Mixed Input Research Articles

Dual-branch network with hypergraph feature augmentation and adaptive logits adjustment for long-tailed visual recognition

The long-tailed distribution presents issues of data scarcity and significant class imbalance, which causes the model to increase the prediction tendency for the head classes and reduce performance on tail classes. The decoupling learning strategy separates the learning process into representation learning and classifier learning to enhance model performance. However, this approach does not delve into the complex semantic relationships within images and overlooks the bias of the model confidence. To address this pervasive issue, we propose a novel algorithm called Hypergraph Feature Augmentation and Adaptive Logits Adjustment for Long-tailed Visual Recognition (HALR) based on a decoupling learning framework. For the representation learning task, we extract hypergraph features from the mixed input samples to capture the global spatial contextual semantic information of the images. For the classifier learning task, we propose an adaptive logits adjustment function that automatically corrects prediction score biases, thereby yielding robust decision boundaries. Extensive experiments on widely-used benchmark datasets, including CIFAR10/100-LT, ImageNet-LT, and iNaturalist 2018, validate the efficacy of HALR. For example, the test accuracy of HALR on CIFAR10-LT with an imbalance factor of 100 is 88.25 %, which is a 17.85 % improvement over the baseline model accuracy and 0.68 % higher than the state-of-the-art model. The results demonstrate that HALR effectively optimizes representation and classifier learning tasks for long-tailed learning, mitigating the negative impact of class imbalance and data scarcity.

Source identification of nitrate in groundwater of an agro-pastoral ecotone in a semi-arid zone, northern China: Coupled evidences from MixSIAR model and DOM fluorescence

In the extensive agro-pastoral ecotone of northern China, groundwater is a valuable resource, but it suffers from severe nitrate pollution. However, the extent of the contributions of agricultural and pastoral activities to nitrate accumulation in the groundwater of the region remains unclear. This study aimed to identify the main sources of groundwater nitrate in Chahannur Basin, a typical agro-pastoral ecotone in the semi-arid zone of northern China, using the MixSIAR model based on dual stable isotopes and dissolved organic matter (DOM) fluorescence. The hydrochemical and isotopic results showed that the groundwater has high concentrations of nitrate (up to 208.19 mg/L) with weak denitrification, and nitrate accumulation is mainly driven by mixed input from different sources. The MixSIAR model results indicated that the largest contributors to groundwater nitrate accumulation are manure & sewage, followed by soil nitrogen, chemical fertilizers, and atmospheric deposition. According to the DOM fluorescence characteristics, the groundwater is strongly affected by livestock and poultry waste, followed by soil humic substances. Moreover, the DOM fluorescence results further supported the MixSIAR model results based on the significant correlation between the contributions of nitrate sources and the percentages of fluorescent components, jointly confirming that the main source of nitrate in groundwater is manure and sewage, followed by soil nitrogen. These findings indicate that coupled evidences from the MixSIAR model and DOM fluorescence could be applied to identify the sources of nitrate in groundwater, and this coupling can provide valuable information for local authorities to achieve sustainable groundwater management.

Robust generation of N-partite N-level singlet states by identical particle interferometry

Abstract We propose an interferometric scheme for generating the totally antisymmetric state of N identical bosons with N internal levels (generalized singlet). This state is a resource for various problems with dramatic quantum advantage. The procedure uses a sequence of Fourier multi-ports, combined with coincidence measurements filtering the results. Successful preparation of the generalized singlet is confirmed when the N particles of the input state stay separate (anti-bunch) on each multiport. The scheme is robust to local lossless noise and works even with a totally mixed input state.

Quantum state estimation based on deep learning

Abstract We used deep learning techniques to construct various models for reconstructing quantum states from a given set of coincidence measurements. Through simulations, we demonstrated that our approach generates functionally equivalent reconstructed states for a wide range of pure and mixed input states. Compared to traditional methods, our system offers the advantage of faster speed. Additionally, by training our system with measurement results containing simulated noise sources, we could show a significant improvement in average fidelity compared to typical reconstruction methods. We also found that constraining the variational manifold to physical states, i.e., positive semi-definite density matrices, greatly enhances the quality of the reconstructed states in the presence of experimental imperfections and noise. Finally, we validated the correctness and superiority of our model by using data generated on IBMQ, a real quantum computer.

