Benchmark Accuracy Research Articles

Three-Dimensional Convolutional Neural Networks Utilizing Molecular Topological Features for Accurate Atomization Energy Predictions.

Deep learning methods provide a novel way to establish a correlation between two quantities. In this context, computer vision techniques such as three-dimensional (3D)-convolutional neural networks become a natural choice to associate a molecular property with its structure due to the inherent 3D nature of a molecule. However, traditional 3D input data structures are intrinsically sparse in nature, which tend to induce instabilities during the learning process, which in turn may lead to underfitted results. To address this deficiency, in this project, we propose to use quantum-chemically derived molecular topological features, namely, localized orbital locator and electron localization function, as molecular descriptors, which provide a relatively denser input representation in a 3D space. Such topological features provide a detailed picture of the atomic and electronic configuration and interatomic interactions in the molecule and hence are ideal for predicting properties that are highly dependent on the physical or electronic structure of the molecule. Herein, we demonstrate the efficacy of our proposed model by applying it to the task of predicting atomization energies for the QM9-G4MP2 data set, which contains ∼134k molecules. Furthermore, we incorporated the Δ-machine learning approach into our model, which enabled us to reach beyond benchmark accuracy levels (∼1.0 kJ mol-1). As a result, we consistently obtain impressive mean absolute errors of the order 0.1 kcal mol-1 (∼0.42 kJ mol-1) versus the G4(MP2) theory using relatively modest models, which could potentially be improved further in a systematic manner using additional compute resources.

Toward Automated Feature Extraction for Deep Learning Classification of Electrocardiogram Signals

Many recent studies have focused on the automatic classification of electrocardiogram (ECG) signals using deep learning (DL) methods. Most rely on existing complex DL methods, such as transfer learning or providing the models with carefully designed extracted features based on domain knowledge. A common assumption is that the deeper and more complex the DL model is, the better it learns. In this study, we propose two different DL models for automatic feature extraction from ECG signals for classification tasks: A CNN-LSTM hybrid model and an attention/transformer-based model with wavelet transform for the dimensional embedding. Both of the models extract the features from time series at the initial layers of the neural networks and can obtain performance at least equal to, if not greater than, many contemporary deep neural networks. To validate our hypothesis, we used three publicly available data-sets to evaluate the proposed models. Our model achieved a benchmark accuracy of 99.92% for fall detection and 99.93% for the PTB database for myocardial infarction versus normal heartbeat classification.

A Modified Jellyfish Search Optimizer with Opposition Based Learning and Biased Passive Swarm Motion

Jellyfish Search Optimizer (JSO) is one of the latest nature inspired optimization algorithms. This paper aims to improve the convergence speed of the algorithm. For the purpose, it identifies two modifications to form a proposed variant. First, it proposes improvement of initial population using Opposition based Learning (OBL). Then it introduces a probability-based replacement of passive swarm motion into moves biased towards the global best. OBL enables the algorithm to start with an improved set of population. Biased moves towards global best improve the exploitation capability of the algorithm. The proposed variant has been tested over 30 benchmark functions and the real world problem of 10-bar truss structure design optimization. The proposed variant has also been compared with other algorithms from the literature for the 10-bar truss structure design. The results show that the proposed variant provides fast convergence for benchmark functions and accuracy better than many algorithms for truss structure design.

KL-MOB: automated COVID-19 recognition using a novel approach based on image enhancement and a modified MobileNet CNN.

