivity is extracted through the formed deep learning network, and the validity of the predicted value is verified by comparing with independent measurements.

Performance analysis and knowledge-based quality assurance of critical organ auto-segmentation for pediatric craniospinal irradiation

Craniospinal irradiation (CSI) is a vital therapeutic approach utilized for young patients suffering from central nervous system disorders such as medulloblastoma. The task of accurately outlining the treatment area is particularly time-consuming due to the presence of several sensitive organs at risk (OAR) that can be affected by radiation. This study aimed to assess two different methods for automating the segmentation process: an atlas technique and a deep learning neural network approach. Additionally, a novel method was devised to prospectively evaluate the accuracy of automated segmentation as a knowledge-based quality assurance (QA) tool. Involving a patient cohort of 100, ranging in ages from 2 to 25 years with a median age of 8, this study employed quantitative metrics centered around overlap and distance calculations to determine the most effective approach for practical clinical application. The contours generated by two distinct methods of atlas and neural network were compared to ground truth contours approved by a radiation oncologist, utilizing 13 distinct metrics. Furthermore, an innovative QA tool was conceptualized, designed for forthcoming cases based on the baseline dataset of 100 patient cases. The calculated metrics indicated that, in the majority of cases (60.58%), the neural network method demonstrated a notably higher alignment with the ground truth. Instances where no difference was observed accounted for 31.25%, while utilization of the atlas method represented 8.17%. The QA tool results showed that the two approaches achieved 100% agreement in 39.4% of instances for the atlas method and in 50.6% of instances for the neural network auto-segmentation. The results indicate that the neural network approach showcases superior performance, and its significantly closer physical alignment to ground truth contours in the majority of cases. The metrics derived from overlap and distance measurements have enabled clinicians to discern the optimal choice for practical clinical application.

An efficient deep neural network channel state estimator for OFDM wireless systems

AbstractChannel state estimation (CSE) is essential for orthogonal frequency division multiplexing (OFDM) wireless systems to deal with multipath channel fading. To attain a high data rate with the use of OFDM technology, an efficient CSE and accurate signal detection are required. The use of machine learning (ML) to improve channel estimates has attracted a lot of attention lately. This is because ML techniques are more adaptable than traditional model-based estimation techniques. The present study proposes a receiver for low-spectrum usage in OFDM wireless systems on Rayleigh fading channels using deep learning (DL) long short-term memory (LSTM). Before online deployment and data retrieval, the proposed DL LSTM estimator gathers channel state information from transmit/receive pairs using offline training. Based on the simulation results of a comparative study, the proposed estimator outperforms conventional channel estimation approaches like minimum mean square error and least squares in noisy and interfering wireless channels. Furthermore, the proposed estimator outperforms the DL bidirectional LSTM (BiLSTM)-based CSE model. In particular, the proposed CSE performs better than other examined estimators with a reduced number of pilots, no cycle prefixes, and no prior knowledge of channel statistics. Because the proposed estimator relies on a DL neural network approach, it holds promise for OFDM wireless communication systems.

Deep Learning for an Automated Image-Based Stem Cell Classification

Hematopoiesis is a process in which hematopoietic stem cells produce other mature blood cells in the bone marrow through cell proliferation and differentiation. The hematopoietic cells are cultured on a petri dish to form a different colony-forming unit (CFU). The idea is to identify the type of CFU produced by the stem cell. Several software has been developed to classify the CFU automatically. However, an automated identification or classification of CFU types has become the main challenge. Most of the current software has common drawbacks, such as the expensive operating cost and complex machines. The purpose of this study is to investigate several selected convolutional neural network (CNN) pre-trained models to overcome these constraints for automated CFU classification. Prior to CFU classification, the images are acquired from mouse stem cells and categorized into three types which are CFU-erythroid (E), CFU-granulocyte/macrophage (GM) and CFU-PreB. These images are then pre-processed before being fed into CNN pre-trained models. The models adopt a deep learning neural network approach to extract informative features from the CFU images Classification performance shows that the models integrated with the pre-processing module can classify the CFUs with high accuracies and shorter computational time with 96.33% on 61 minutes and 37 seconds, respectively. Hence, this work finding could be used as the baseline reference for further research.