MixTrain: accelerating DNN training via input mixing.

Training Deep Neural Networks (DNNs) places immense compute requirements on the underlying hardware platforms, expending large amounts of time and energy. An important factor contributing to the long training times is the increasing dataset complexity required to reach state-of-the-art performance in real-world applications. To address this challenge, we explore the use of input mixing, where multiple inputs are combined into a single composite input with an associated composite label for training. The goal is for training on the mixed input to achieve a similar effect as training separately on each the constituent inputs that it represents. This results in a lower number of inputs (or mini-batches) to be processed in each epoch, proportionally reducing training time. We find that naive input mixing leads to a considerable drop in learning performance and model accuracy due to interference between the forward/backward propagation of the mixed inputs. We propose two strategies to address this challenge and realize training speedups from input mixing with minimal impact on accuracy. First, we reduce the impact of inter-input interference by exploiting the spatial separation between the features of the constituent inputs in the network's intermediate representations. We also adaptively vary the mixing ratio of constituent inputs based on their loss in previous epochs. Second, we propose heuristics to automatically identify the subset of the training dataset that is subject to mixing in each epoch. Across ResNets of varying depth, MobileNetV2 and two Vision Transformer networks, we obtain upto 1.6 × and 1.8 × speedups in training for the ImageNet and Cifar10 datasets, respectively, on an Nvidia RTX 2080Ti GPU, with negligible loss in classification accuracy.

Accuracy of vestibular schwannoma segmentation using deep learning models - a systematic review & meta-analysis.

Following 2020 PRISMA guidelines, a search across four databases was conducted. Inclusion criteria focused on articles using DL for VS MR image segmentation. The primary metric was the Dice score, supplemented by relative volume error (RVE) and average symmetric surface distance (ASSD). The search process identified 752 articles, leading to 11 studies for meta-analysis. A QUADAS- 2 analysis revealed varying biases. The overall Dice score for 56 models was 0.89 (CI: 0.88-0.90), with high heterogeneity (I2 = 95.9%). Subgroup analyses based on DL architecture, MRI inputs, and testing set sizes revealed performance variations. 2.5D DL networks demonstrated comparable efficacy to 3D networks. Imaging input analyses highlighted the superiority of contrast-enhanced T1-weighted imaging and mixed MRI inputs. This study fills a gap in systematic review in the automated segmentation of VS using DL techniques. Despite promising results, limitations include publication bias and high heterogeneity. Future research should focus on standardized designs, larger testing sets, and addressing biases for more reliable results. DL have promising efficacy in VS diagnosis, however further validation and standardization is needed. In conclusion, this meta-analysis provides comprehensive review into the current landscape of automated VS segmentation using DL. The high Dice score indicates promising agreement in segmentation, yet challenges like bias and heterogeneity must be addressed in the future research.

Optimizing SEM parameters for segmentation with AI – Part 2: Designing and training a regression model

Selecting the best microscope parameters for optimal image quality currently relies on microscopists; there exist no procedures or guidelines for tuning parameters to ensure the desired image quality is achieved. More importantly, for quantitative analysis purposes, adequate image quality for segmentation should be prioritized. This paper is the second of two parts, describing a regression model, mixed input, multiple output with Keras TensorFlow, trained to predict the beam energy and probe current, two important parameters for image quality. Specifically, parameters are predicted to optimize the image quality for segmentation, using a generated training set, as described in Part 1 of this paper. Model performance is then tested on models trained with multiple different training sets, and with different proportions of simulated and acquired data. First, to examine the impact of the training set on the prediction accuracy and then, to evaluate the importance of including real data during training. The model successfully predicted the beam energy and probe current to set on the microscope to improve image quality for segmentation. Models trained with both simulated and acquired data performed the best, as evaluated by their efficacy at improving the image quality for feature segmentation.

