Sampling-based Analysis Research Articles

AI-Hub 국내 신호등 데이터셋의 표본추출 기반 주석 오류 분석

본 논문은 AI-Hub 신호등 데이터셋에서 추출한 표본의 정밀한 분석을 통해 주석오류를 분류하는 방법을 제시하고, 주석생성 작업자의 주석 오류와 검증 시스템 오류를 구분함으로써 데이터셋 정제 방향을 제시한다. 또한, 데이터셋의 주석오류가 SSD와 YOLOv4의 객체 검출 정확도에 미치는 영향을 실험적으로 검증한다. This paper presents a labeling-error classification through the precise sampling-based analysis of the AI-Hub traffic light datasets, and suggests the direction of dataset refinement by distinguishing the crowdsourcing worker’s labeling errors from those by the verification system. We also experimentally validate the impacts of labeling errors in the dataset on the mAP of two representative object detection networks: SSD-512 and Scaled-YOLOv4-P6.

Ligand Dilution Analysis Facilitates Aptamer Binding Characterization at the Single-Molecule Level.

Cell-specific aptamers offer a powerful tool to study membrane receptors at the single-molecule level. Most target receptors of aptamers are highly expressed on the cell surface, but difficult to analyze in situ because of dense distribution and fast velocity. Therefore, we herein propose a random sampling-based analysis strategy termed ligand dilution analysis (LDA) for easily implemented aptamer-based receptor study. Receptor density on the cell surface can be calculated based on a regression model. By using a synergistic ligand dilution design, colocalization and differentiation of aptamer and monoclonal antibody (mAb) binding on a single receptor can be realized. Once this is accomplished, precise binding site and detailed aptamer-receptor binding mode can be further determined using molecular docking and molecular dynamics simulation. The ligand dilution strategy also sets the stage for an aptamer-based dynamics analysis of two- and three-dimensional motion and fluctuation of highly expressed receptors on the live cell membrane.

A sequential optimal Latin hypercube design method using an efficient recursive permutation evolution algorithm

Latin hypercube design (LHD) is one of the most frequently used sampling methods. However, most LHDs generate data samples in a manner that hinders computational efficiency and space-filling performance when high dimensions and large samples are involved. Therefore, a sequential recursive evolution Latin hypercube design (RELHD) is proposed in this article, which adopts a permutation inheritance algorithm to update and optimize the LHD. A recursive split algorithm is also proposed and used to enhance the computational efficiency by dividing the sample set into smaller subsets. Numerical experiments demonstrate that the space-filling quality of the RELHD compares well with the enhanced stochastic evolutionary algorithm (ESE) in complex problems with large samples and high dimensions, with RELHD having a significantly higher computational efficiency than ESE. Finally, the sequential approach of RELHD proves to be a more efficient strategy when dealing with sampling-based analysis problems.

Sage

We propose Sage, a system for uncertain network analysis. Algorithms for uncertain network analysis require large amounts of memory and computing resources as they sample a large number of network instances and run analysis on them. Sage makes uncertain network analysis simple and efficient. By extending the edge-centric programming model, Sage makes writing sampling-based analysis algorithms as simple as writing conventional graph algorithms in Pregel-like systems. Moreover, Sage proposes four optimization techniques, namely, deterministic sampling, hybrid gathering, schedule-aware caching, and copy-on-write attributes, that exploit common properties of uncertain network analysis. Extensive evaluation of Sage with eight algorithms on six real-world networks shows that the four optimizations in Sage jointly improve performance by up to 13.9X and on average 2.7X.

Hybrid active learning method for non-probabilistic reliability analysis with multi-super-ellipsoidal model

Non-probabilistic reliability analysis is of great importance in both reliability measure and reliability based design, the efficiency and precision of non-probabilistic reliability analysis, as well as uncertainty quantification, have attracted great attention currently. In this study, considering the coexistence of correlated and independent uncertain-but-bounded variables in engineering applications, a multi-super-ellipsoidal model is used to quantify the uncertainties. Furthermore, inspired by both “norm” based reliability index and “volume-ratio” based reliability index, a hybrid non-probabilistic reliability index is derived to measure the reliability extent more accurately and intuitively. To improve the accuracy and efficiency of solving the hybrid non-probabilistic reliability index, an effective Kriging based hybrid active learning method (HALM) is further developed. Finally, four examples are used to verify the effectiveness and robustness of the proposed HALM. The results show that the selection of multi-super-ellipsoidal model has a certain effect on the estimation of reliability index. Compared with the sampling-based analysis method, HALM presents better performance in terms of the trade-of between in computational efficiency and accuracy.