A deep learning solution for crystallographic structure determination

The general de novo solution of the crystallographic phase problem is difficult and only possible under certain conditions. This paper develops an initial pathway to a deep learning neural network approach for the phase problem in protein crystallography, based on a synthetic dataset of small fragments derived from a large well curated subset of solved structures in the Protein Data Bank (PDB). In particular, electron-density estimates of simple artificial systems are produced directly from corresponding Patterson maps using a convolutional neural network architecture as a proof of concept.

A Deep-Learning Neural Network Approach for Secure Wireless Communication in the Surveillance of Electronic Health Records

The electronic health record (EHR) surveillance process relies on wireless security administered in application technology, such as the Internet of Things (IoT). Automated supervision with cutting-edge data analysis methods may be a viable strategy to enhance treatment in light of the increasing accessibility of medical narratives in the electronic health record. EHR analysis structured data structure code was used to obtain data on initial fatality risk, infection rate, and hazard ratio of death from EHRs for prediction of unexpected deaths. Patients utilizing EHRs in general must keep in mind the significance of security. With the rise of the IoT and sensor-based Healthcare 4.0, cyber-resilience has emerged as a need for the safekeeping of patient information across all connected devices. Security for access, amendment, and storage is cumulatively managed using the common paradigm. For improving the security of surveillance in the aforementioned services, this article introduces an endorsed joint security scheme (EJSS). This scheme recognizes the EHR utilization based on the aforementioned processes. For each process, different security measures are administered for sustainable security. Access control and storage modification require relative security administered using mutual key sharing between the accessing user and the EHR database. In this process, the learning identifies the variations in different processes for reducing adversarial interruption. The federated learning paradigm employed in this scheme identifies concurrent adversaries in the different processes initiated at the same time. Differentiating the adversaries under each process strengthens mutual authentication using individual attributes. Therefore, individual surveillance efficiency through log inspection and adversary detection is improved for heterogeneous and large-scale EHR databases.

Explainable Goal-driven Agents and Robots - A Comprehensive Review

Recent applications of autonomous agents and robots have brought attention to crucial trust-related challenges associated with the current generation of artificial intelligence (AI) systems. AI systems based on the connectionist deep learning neural network approach lack capabilities of explaining their decisions and actions to others, despite their great successes. Without symbolic interpretation capabilities, they are ‘black boxes’, which renders their choices or actions opaque, making it difficult to trust them in safety-critical applications. The recent stance on the explainability of AI systems has witnessed several approaches to eXplainable Artificial Intelligence (XAI) ; however, most of the studies have focused on data-driven XAI systems applied in computational sciences. Studies addressing the increasingly pervasive goal-driven agents and robots are sparse at this point in time. This paper reviews approaches on explainable goal-driven intelligent agents and robots, focusing on techniques for explaining and communicating agents’ perceptual functions (e.g., senses, vision) and cognitive reasoning (e.g., beliefs, desires, intentions, plans, and goals) with humans in the loop. The review highlights key strategies that emphasize transparency, understandability, and continual learning for explainability. Finally, the paper presents requirements for explainability and suggests a road map for the possible realization of effective goal-driven explainable agents and robots.

Forecasting office rents with ensemble models – the case for European real estate markets