The emergence of the novel coronavirus pneumonia (COVID-19) pandemic at the end of 2019 led to worldwide chaos. However, the world breathed a sigh of relief when a few countries announced the development of a vaccine and gradually began to distribute it. Nevertheless, the emergence of another wave of this pandemic returned us to the starting point. At present, early detection of infected people is the paramount concern of both specialists and health researchers. This paper proposes a method to detect infected patients through chest x-ray images by using the large dataset available online for COVID-19 (COVIDx), which consists of 2128 X-ray images of COVID-19 cases, 8,066 normal cases, and 5,575 cases of pneumonia. A hybrid algorithm is applied to improve image quality before undertaking neural network training. This algorithm combines two different noise-reduction filters in the image, followed by a contrast enhancement algorithm. To detect COVID-19, we propose a novel convolution neural network (CNN) architecture called KL-MOB (COVID-19 detection network based on the MobileNet structure). The performance of KL-MOB is boosted by adding the Kullback–Leibler (KL) divergence loss function when trained from scratch. The KL divergence loss function is adopted for content-based image retrieval and fine-grained classification to improve the quality of image representation. The results are impressive: the overall benchmark accuracy, sensitivity, specificity, and precision are 98.7%, 98.32%, 98.82% and 98.37%, respectively. These promising results should help other researchers develop innovative methods to aid specialists. The tremendous potential of the method proposed herein can also be used to detect COVID-19 quickly and safely in patients throughout the world.

Leveraging Neuroimaging Tools to Assess Precision and Accuracy in an Alzheimer's Disease Neuropathologic Sampling Protocol.

IntroductionThe study of Alzheimer’s disease investigates topographic patterns of degeneration in the context of connected networks comprised of functionally distinct domains using increasingly sophisticated molecular techniques. Therefore, obtaining high precision and accuracy of neuropathologic tissue sampling will enhance the reliability of molecular studies and contribute to the understanding of Alzheimer’s disease pathology. Neuroimaging tools can help assess these aspects of current sampling protocols as well as contribute directly to their improvement.MethodsUsing a virtual sampling method on magnetic resonance images (MRIs) from 35 participants (21 women), we compared the precision and accuracy of traditional neuropathologic vs. neuroimaging-guided sampling. The impact of the resulting differences was assessed by evaluating the functional connectivity pattern of regions selected by each approach.ResultsVirtual sampling using the traditional neuropathologic approach had low neuroanatomical precision and accuracy for all cortical regions tested. Neuroimaging-guided strategies narrowed these gaps. Discrepancies in the location of traditional and neuroimaging-guided samples corresponded to differences in fMRI measures of functional connectivity.DiscussionIntegrating neuroimaging tools with the neuropathologic assessment will improve neuropathologic-neuroimaging correlations by helping to ensure specific functional domains are accurately sampled for quantitative molecular neuropathologic applications. Our neuroimaging-based simulation of current sampling practices provides a benchmark of precision and accuracy against which to measure improvements when using novel tissue sampling approaches. Our results suggest that relying on gross landmarks alone to select samples at autopsy leads to significant variability, even when sampled by the same neuropathologist. Further, this exercise highlights how sampling precision could be enhanced if neuroimaging were integrated with the standard neuropathologic assessment. More accurate targeting and improved biological homogeneity of sampled brain tissue will facilitate the interpretation of neuropathological analyses in AD and the downstream research applications of brain tissue from biorepositories.

An assessment of high-resolution wind speeds downscaled with the Weather Research and Forecasting Model for coastal areas in Ghana

Ghana produces over 50% of its electrical energy demands from fossil-fuelled thermal plants. To increase the proportion of renewable energy in the national energy generation, a Renewable Energy Master Plan (REMP) which seeks, among others, to shift the country's national energy generation capacity towards more renewable energy sources has been developed. The REMP noted that inadequate data on renewable energy sources such as wind is one of the challenges to achieving this target. In this regard, this paper assessed the open-source Weather Research and Forecasting Model, as a tool for generating wind resource data. The WRF model is often used to downscale meteorological datasets for wind resources assessments. However, due to diverse model options, performance assessments are required to establish the accuracy and suitability of a model configuration for an application in an area. This paper assessed the performance of a Weather Research and Forecasting Model configuration that is based on previous verification studies. In evaluation, data accuracy benchmarks were generally met by the downscaled wind data. A wind map that was generated was observed to be generally accurate and better than the previous 2001 wind map for Ghana. It is presumed that the configuration is suitable for wind mapping activities for the coastal areas in Ghana, and probably neighbouring countries. However, for downscaling time-series data, further studies are recommended.