Microbiota long-term dynamics and prediction of acute graft-versus-host disease in pediatric allogeneic stem cell transplantation

BackgroundPatients undergoing allogeneic hematopoietic stem cell transplantation (HSCT) exhibit changes in their gut microbiota and are experiencing a range of complications, including acute graft-versus-host disease (aGvHD). It is unknown if, when, and under which conditions a re-establishment of microbial and immunological homeostasis occurs. It is also unclear whether microbiota long-term dynamics occur at other body sites than the gut such as the mouth or nose. Moreover, it is not known whether the patients’ microbiota prior to HSCT holds clues to whether the patient would suffer from severe complications subsequent to HSCT. Here, we take a holobiont perspective and performed an integrated host-microbiota analysis of the gut, oral, and nasal microbiota in 29 children undergoing allo-HSCT.ResultsThe bacterial diversity decreased in the gut, nose, and mouth during the first month and reconstituted again 1–3 months after allo-HSCT. The microbial community composition traversed three phases over 1 year. Distinct taxa discriminated the microbiota temporally at all three body sides, including Enterococcus spp., Lactobacillus spp., and Blautia spp. in the gut. Of note, certain microbial taxa appeared already changed in the patients prior to allo-HSCT as compared with healthy children. Acute GvHD occurring after allo-HSCT could be predicted from the microbiota composition at all three body sites prior to HSCT. The reconstitution of CD4+ T cells, TH17, and B cells was associated with distinct taxa of the gut, oral, and nasal microbiota.ConclusionsThis study reveals for the first time bacteria in the mouth and nose that may predict aGvHD. Monitoring of the microbiota at different body sites in HSCT patients and particularly through involvement of samples prior to transplantation may be of prognostic value and could assist in guiding personalized treatment strategies. The identification of distinct bacteria that have a potential to predict post-transplant aGvHD might provide opportunities for an improved preventive clinical management, including a modulation of microbiomes. The host-microbiota associations shared between several body sites might also support an implementation of more feasible oral and nasal swab sampling-based analyses. Altogether, the findings suggest that the microbiota and host factors together could provide actionable information to guiding precision medicine.92cAJ6KC3SeHUVAFGfZKuDVideo

ENGINEERING PROBABILISTIC ANALYSIS USING MODEL REDUCTION AND AUGMENTED RADIAL BASIS FUNCTIONS

Probabilistic analysis of practical engineering problems has long been based on traditional sampling-based approaches, such as Monte Carlo Simulations (MCS) and gradient-based first-order and second-order methods. Since the finite element (FE) or other numerical methods are required to evaluate engineering system responses, such as forces or displacements, it is not efficient to directly integrate FE and sampling-based analysis approaches. Over the years, various approximate methods have been developed and applied to the reliability analysis of engineering problems. In this study, an efficient model reduction technique based on high-dimensional model reduction (HDMR) method has been developed using augmented radial basis functions (RBFs). The basic idea is to use augmented RBFs to construct HDMR component functions. The first-order HDMR model only requires sample points along each variable axis. The HDMR model, once created and used to explicitly express a performance function, is further combined with MCS to perform probabilistic calculations. As test problems, a mathematical problem and a 10-bar truss example are studied using the proposed reliability analysis approach. The proposed method works well, and accurate reliability analysis results are found with a small number of original performance function evaluations, i.e., FE simulations.

Formal Analysis of Unmanned Aerial Vehicles Using Higher-Order-Logic Theorem Proving

The continuous dynamics of unmanned aerial vehicles (UAVs) are generally modeled as a set of differential equations. Traditionally, these continuous dynamics of UAVs are analyzed using paper-and-pencil proof and computer-based testing or simulations to study the performance, stability, and various other control characteristics of the aircraft flying in the air. However, these techniques suffer from their inherent limitations such as human error proneness, sampling-based analysis, approximations of the mathematical results, and the usage of unverified algorithms. Thus, these methods cannot be trusted when considering the utility of UAVs in many safety-critical applications. To overcome the limitations of the aforementioned techniques, it is proposed to use higher-order-logic theorem proving for formally analyzing the continuous dynamics of UAVs. In particular, a formalization of complex-valued matrices in higher-order logic is provided using the HOL Light theorem prover, which is in turn used for the formalization of the navigation’s and aircraft’s body-fixed frames, as well as their associated transformations. Formal reasoning support is also provided for analyzing the multiple-input multiple-output systems, which are in turn used for formally analyzing the continuous dynamics of UAVs using HOL Light. For illustration, we use our proposed framework for the formal stability analysis of the CropCam UAV using HOL Light.