Aliphatic and polycyclic aromatic compounds in coal and coal-based solid wastes: Relationship with coal-forming paleoenvironment and implications for environmental pollution

In this study, coal and coal-based solid wastes (coal gangue, fly ash, bottom ash, desulfurized gypsum and tar residue) were collected from major coal mines, power plants and coking plants in Lianghuai mining area (LH), China, and were analyzed for 76 polycyclic aromatic compounds (PACs), 27 n-alkanes and 2 isoprenoids (phytane and pristane). The total n-alkanes concentrations and ∑76 PACs in raw coals (640 ± 600 and 180 ± 87 μg/g) were higher than those in coal-based solid wastes (47 ± 40 and 24 ± 25 μg/g), but were lower than those in tar residue (3700 and 63,000 μg/g). It was discovered that the depositional paleoenvironment in LH was mostly a lacustrine and freshwater environment with oxidizing conditions and mixed organic matter input, but the Huainan coalfield had stronger oxidizing conditions and more input of terrestrial organic matter than that of the Huaibei coalfield. Alkylated PACs made up 56 ± 12 % of the ∑76PACs in raw coals, whereas solid wastes mainly consisted of 16 EPA PAHs (66 ± 16 %). Coal combustion and gangue weathering altered the structural properties of n-alkanes and PACs, resulting in a significant loss of n-alkanes and PACs, a higher proportion of parent PACs, and an increased abundance of short n-alkanes in the products (No apparent change of n-alkanes composition was observed through gangue weathering). The toxicity of PACs in raw coal and its solid wastes in LH from high to low was tar residue, raw coal, coal gangue, and coal-fired products. This investigation further confirmed that traditional diagnostic ratios may distort source information, and that they should not be used to assess PACs sources from raw coal particles or coal gangues, but rather to identify combustion sources near the point source. In addition, Retene/(Retene + Chrysene) < 0.03 may indicate direct contamination of raw coal particles.

The tubular baffled reactor and its potential for the biological methanation of carbon dioxide

The biological Power to Methane process (PtM) is gaining ground as an answer to the long-term renewable energy storage problem. Methane is an efficient hydrogen carrier, has an established worldwide transport infrastructure and can serve as a link between renewable power generation and a circular carbon economy. One of the defining factors regarding the scalability of the PtM process is the design of the reactor as it can determine the production rate/energy expenditure ratio. The tubular baffled reactor, a popular reactor design within the chemical industry has been assessed in the present study as a biomethanation reactor for the first time. The experiments were conducted with mixed cultures and the results point to high gas-liquid mass transfer capabilities as indicated by the methanation rates achieved (>90 % CH4 at 270 L/L/d mixed gas input rate). The gas/liquid flow ratio appears to have a stronger effect on methanation than the gas residence time. The working length of the reactor determines the pressure drop experienced by the culture, with higher pressure drops showing a negative correlation to methanogenesis.

Goats on the Move: Evaluating Machine Learning Models for Goat Activity Analysis Using Accelerometer Data.

Putting sensors on the bodies of animals to automate animal activity recognition and gain insight into their behaviors can help improve their living conditions. Although previous hard-coded algorithms failed to classify complex time series obtained from accelerometer data, recent advances in deep learning have improved the task of animal activity recognition for the better. However, a comparative analysis of the generalizing capabilities of various models in combination with different input types has yet to be addressed. This study experimented with two techniques for transforming the segmented accelerometer data to make them more orientation-independent. The methods included calculating the magnitude of the three-axis accelerometer vector and calculating the Discrete Fourier Transform for both sets of three-axis data as the vector magnitude. Three different deep learning models were trained on this data: a Multilayer Perceptron, a Convolutional Neural Network, and an ensemble merging both called a hybrid Convolutional Neural Network. Besides mixed cross-validation, every model and input type combination was assessed on a goat-wise leave-one-out cross-validation set to evaluate its generalizing capability. Using orientation-independent data transformations gave promising results. A hybrid Convolutional Neural Network with L2-norm as the input combined the higher classification accuracy of a Convolutional Neural Network with the lower standard deviation of a Multilayer Perceptron. Most of the misclassifications occurred for behaviors that display similar accelerometer traces and minority classes, which could be improved in future work by assembling larger and more balanced datasets.