PERFECTION OF PAPER TRANSPORTATION SETTINGS TO GET PROPERTIES ON MACHINE KOMORI LITHRONE G40

The perfection of the paper transportation settings raised in the Final Project work aims to achieve print accuracy, the benchmark for print accuracy is the achievement of perfection of print ink stacks on the printed results guided by the register point / print mark. Overview of the need for paper transport settings so that the printing process can run smoothly. In making this final project, the author uses a descriptive method, namely a method that explains and explains about paper transportation. The formulation of the problem that is taken in this final project is how to set up paper transportation, problems that arise during paper transportation, and solutions that can be done to overcome these problems. In this paper, the authors use literature review, field observations, and interviews related to data. From the results of these observations, it can be shown that the problems that occur at the time are caused by paper transportation settings that are not in accordance with standard work procedures. Suggestions put forward by the authors need to apply standardization of settings and provision of training for each operator who worksKeywords— Paper transport, print accuracy, tuning

Single-cell conventional pap smear image classification using pre-trained deep neural network architectures

BackgroundAutomating cytology-based cervical cancer screening could alleviate the shortage of skilled pathologists in developing countries. Up until now, computer vision experts have attempted numerous semi and fully automated approaches to address the need. Yet, these days, leveraging the astonishing accuracy and reproducibility of deep neural networks has become common among computer vision experts. In this regard, the purpose of this study is to classify single-cell Pap smear (cytology) images using pre-trained deep convolutional neural network (DCNN) image classifiers. We have fine-tuned the top ten pre-trained DCNN image classifiers and evaluated them using five class single-cell Pap smear images from SIPaKMeD dataset. The pre-trained DCNN image classifiers were selected from Keras Applications based on their top 1% accuracy.ResultsOur experimental result demonstrated that from the selected top-ten pre-trained DCNN image classifiers DenseNet169 outperformed with an average accuracy, precision, recall, and F1-score of 0.990, 0.974, 0.974, and 0.974, respectively. Moreover, it dashed the benchmark accuracy proposed by the creators of the dataset with 3.70%.ConclusionsEven though the size of DenseNet169 is small compared to the experimented pre-trained DCNN image classifiers, yet, it is not suitable for mobile or edge devices. Further experimentation with mobile or small-size DCNN image classifiers is required to extend the applicability of the models in real-world demands. In addition, since all experiments used the SIPaKMeD dataset, additional experiments will be needed using new datasets to enhance the generalizability of the models.

Interactions between large molecules pose a puzzle for reference quantum mechanical methods

Quantum-mechanical methods are used for understanding molecular interactions throughout the natural sciences. Quantum diffusion Monte Carlo (DMC) and coupled cluster with single, double, and perturbative triple excitations [CCSD(T)] are state-of-the-art trusted wavefunction methods that have been shown to yield accurate interaction energies for small organic molecules. These methods provide valuable reference information for widely-used semi-empirical and machine learning potentials, especially where experimental information is scarce. However, agreement for systems beyond small molecules is a crucial remaining milestone for cementing the benchmark accuracy of these methods. We show that CCSD(T) and DMC interaction energies are not consistent for a set of polarizable supramolecules. Whilst there is agreement for some of the complexes, in a few key systems disagreements of up to 8 kcal mol−1 remain. These findings thus indicate that more caution is required when aiming at reproducible non-covalent interactions between extended molecules.