GP3: A Sampling-based Analysis Framework for Gaussian Processes

Although machine learning is increasingly applied in control approaches, only few methods guarantee certifiable safety, which is necessary for real world applications. These approaches typically rely on well-understood learning algorithms, which allow formal theoretical analysis. Gaussian process regression is a prominent example among those methods, which attracts growing attention due to its strong Bayesian foundations. Even though many problems regarding the analysis of Gaussian processes have a similar structure, specific approaches are typically tailored for them individually, without strong focus on computational efficiency. Thereby, the practical applicability and performance of these approaches is limited. In order to overcome this issue, we propose a novel framework called GP3, general purpose computation on graphics processing units for Gaussian processes, which allows to solve many of the existing problems efficiently. By employing interval analysis, local Lipschitz constants are computed in order to extend properties verified on a grid to continuous state spaces. Since the computation is completely parallelizable, the computational benefits of GPU processing are exploited in combination with multi-resolution sampling in order to allow high resolution analysis.

An efficient local search-based genetic algorithm for constructing optimal Latin hypercube design

ABSTRACTLatin hypercube design (LHD) is a multi-stratified sampling method, which has been frequently used in sampling-based analysis. To achieve good space-filling quality of LHD, an efficient method, termed local search-based genetic algorithm (LSGA), is proposed in this article for constructing an optimal LHD. LSGA adopts modified order crossover, probabilistic mutation and adaptive selection operators to enrich population diversity and speed up convergence. A local search strategy is also presented in the approach to enhance the search ability. The performance of the proposed method is compared with several established methods in three perspectives, namely space-filling quality, computational efficiency and predictive accuracy of the metamodel. Several numerical experiments with distinct dimensions and numbers of design points are studied, and the results demonstrate that the proposed method performs better than other methods when dealing with LHD construction issues with high dimension and a large number of sampling points.

Continuous monitoring of water quality at aeration plant with potentiometric sensor array

The potential of multisensor array in continuous on-line monitoring of processed water quality at aeration plant is explored. The responses of 23 potentiometric sensors were continuously registered every seven seconds through the 26 days of experiment in a special container having connection with outlet water line of the aeration plant. Using chemometric tools it is shown that potentiometric “electronic tongue” is capable of evaluation of two important parameters of water quality: ammonium nitrogen and nitrate nitrogen. Unlike traditional sampling-based analysis the results of multisensor are available immediately in a real-time mode. Moreover, the achieved precision is sufficient to monitor possible alarm events. The employment of topological data analysis allowed for exploration of the very large dataset (295,828 measurements) accumulated through the long period of continuous measurements with sensors and for judgment on stability of water quality.

Sensitivity of the simulated Oxygen Minimum Zone to biogeochemical processes at an oligotrophic site in the Arabian Sea

Oxygen minimum zones (OMZs) are large, low-oxygen areas in the global oceans. Although OMZs represent a serious threat to ecosystem functioning and services, our capability of modelling the main biogeochemical processes driving OMZ dynamic are still limited. Here we performed a full sensitivity analysis of a complex ecosystem model to rank the most important biogeochemical parameters influencing the simulation of the OMZ at an oligotrophic site in the open Arabian Sea. We applied a one-dimensional configuration of the European Regional Seas Ecosystem Model (ERSEM) - here advanced by including denitrification - coupled with the General Ocean Turbulence Model (GOTM). The coupled model was skilled in simulating the vertical gradients of climatological data of oxygen and nutrients. The sensitivity analysis of the model was carried out in two steps: i) a preliminary Morris screening analysis of 207 ERSEM parameters, which selected the three most influential groups of parameters; and ii) a subsequent Monte Carlo sampling-based analysis for ranking the importance of the 38 parameters within the three selected groups. Overall, the four most important parameters for the simulation of the minimum oxygen concentration were found to be: 1) the cubic half saturation constant for oxygenic control of denitrification; 2) the parameter regulating the fraction of ingested matter excreted by heterotrophic nanoflagellates; 3) the bacterial efficiency at low oxygen levels; and 4) the specific rate of bacterial release of capsular material. Based on these findings, and assuming that the ranking of the model parameters reflects the relevance of the process they characterize, we present a conceptual model describing the most important biogeochemical processes affecting the OMZ at the study site. Our results suggest that including bacteria explicitly in ecosystem models is useful to simulate and predict OMZs, provided that efforts are invested in estimating parameters characterizing the microbial loop in marine ecosystems.

Progressive Latin Hypercube Sampling: An efficient approach for robust sampling-based analysis of environmental models

Efficient sampling strategies that scale with the size of the problem, computational budget, and users’ needs are essential for various sampling-based analyses, such as sensitivity and uncertainty analysis. In this study, we propose a new strategy, called Progressive Latin Hypercube Sampling (PLHS), which sequentially generates sample points while progressively preserving the distributional properties of interest (Latin hypercube properties, space-filling, etc.), as the sample size grows. Unlike Latin hypercube sampling, PLHS generates a series of smaller sub-sets (slices) such that (1) the first slice is Latin hypercube, (2) the progressive union of slices remains Latin hypercube and achieves maximum stratification in any one-dimensional projection, and as such (3) the entire sample set is Latin hypercube. The performance of PLHS is compared with benchmark sampling strategies across multiple case studies for Monte Carlo simulation, sensitivity and uncertainty analysis. Our results indicate that PLHS leads to improved efficiency, convergence, and robustness of sampling-based analyses.