Accelerating regional-scale groundwater flow simulations with a hybrid deep neural network model incorporating mixed input types: A case study of the northeast Qatar aquifer

ABSTRACT This study presents the ‘Dual Path CNN-MLP’, a novel hybrid deep neural network (DNN) architecture that merges the strengths of convolutional neural networks (CNNs) and multilayer perceptrons (MLPs) for regional groundwater flow simulations. This model stands out from previous DNN approaches by managing mixed input types, including both imagery and numerical vectors. Such flexibility allows the diverse nature of groundwater data to be efficiently utilized without the need to convert it into a uniform format, which often leads to oversimplification or unnecessary expansion of the dataset. When applied to the northeast Qatar aquifer, the model demonstrates high accuracy in simulating transient groundwater flow fields, benchmarked against the well-established MODFLOW model. The model's efficacy is confirmed through k-fold cross-validation, showing an error margin of less than 12% across all examined locations. The study also examines the model's ability to perform uncertainty analysis using Monte Carlo simulations, finding that it achieves around 1% average absolute percentage error in estimating the mean hydraulic head. Errors are mostly found in areas with significant variations in the hydraulic head. Switching to this ML model from the conventional MODFLOW simulator boosts computational efficiency by about 99%, showcasing its advantage for tasks like uncertainty analysis in repetitive groundwater simulations.

Emergence of differential object marking in Asia Minor Greek

Abstract This paper investigates the emergence of differential object marking (DOM) in the Asia Minor Greek dialect of Pharasa (PhG) under contact with Turkish. We show that DOM in Turkish and PhG are both instances of structural accusative case and DOM can be formally modeled as context sensitive dependent case. We propose that two factors caused the emergence of DOM in PhG, namely (i) case neutralization in indefinite contexts, and (ii) an increase in the number of V-NP idioms borrowed from Turkish where the NP is in bare form. These perturbations led to a significant change in the overall data created by the community resulting in mixed input for the younger generations. Once the amount of bare NPs passed a certain threshold, a divergent grammar became inevitable. We test our proposal using an abductive generalization learning algorithm based on the Tolerance Principle and running a number of simulations. Our simulation results confirm our hypothesis.

Global sensitivity analysis and optimization of multi-stage TEG based on random and fuzzy mixed input variables

In practical engineering, uncertainties in thermoelectric power generation arise from various factors such as material properties, geometric dimensions, assembly processes, and operational fluctuations of the thermoelectric components. Addressing subjective and objective uncertainties in the design parameters of multi-stage thermoelectric generators (TEGs), this study establishes a robust TEG model incorporating non-equilibrium thermodynamics. Fuzzy and stochastic theories are employed for uncertainty representation, and a variance-based global sensitivity analysis is conducted. The results indicate that the uncertainties in the design variables have varying effects on the output responses of the three-stage TEG. Among them, the temperature of the hot and cold sources has the most significant impact on the output responses（STHP: 0.5018, STLP:0.2199, STHη: 0.5436, STLη: 0.3336）, Parameters like Seebeck coefficient, thermal resistance, thermal conductivity, and operating current exhibit relatively minor effects. The coefficient of variation of the output response varies consistently with changes in the standard deviation or coefficient of variation of input variables, following a similar order as the sensitivity indicators. The variations in the distribution parameters of the key design variable, i.e., the temperature of the hot and cold sources, do not affect the sensitivity index rankings of the variables concerning the system output responses. As the distribution parameters of the hot and cold source temperatures gradually increase, the sensitivity indices of the design variables to the system output responses exhibit different trends, indicating a change in the degree of sensitivity to the variations in distribution parameters of the design variables. Optimization based on global sensitivity analysis enhances power output by 43.37 %, reaching a maximum of 11.37 W under constraints compared to the original 7.93 W. This research provides guidance for designing and optimizing thermoelectric power systems under mixed uncertainties.