DarkASDNet: Classification of ASD on Functional MRI Using Deep Neural Network

Non-invasive whole-brain scans aid the diagnosis of neuropsychiatric disorder diseases such as autism, dementia, and brain cancer. The assessable analysis for autism spectrum disorders (ASD) is rationally challenging due to the limitations of publicly available datasets. For diagnostic or prognostic tools, functional Magnetic Resonance Imaging (fMRI) exposed affirmation to the biomarkers in neuroimaging research because of fMRI pickup inherent connectivity between the brain and regions. There are profound studies in ASD with introducing machine learning or deep learning methods that have manifested advanced steps for ASD predictions based on fMRI data. However, utmost antecedent models have an inadequacy in their capacity to manipulate performance metrics such as accuracy, precision, recall, and F1-score. To overcome these problems, we proposed an avant-garde DarkASDNet, which has the competence to extract features from a lower level to a higher level and bring out promising results. In this work, we considered 3D fMRI data to predict binary classification between ASD and typical control (TC). Firstly, we pre-processed the 3D fMRI data by adopting proper slice time correction and normalization. Then, we introduced a novel DarkASDNet which surpassed the benchmark accuracy for the classification of ASD. Our model's outcomes unveil that our proposed method established state-of-the-art accuracy of 94.70% to classify ASD vs. TC in ABIDE-I, NYU dataset. Finally, we contemplated our model by performing evaluation metrics including precision, recall, F1-score, ROC curve, and AUC score, and legitimize by distinguishing with recent literature descriptions to vindicate our outcomes. The proposed DarkASDNet architecture provides a novel benchmark approach for ASD classification using fMRI processed data.

Mobile sonouroflowmetry using voiding sound and volume

Uroflowmetry (UF) is a common clinic-based non-invasive test to diagnose Lower Urinary Tract Dysfunction (LUTD). Accurate home-based uroflowmetry methods are needed to conveniently conduct repeated uroflowmetries when patients are physiologically ready to urinate. To this end, we propose and evaluate a novel mobile sonouroflowmetry (SUF) method that estimates the urinary flow rate from a sound signal recorded using a mobile phone. By linearly mapping the total sound energy to the total voided volume, the sound energy curve is transformed to a flow rate curve allowing the estimation of the flow rate over time. An evaluation using data from 44 healthy young men showed high similarity between the UF and SUF flow rates with a mixed-effects model correlation coefficient of 0.993 and a mean root mean square error of 2.37 ml/s. Maximum flow rates were estimated with an average absolute error of 2.41 ml/s. Future work on mobile uroflowmetry can use these results as an initial benchmark for flow rate estimation accuracy.

Fighting Deepfakes Using Body Language Analysis

Recent improvements in deepfake creation have made deepfake videos more realistic. Moreover, open-source software has made deepfake creation more accessible, which reduces the barrier to entry for deepfake creation. This could pose a threat to the people’s privacy. There is a potential danger if the deepfake creation techniques are used by people with an ulterior motive to produce deepfake videos of world leaders to disrupt the order of countries and the world. Therefore, research into the automatic detection of deepfaked media is essential for public security. In this work, we propose a deepfake detection method using upper body language analysis. Specifically, a many-to-one LSTM network was designed and trained as a classification model for deepfake detection. Different models were trained by varying the hyperparameters to build a final model with benchmark accuracy. We achieved 94.39% accuracy on the deepfake test set. The experimental results showed that upper body language can effectively detect deepfakes.

Machine learning for patient risk stratification: standing on, or looking over, the shoulders of clinicians?

Machine learning can help clinicians to make individualized patient predictions only if researchers demonstrate models that contribute novel insights, rather than learning the most likely next step in a set of actions a clinician will take. We trained deep learning models using only clinician-initiated, administrative data for 42.9 million admissions using three subsets of data: demographic data only, demographic data and information available at admission, and the previous data plus charges recorded during the first day of admission. Models trained on charges during the first day of admission achieve performance close to published full EMR-based benchmarks for inpatient outcomes: inhospital mortality (0.89 AUC), prolonged length of stay (0.82 AUC), and 30-day readmission rate (0.71 AUC). Similar performance between models trained with only clinician-initiated data and those trained with full EMR data purporting to include information about patient state and physiology should raise concern in the deployment of these models. Furthermore, these models exhibited significant declines in performance when evaluated over only myocardial infarction (MI) patients relative to models trained over MI patients alone, highlighting the importance of physician diagnosis in the prognostic performance of these models. These results provide a benchmark for predictive accuracy trained only on prior clinical actions and indicate that models with similar performance may derive their signal by looking over clinician’s shoulders—using clinical behavior as the expression of preexisting intuition and suspicion to generate a prediction. For models to guide clinicians in individual decisions, performance exceeding these benchmarks is necessary.