Computational Assessment of the Relation Between Embolism Source and Embolus Distribution to the Circle of Willis for Improved Understanding of Stroke Etiology.

Stroke caused by an embolism accounts for about a third of all stroke cases. Understanding the source and cause of the embolism is critical for diagnosis and long-term treatment of such stroke cases. The complex nature of the transport of an embolus within large arteries is a primary hindrance to a clear understanding of embolic stroke etiology. Recent advances in medical image-based computational hemodynamics modeling have rendered increasing utility to such techniques as a probe into the complex flow and transport phenomena in large arteries. In this work, we present a novel, patient-specific, computational framework for understanding embolic stroke etiology, by combining image-based hemodynamics with discrete particle dynamics and a sampling-based analysis. The framework allows us to explore the important question of how embolism source manifests itself in embolus distribution across the various major cerebral arteries. Our investigations illustrate prominent numerical evidence regarding (i) the size/inertia-dependent trends in embolus distribution to the brain; (ii) the relative distribution of cardiogenic versus aortogenic emboli among the anterior, middle, and posterior cerebral arteries; (iii) the left versus right brain preference in cardio-emboli and aortic-emboli transport; and (iv) the source-destination relationship for embolisms affecting the brain.

Identification and Validation of Genetic Variants that Influence Transcription Factor and Cell Signaling Protein Levels

Many genetic variants associated with human disease have been found to be associated with alterations in mRNA expression. Although it is commonly assumed that mRNA expression changes will lead to consequent changes in protein levels, methodological challenges have limited our ability to test the degree to which this assumption holds true. Here, we further developed the micro-western array approach and globally examined relationships between human genetic variation and cellular protein levels. We collected more than 250,000 protein level measurements comprising 441 transcription factor and signaling protein isoforms across 68 Yoruba (YRI) HapMap lymphoblastoid cell lines (LCLs) and identified 12 cis and 160 trans protein level QTLs (pQTLs) at a false discovery rate (FDR) of 20%. Whereas up to two thirds of cis mRNA expression QTLs (eQTLs) were also pQTLs, many pQTLs were not associated with mRNA expression. Notably, we replicated and functionally validated a trans pQTL relationship between the KARS lysyl-tRNA synthetase locus and levels of the DIDO1 protein. This study demonstrates proof of concept in applying an antibody-based microarray approach to iteratively measure the levels of human proteins and relate these levels to human genome variation and other genomic data sets. Our results suggest that protein-based mechanisms might functionally buffer genetic alterations that influence mRNA expression levels and that pQTLs might contribute phenotypic diversity to a human population independently of influences on mRNA expression.

A novel image-analysis toolbox enabling quantitative analysis of root system architecture.

We present in this paper a novel, semiautomated image-analysis software to streamline the quantitative analysis of root growth and architecture of complex root systems. The software combines a vectorial representation of root objects with a powerful tracing algorithm that accommodates a wide range of image sources and quality. The root system is treated as a collection of roots (possibly connected) that are individually represented as parsimonious sets of connected segments. Pixel coordinates and gray level are therefore turned into intuitive biological attributes such as segment diameter and orientation as well as distance to any other segment or topological position. As a consequence, user interaction and data analysis directly operate on biological entities (roots) and are not hampered by the spatially discrete, pixel-based nature of the original image. The software supports a sampling-based analysis of root system images, in which detailed information is collected on a limited number of roots selected by the user according to specific research requirements. The use of the software is illustrated with a time-lapse analysis of cluster root formation in lupin (Lupinus albus) and an architectural analysis of the maize (Zea mays) root system. The software, SmartRoot, is an operating system-independent freeware based on ImageJ and relies on cross-platform standards for communication with data-analysis software.

Sampling-Based Analysis of a Spatially Distributed Model for Liver Metabolism at Steady State

In order for predictive models of complex metabolic systems to be a suitable replacement for in vivo experiments, they need to capture the essential features of the systems while remaining computationally tractable. Mathematical models of hepatic metabolism need to be spatially distributed if they are used to test whether the spatial location has any influence on the metabolic processes. The main computational challenge of a spatially distributed metabolic model is the large number of parameters. In this paper we describe a spatially distributed two-compartment (blood and tissue) model for liver metabolism, which can be used for various in silico experiments. The model consists of a system of partial differential equations describing the change of concentrations of metabolites and metabolic intermediates in time and space due to convection, diffusion, transport between the compartments, and biochemical reactions. After the introduction of spatial discretization by a finite difference approximation we view...