Fecal biomarkers in Italian anthropogenic soil horizons and deposits from Middle Ages and bronze age

Archaeological excavations in urban and rural contexts often uncover dark homogeneous anthropogenic deposits, soils and soil horizons, known as Dark Earths, Cultural Layers and Anthrosols. Major scientific questions arise about the processes that lead to the formation of these soils and deposits, as they are often related to a complex combination of environmental, climatic, and anthropogenic factors. Many studies focused on the morphological and physical-chemical investigation of the archaeological remains found as well as the inorganic constituent of the soil. The organic fraction has been only roughly investigated and studies carried out on a molecular level are very limited. However, a more refined investigation of the chemical constituents of these soils can potentially reveal pivotal insights on the archaeological context such as past domestic habits and/or urban organization. It has been demonstrated that, among organic matter constituents, fecal steroids are promising biomarkers to track fecal input in many environmental matrixes such as water, soils, and sediments. In this paper we determine the fecal sterols and stanols content in dark homogeneous anthropogenic deposits and soils found in seven archaeological sites in Italy (from Verona, Como, Florence, Mel and Conegliano) dated from the Late Bronze Age and the Medieval Period, aiming to assess the major contribution of the fecal input using diagnostic fecal indexes obtained from specific fecal sterol proportions. To this purpose, a baseline was established by identifying the fecal fingerprint of domestic herbivores (caprids, ovines, equines and bovines) and omnivores (pigs). Based on their fecal fingerprint, we classified the deposits of seven studied sites into four distinct categories: i) the medieval sites of Verona, presented a strong mixed fecal input from both herbivores and humans; ii) the medieval sites of Como and Florence which showed a similar pattern, with a predominance of human fecal input over livestock; iii) the late medieval soils of Mel and the late bronze age Cultural Layers of Castellar di Leppia (Verona) which exhibited a generally scarce fecal contamination while iv) the final bronze age site of Conegliano (Treviso), a strong fecal contamination from herbivores was detected. These results thus demonstrate that for the Medieval Dark Earth the socio-cultural processes leading to the formation of the Dark Earth differ in terms of sewage and waste management and/or in animal husbandry practices.

On the Design of a Stepped Mixed Signal Transmission Line with Defected Microstrip Structure for High-Frequency PCB Traces

The paper addresses the challenge of crosstalk in neighboring paths on PCBs, emphasizing its growing importance with the advancement of electronic circuits in terms of speed and frequency. The proposed solution involves the introduction of a novel W-shaped Defected Microstrip Structure (DMS) on a step-shaped coupled transmission lines to mitigate near-end crosstalk (NEXT) and far-end crosstalk (FEXT) using mixed signal input. A novel W-shaped defective microstrip structure (DMS) that is etched on step-shaped coupled transmission lines. The proposed technique focuses on minimizing the electromagnetic coupling strength with a minimum spacing of one time the width of the trace. It optimizes the ratio of capacitive coupling and inductive coupling between the coupled microstrip transmission lines to improve crosstalk. The frequency domain simulation, conducted using Computer Simulation Technology (CST), indicates promising results. The insertion loss (|S21|), near-end crosstalk (|S31|), and far-end crosstalk (|S41|) of the suggested structure are reported as 1.425dB, &lt;25dB, and &lt;25dB, respectively, over the frequency range of 0 to 10 GHz. Eye diagram characterizations are done and measured in terms of eye height, eye width, jitter and data rate are reported as 0.942V, 695.5ps, 6.69ps and 1.4Gbps respectively.

Acquisition pattern and the role of vocabulary and language experience in the acquisition of inflectional grammar by Mandarin-English speaking preschoolers.

Australian Mandarin-English bilingual preschoolers must acquire linguistic structures that occur only in the community language (e.g., English inflectional grammar). This study investigated how they acquire such structures and any relationship between linguistic knowledge and language experience on their performance. Twenty 4-6-year-olds showed known monolingual acquisition patterns with good performance for producing the progressive, developing ability for plurals, but only emerging ability for past and present tense. Better performance was related to a larger English vocabulary, more mixed language input and use, but less Mandarin input and use. On average, these children received less than 50% input in English and were performing behind monolinguals.

The Genome of Arsenophonus sp. and Its Potential Contribution in the Corn Planthopper, Peregrinus maidis.