Low-Scaling GW with Benchmark Accuracy and Application to Phosphorene Nanosheets

GW is an accurate method for computing electron addition and removal energies of molecules and solids. In a conventional GW implementation, however, its computational cost is O(N4) in the system size N, which prohibits its application to many systems of interest. We present a low-scaling GW algorithm with notably improved accuracy compared to our previous algorithm [J. Phys. Chem. Lett. 2018, 9, 306-312]. This is demonstrated for frontier orbitals using the GW100 benchmark set, for which our algorithm yields a mean absolute deviation of only 6 meV with respect to canonical implementations. We show that also excitations of deep valence, semicore, and unbound states match conventional schemes within 0.1 eV. The high accuracy is achieved by using minimax grids with 30 grid points and the resolution of the identity with the truncated Coulomb metric. We apply the low-scaling GW algorithm with improved accuracy to phosphorene nanosheets of increasing size. We find that their fundamental gap is strongly size-dependent varying from 4.0 eV (1.8 nm × 1.3 nm, 88 atoms) to 2.4 eV (6.9 nm × 4.8 nm, 990 atoms) at the evGW0@PBE level.

Colorectal Histology Tumor Detection Using Ensemble Deep Neural Network

With a mortality rate of approximately 33.33%, Colorectal cancer serves as the second most prevalent malignant tumor type in the world. AI-guided clinical care/tool can help in reducing health disparities, specifically in resource-constrained regions. In this paper, using multi-class tissue features, we proposed an Ensemble Deep Neural Network to Tumor in Colorectal Histology images. On two different publicly available datasets: NCT-CRC-HE-100K (107,180 images) and Colorectal Histology (5000 images), we achieved accuracies of 96.16% and 92.83%, respectively. When datasets are combined, it provided a benchmark accuracy of 99.13%. We efficiently used resourced data, thereby achieving results that outperformed the state-of-the-art works.

Bangla language textual image description by hybrid neural network model

Automatic image captioning task in different language is a challenging task which has not been well investigated yet due to the lack of dataset and effective models. It also requires good understanding of scene and contextual embedding for robust semantic interpretation of images for natural language image descriptor. To generate image descriptor in Bangla, we created a new Bangla dataset of images paired with target language label, named as Bangla Natural Language Image to Text (BNLIT) dataset. To deal with the image understanding, we propose a hybrid encoder-decoder model based on encoder-decoder architecture and the model is evaluated on our newly created dataset. This proposed approach achieves significance performance improvement on task of semantic retrieval of images. Our hybrid model uses the Convolutional Neural<br />Network as an encoder whereas the Bidirectional Long Short Term Memory is used for the sentence representation that decreases the computational complexities without trading off the exactness of the descriptor. The model yielded benchmark accuracy in recovering Bangla natural language and we also conducted a thorough numerical analysis of the model performance on the BNLIT dataset.