PurposeCommercial real estate and office rental values, in particular, have long been the focus of research. Several forecasting frameworks for office rental values in multivariate and univariate fashions have been proposed. Recent developments in time series forecasting using machine learning and deep learning methods offer an opportunity to update traditional univariate forecasting frameworks.Design/methodology/approachWith the aim to extend research on univariate rent forecasting a hybrid methodology combining both ARIMA and a neural network model is proposed to exploit the unique strengths of both methods in linear and nonlinear modelling. N-BEATS, a deep learning algorithm that has demonstrated state-of-the-art forecasting performance in major forecasting competitions, are explained. With the ARIMA model, it is jointly applied to the office rental dataset to produce forecasts for four-quarters ahead.FindingsWhen the approach is applied to a dataset of 21 major European office cities, the results show that the ensemble model can be an effective approach to improve the prediction accuracy achieved by each of the models used separately.Practical implicationsReal estate forecasting is essential for assessing the value of managing portfolios and for evaluating investment strategies. The approach applied in this paper confirms the heterogeneity of real estate markets. The application of mixed modelling via linear and nonlinear methods decreases the uncertainty of abrupt changes in rents.Originality/valueTo the best of the authors' knowledge, no such application of a hybrid model updating classical statistical forecasting with a deep learning neural network approach in the field of commercial real estate rent forecasting has been undertaken.

Factors influencing the acceptance of telemedicine in the Philippines

The utilization of telemedicine has shown to increase access to healthcare services and bridge gaps in existing healthcare system issues. Despite these benefits, the technology is not currently widely accepted nor utilized in the Philippines, notwithstanding its worldwide recognition as a reliable alternative to face-to-face consultations. Issues regarding the slow utilization among Filipinos of telemedicine significantly affect its implementation. This study aimed to measure significant factors affecting the Philippines' behavioral adaptability of telemedicine application among Filipino users, through the Unified Theory of Acceptance and Use of Technology (UTAUT2) model. A total of 533 Filipinos voluntarily answered a self-administered questionnaire for 11 variables. Structural equation modeling and deep learning neural network approach showed that usage behavior (UB) was the key factor in determining user intention to use telemedicine, subsequently leading to perceived user adoption. In addition, performance expectancy, effort expectancy, hedonic motivation, social influence, and resistance to use was found to have a significant effect on UB. This study is the first study on Filipinos’ behavioral perception and acceptance of telemedicine. The findings of this study would be beneficial to researchers and the government—specifically the health sector in the Philippines for the implementation and promotion of telemedicine. The findings of this study could be adopted in technology acceptance in healthcare services among developing countries.

A deep learning neural network approach for predicting the factors influencing heavy-metal adsorption by clay minerals

AbstractThe treatment of water containing heavy metals has attracted increasing attention because the ingestion of such water poses risks to human health. Due to their relatively large specific surface areas and surface charges, clay minerals play a significant role in the adsorption of heavy metals in water. However, the major factors that influence the adsorption rates of clay minerals are not well understood, and thus methods to predict the sorption of heavy metals by clay minerals are lacking. A method that can identify the most appropriate clay minerals for removal of a given heavy metal, based on the predicted sorption of the clay minerals, is required. This paper presents a widely applicable deep learning neural network approach that yielded excellent predictions of the influence of the sorption ratio on the adsorption of heavy metals by clay minerals. The neural network model was based on datasets of heavy-metal parameters that are available generally. It yielded highly accurate predictions of the adsorption rate based on training data from the dataset and was able to account for a wide range of input parameters. A Pearson sensitivity analysis was used to determine the contributions of individual input parameters to the adsorption rates predicted by the neural network. This newly developed method can predict the major factors influencing heavy-metal adsorption rates. The model described here could be applied in a wide range of scenarios.

Investigating the quasi-liquid layer on ice surfaces: a comparison of order parameters.

Ice surfaces are characterized by pre-melted quasi-liquid layers (QLLs), which mediate both crystal growth processes and interactions with external agents. Understanding QLLs at the molecular level is necessary to unravel the mechanisms of ice crystal formation. Computational studies of the QLLs heavily rely on the accuracy of the methods employed for identifying the local molecular environment and arrangements, discriminating between solid-like and liquid-like water molecules. Here we compare the results obtained using different order parameters to characterize the QLLs on hexagonal ice (Ih) and cubic ice (Ic) model surfaces investigated with molecular dynamics (MD) simulations in a range of temperatures. For the classification task, in addition to the traditional Steinhardt order parameters in different flavours, we select an entropy fingerprint and a deep learning neural network approach (DeepIce), which are conceptually different methodologies. We find that all the analysis methods give qualitatively similar trends for the behaviours of the QLLs on ice surfaces with temperature, with some subtle differences in the classification sensitivity limited to the solid-liquid interface. The thickness of QLLs on the ice surface increases gradually as the temperature increases. The trends of the QLL size and of the values of the order parameters as a function of temperature for the different facets may be linked to surface growth rates which, in turn, affect crystal morphologies at lower vapour pressure. The choice of the order parameter can be therefore informed by computational convenience except in cases where a very accurate determination of the liquid-solid interface is important.