The co-evolution between symbionts and their insect hosts has led to intricate functional interdependencies. Advances in DNA-sequencing technologies have not only reduced the cost of sequencing but, with the advent of highly accurate long-read methods, have also enabled facile genome assembly even using mixed genomic input, thereby allowing us to more easily assess the contribution of symbionts to their insect hosts. In this study, genomic data recently generated from Peregrinus maidis was used to assemble the genome of a bacterial symbiont, Pm Arsenophonus sp. This ~4.9-Mb assembly is one of the largest Arsenophonus genomes reported to date. The Benchmarking Universal Single-Copy Orthologs (BUSCO) result indicates that this Pm Arsenophonus assembly has a high degree of completeness, with 96% of the single-copy Enterobacterales orthologs found. The identity of the Pm Arsenophonus sp. was further confirmed by phylogenetic analysis. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis indicates a major contribution by Pm Arsenophonus sp. to the biosynthesis of B vitamins and essential amino acids in P. maidis, where threonine and lysine production is carried out solely by Pm Arsenophonus sp. This study not only provides deeper insights into the evolutionary relationships between symbionts and their insect hosts, but also adds to our understanding of insect biology, potentially guiding the development of novel pest control methods.

Quantum Kolmogorov complexity and quantum correlations in deterministic-control quantum Turing machines

This work presents a study of Kolmogorov complexity for general quantum states from the perspective of deterministic-control quantum Turing Machines (dcq-TM). We extend the dcq-TM model to incorporate mixed state inputs and outputs, and define dcq-computable states as those that can be approximated by a dcq-TM. Moreover, we introduce (conditional) Kolmogorov complexity of quantum states and use it to study three particular aspects of the algorithmic information contained in a quantum state: a comparison of the information in a quantum state with that of its classical representation as an array of real numbers, an exploration of the limits of quantum state copying in the context of algorithmic complexity, and study of the complexity of correlations in quantum systems, resulting in a correlation-aware definition for algorithmic mutual information that satisfies symmetry of information property.

Organic Geochemical Evaluation of the Black Shale interbeds Within the Lithological Succession of the Al Aziziyah Formation, NW Libya

The samples under investigation represent the black shale in interbeds within the Al Aziziyah Formation, located in the NW part of Libya and were analysed using geochemical techniques method and a variety of organic geochemical parameters. The aim of this study is to assess the type of organic matter, thermal maturity, and environment of deposition, based on biomarker distributions and Rock-Eval pyrolysis. The Gas chromatograms of the saturated hydrocarbon fractions of the study samples display a smooth high-end member distribution of the n- n-alkanes extending beyond nC34. The sample displayed evidence of biodegradation as suggested by the presence of the loss of some of the lower molecular weight n-alkanes. The predominance of high molecular weight n-alkane in the study sample suggests a significant input of higher land plant organic matter into these sediments. Geochemical (Rock-Eval, Tmax. values ranging from 416-4390C typical of immature range). Another related feature of these rocks extracted is the presence of a high relative abundance of gammacerane, indicating an anoxic marine hypersaline source depositional environment. The relatively high abundance of common land-plant-derived biomarkers, such as oleananes, is a clear indication of major terrigenous input to the source of these extractable organic matters. The high abundance of C24 tetracyclic terpanes could be associated with higher land-plant, algae, or microbial sources, thereby suggesting a mixed source input. TOC valued &lt; 0.5 wt.%, in the analyzed samples can be considered as a poor source rock generative potential. Based on this study, it can be said that the interbedded shales of the Al Aziziyah Formation seem to have received substantial amounts of land-derived organic matter that has been transported into an open marine subtidal depositional setting had prevailed and was deposited in a photosynthetic-organisms dominated environment (thereby suggesting a mixed source input).

SMT 2.0: A Surrogate Modeling Toolbox with a focus on hierarchical and mixed variables Gaussian processes

The Surrogate Modeling Toolbox (SMT) is an open-source Python package that offers a collection of surrogate modeling methods, sampling techniques, and a set of sample problems. This paper presents SMT 2.0, a major new release of SMT that introduces significant upgrades and new features to the toolbox. This release adds the capability to handle mixed-variable surrogate models and hierarchical variables. These types of variables are becoming increasingly important in several surrogate modeling applications. SMT 2.0 also improves SMT by extending sampling methods, adding new surrogate models, and computing variance and kernel derivatives for Kriging. This release also includes new functions to handle noisy and use multi-fidelity data. To the best of our knowledge, SMT 2.0 is the first open-source surrogate library to propose surrogate models for hierarchical and mixed inputs. This open-source software is distributed under the New BSD license.22https://github.com/SMTorg/SMT.