The Stratic Defuzzifier for discretised general type-2 fuzzy sets

Stratification is a feature of the type-reduced set of the general type-2 fuzzy set, from which a new technique for general type-2 defuzzification, Stratic Defuzzification, may be derived. Existing defuzzification strategies are summarised. The stratified structure is described, after which the Stratic Defuzzifier is presented and contrasted experimentally for accuracy and efficiency with both the Exhaustive Method of Defuzzification (to benchmark accuracy) and the α-Planes/Karnik–Mendel Iterative Procedure strategy, employing 5, 11, 21, 51 and 101 α-planes. The Stratic Defuzzifier is shown to be much faster than the Exhaustive Defuzzifier. In fact the Stratic Defuzzifier and the α-Planes/Karnik–Mendel Iterative Procedure Method are comparably speedy; the speed of execution correlates with the number of planes participating in the defuzzification process. The accuracy of the Stratic Defuzzifier is shown to be excellent. It is demonstrated to be more accurate than the α-Planes/Karnik–Mendel Iterative Procedure Method in four of six test cases, regardless of the number of α-planes employed. In one test case, it is less accurate than the α-Planes/Karnik–Mendel Iterative Procedure Method, regardless of the number of α-planes employed. In the remaining test case, the α-Planes/Karnik–Mendel Iterative Procedure Method with 11 α-Planes gives the most accurate result, with the Stratic Defuzzifier coming second.

Prediction of Protein Mutational Free Energy: Benchmark and Sampling Improvements Increase Classification Accuracy.

Software to predict the change in protein stability upon point mutation is a valuable tool for a number of biotechnological and scientific problems. To facilitate the development of such software and provide easy access to the available experimental data, the ProTherm database was created. Biases in the methods and types of information collected has led to disparity in the types of mutations for which experimental data is available. For example, mutations to alanine are hugely overrepresented whereas those involving charged residues, especially from one charged residue to another, are underrepresented. ProTherm subsets created as benchmark sets that do not account for this often underrepresent tense certain mutational types. This issue introduces systematic biases into previously published protocols’ ability to accurately predict the change in folding energy on these classes of mutations. To resolve this issue, we have generated a new benchmark set with these problems corrected. We have then used the benchmark set to test a number of improvements to the point mutation energetics tools in the Rosetta software suite.

BERT-Caps: A Transformer-Based Capsule Network for Tweet Act Classification

Identification of speech acts provides essential cues in understanding the pragmatics of a user utterance. It typically helps in comprehending the communicative intention of a speaker. This holds true for conversations or discussions on any fora, including social media platforms, such as Twitter. This article presents a novel tweet act classifier (speech act for Twitter) for assessing the content and intent of tweets, thereby exploring the valuable communication among the tweeters. With the recent success of Bidirectional Encoder Representations from Transformers (BERT), a newly introduced language representation model that provides pretrained deep bidirectional representations of vast unlabeled data, we introduce BERT-Caps that is built on top of BERT. The proposed model tends to learn traits and attributes by leveraging from the joint optimization of features from the BERT and capsule layer to develop a robust classifier for the task. Some Twitter-specific symbols are also included in the model to observe its influence and importance. The proposed model attained a benchmark accuracy of 77.52% and outperformed several strong baselines and state-of-the-art approaches.

Shale gas production forecasting is an ill-posed inverse problem and requires regularization

Abstract Decline curve analysis (DCA)—the extrapolation of a production curve model fitted to a well’s past production—remains the standard approach for forecasting unconventional oil and gas production. A scaling curve based on a fractured shale gas reservoir model was recently proposed as a way of connecting this approach with underlying physics but as this paper shows, it actually generates worse predictions than the traditional non-physical modified Arps curve. DCA is fundamentally an ill-posed inverse problem with the defining characteristic of model sloppiness, or parameter correlation. Today’s unconventional resource forecasts can be substantially improved by using information from offset wells to reduce ill-posedness through Tikhonov regularization. This versatile approach nearly matches a deep neural network approach introduced here, which has practical limitations but offers a model-neutral benchmark of achievable extrapolation accuracy. There is a natural connection between regularization and a Bayesian formulation which is also highlighted. This paper evaluates long-term forecasting accuracy for these techniques using historic production data from 4457 Barnett shale wells, and reveals that the overlooked step of regularization is more critical than choice of model.