Channel state information estimation for 5G wireless communication systems: recurrent neural networks approach.

In this study, a deep learning bidirectional long short-term memory (BiLSTM) recurrent neural network-based channel state information estimator is proposed for 5G orthogonal frequency-division multiplexing systems. The proposed estimator is a pilot-dependent estimator and follows the online learning approach in the training phase and the offline approach in the practical implementation phase. The estimator does not deal with complete a priori certainty for channels’ statistics and attains superior performance in the presence of a limited number of pilots. A comparative study is conducted using three classification layers that use loss functions: mean absolute error, cross entropy function for kth mutually exclusive classes and sum of squared of the errors. The Adam, RMSProp, SGdm, and Adadelat optimisation algorithms are used to evaluate the performance of the proposed estimator using each classification layer. In terms of symbol error rate and accuracy metrics, the proposed estimator outperforms long short-term memory (LSTM) neural network-based channel state information, least squares and minimum mean square error estimators under different simulation conditions. The computational and training time complexities for deep learning BiLSTM- and LSTM-based estimators are provided. Given that the proposed estimator relies on the deep learning neural network approach, where it can analyse massive data, recognise statistical dependencies and characteristics, develop relationships between features and generalise the accrued knowledge for new datasets that it has not seen before, the approach is promising for any 5G and beyond communication system.

Key Physicochemical Properties Dictating Gastrointestinal Bioaccessibility of Microplastics-Associated Organic Xenobiotics: Insights from a Deep Learning Approach.

A potential risk from human uptake of microplastics is the release of plastics-associated xenobiotics, but the key physicochemical properties of microplastics controlling this process are elusive. Here, we show that the gastrointestinal bioaccessibility, assessed using an in vitro digestive model, of two model xenobiotics (pyrene, at 391-624 mg/kg, and 4-nonylphenol, at 3054-8117 mg/kg) bound to 18 microplastics (including pristine polystyrene, polyvinyl chloride, polyethylene terephthalate, polypropylene, thermoplastic polyurethane, and polyethylene, and two artificially aged samples of each polymer) covered wide ranges: 16.1-77.4% and 26.4-83.8%, respectively. Sorption/desorption experiments conducted in simulated gastric fluid indicated that structural rigidity of polymers was an important factor controlling bioaccessibility of the nonpolar, nonionic pyrene, likely by inducing physical entrapment of pyrene in porous domains, whereas polarity of microplastics controlled bioaccessibility of 4-nonylphenol, by regulating polar interactions. The changes of bioaccessibility induced by microplastics aging corroborated the important roles of polymeric structures and surface polarity in dictating sorption affinity and degree of desorption hysteresis, and consequently, gastrointestinal bioaccessibility. Variance-based global sensitivity analysis using a deep learning neural network approach further revealed that micropore volume was the most important microplastics property controlling bioaccessibility of pyrene, whereas the O/C ratio played a key role in dictating the bioaccessibility of 4-nonylphenol in the gastric tract.

Deep Learning Neural Network Approach for Predicting the Sorption of Ionizable and Polar Organic Pollutants to a Wide Range of Carbonaceous Materials.

Most contaminants of emerging concern are polar and/or ionizable organic compounds, whose removal from engineered and environmental systems is difficult. Carbonaceous sorbents include activated carbon, biochar, fullerenes, and carbon nanotubes, with applications such as drinking water filtration, wastewater treatment, and contaminant remediation. Tools for predicting sorption of many emerging contaminants to these sorbents are lacking because existing models were developed for neutral compounds. A method to select the appropriate sorbent for a given contaminant based on the ability to predict sorption is required by researchers and practitioners alike. Here, we present a widely applicable deep learning neural network approach that excellently predicted the conventionally used Freundlich isotherm fitting parameters log KF and n (R2 > 0.98 for log KF, and R2 > 0.91 for n). The neural network models are based on parameters generally available for carbonaceous sorbents and/or parameters freely available from online databases. A freely accessible graphical user interface is provided.

A typology of adopters and nonadopters of improved sorghum seeds in Tanzania: A deep learning neural network approach

Abstract For more than three decades, the direction of agricultural research and extension efforts have been toward developing improved seeds for agricultural transformation in Sub-Saharan Africa. Despite these efforts and substantial investment in physical and human capital, the adoption of improved seeds has remained marginal. One of the factors constraining adoption is limited choices among heterogeneous small-scale farmers often targeted by fit-for-all agricultural technologies. In this paper, we typify small-scale sorghum producers in Tanzania based on the socio-economic characteristics of farmers that include a propensity for adoption and intensity of adoption. The two variables are predicted using a deep learning neural network with back-propagation. The visualization of identified adopters and nonadopters groups is achieved using t-distributed stochastic neighbor embedding. Knowing the typology of farmers is a critical first step when the goal is scaling-up the adoption process through tailored advisory services. Results show that sorghum producers in Tanzania are heterogeneous, and there is a need for developing targeted agricultural innovations and public policies that serve specific groups of farmers. Since Tanzania agricultural policies are formulated at the national level, there must be room for adjustment by regional and district levels authorities to reflect local demand for services.

A novel deep learning neural network approach for predicting flash flood susceptibility: A case study at a high frequency tropical storm area

This research proposes and evaluates a new approach for flash flood susceptibility mapping based on Deep Learning Neural Network (DLNN)) algorithm, with a case study at a high-frequency tropical storm area in the northwest mountainous region of Vietnam. Accordingly, a DLNN structure with 192 neurons in 3 hidden layers was proposed to construct an inference model that predicts different levels of susceptibility to flash flood. The Rectified Linear Unit (ReLU) and the sigmoid were selected as the activate function and the transfer function, respectively, whereas the Adaptive moment estimation (Adam) was used to update and optimize the weights of the DLNN. A database for the study area, which includes factors of elevation, slope, curvature, aspect, stream density, NDVI, soil type, lithology, and rainfall, was established to train and validate the proposed model. Feature selection was carried out for these factors using the Information gain ratio. The results show that the DLNN attains a good prediction accuracy with Classification Accuracy Rate = 92.05%, Positive Predictive Value = 94.55% and Negative Predictive Value = 89.55%. Compared to benchmarks, Multilayer Perceptron Neural Network and Support Vector Machine, the DLNN performs better; therefore, it could be concluded that the proposed hybridization of GIS and deep learning can be a promising tool to assist the government authorities and involving parties in flash flood mitigation and land-use planning.

Computational Protein Design with Deep Learning Neural Networks

Computational protein design has a wide variety of applications. Despite its remarkable success, designing a protein for a given structure and function is still a challenging task. On the other hand, the number of solved protein structures is rapidly increasing while the number of unique protein folds has reached a steady number, suggesting more structural information is being accumulated on each fold. Deep learning neural network is a powerful method to learn such big data set and has shown superior performance in many machine learning fields. In this study, we applied the deep learning neural network approach to computational protein design for predicting the probability of 20 natural amino acids on each residue in a protein. A large set of protein structures was collected and a multi-layer neural network was constructed. A number of structural properties were extracted as input features and the best network achieved an accuracy of 38.3%. Using the network output as residue type restraints improves the average sequence identity in designing three natural proteins using Rosetta. Moreover, the predictions from our network show ~3% higher sequence identity than a previous method. Results from this study may benefit further development of computational protein design methods.