Intelligent Sampling Method Research Articles

Attention2majority: Weak multiple instance learning for regenerative kidney grading on whole slide images.

Deep learning consistently demonstrates high performance in classifying and segmenting medical images like CT, PET, and MRI. However, compared to these kinds of images, whole slide images (WSIs) of stained tissue sections are huge and thus much less efficient to process, especially for deep learning algorithms. To overcome these challenges, we present attention2majority, a weak multiple instance learning model to automatically and efficiently process WSIs for classification. Our method initially assigns exhaustively sampled label-free patches with the label of the respective WSIs and trains a convolutional neural network to perform patch-wise classification. Then, an intelligent sampling method is performed in which patches with high confidence are collected to form weak representations of WSIs. Lastly, we apply a multi-head attention-based multiple instance learning model to do slide-level classification based on high-confidence patches (intelligently sampled patches). Attention2majority was trained and tested on classifying the quality of 127 WSIs (of regenerated kidney sections) into three categories. On average, attention2majority resulted in 97.4%±2.4 AUC for the four-fold cross-validation. We demonstrate that the intelligent sampling module within attention2majority is superior to the current state-of-the-art random sampling method. Furthermore, we show that the replacement of random sampling with intelligent sampling in attention2majority results in its performance boost (from 94.9%±3.1 to 97.4%±2.4 average AUC for the four-fold cross-validation). We also tested a variation of attention2majority on the famous Camelyon16 dataset, which resulted in 89.1%±0.8 AUC1. When compared to random sampling, the attention2majority demonstrated excellent slide-level interpretability. It also provided an efficient framework to arrive at a multi-class slide-level prediction.

A New Representative Sampling Method for Series Size Rock Joint Surfaces

The greatest variability in both shear strength and roughness exists for joint samples with smaller size, which underscores the necessity of performing representative sampling. This study aims to provide a representative sampling method for series size joint surfaces. The progressive coverage statistical method is introduced to provide the sufficient sample capacity for series sampling sizes by setting different propulsion spaces. The statistical law of the joint surface morphology at different sampling sizes is measured by the 3D roughness parameter with {{theta }}_{max }^{ast }/({C}+1). Through an application in nine natural large-scale rock joints, nine consecutive sampling sizes from 100 mm × 100 mm to 900 mm × 900 mm are selected and 121 samples are successfully acquired from each sampling size. According to the frequency distribution of roughness statistics, a new sampling method combining the layering principle and K-medoids clustering algorithm is proposed to screen representative joint samples for each sampling size. The sampling results that meet the test accuracy requirements suggest the possibility of realizing an intelligent sampling method. In addition, the representative of the interlayer cluster center is validated. Finally, the comparison results with the traditional stratified sampling method prove that the proposed method has better stability.

Domain-specific Gaussian process-based intelligent sampling for inspection planning of complex surfaces

Precision measurement of complex surfaces requires intensive sampling for fully characterising the surface geometry and reducing the measurement uncertainty, which is, however, less efficient when the data are costly to acquire. This paper presents a Gaussian process (GP)-based intelligent sampling method for achieving well balance between the measurement efficiency and accuracy. The method makes use of GP to model the surface with domain-specific composite covariance kernel functions. The statistical nature of the GP makes it capable of giving credibility to the arbitrary prediction over the entire established model which can be used in a critical criterion to perform intelligent sampling of the surfaces. The method is independent from the coordinate frames, which makes the sampling plan easily utilised without accurate pre-positioning in actual measurement. The effectiveness of the method is verified through a series of comparison study and actual application in measuring a multi-scaled complex mould insert on coordinate measuring machine.

CAMS-RS: Clustering Algorithm for Large-Scale Mass Spectrometry Data Using Restricted Search Space and Intelligent Random Sampling.

High-throughput mass spectrometers can produce massive amounts of redundant data at an astonishing rate with many of them having poor signal-to-noise (S/N) ratio. These low S/N ratio spectra may not get interpreted using conventional spectra-to-database matching techniques. In this paper, we present an efficient algorithm, CAMS-RS (Clustering Algorithm for Mass Spectra using Restricted Space and Sampling) for clustering of raw mass spectrometry data. CAMS-RS utilizes a novel metric (called F-set) that exploits the temporal and spatial patterns to accurately assess similarity between two given spectra. The F-set similarity metric is independent of the retention time and allows clustering of mass spectrometry data from independent LC-MS/MS runs. A novel restricted search space strategy is devised to limit the comparisons of the number of spectra. An intelligent sampling method is executed on individual bins that allow merging of the results to make the final clusters. Our experiments, using experimentally generated data sets, show that the proposed algorithm is able to cluster spectra with high accuracy and is helpful in interpreting low S/N ratio spectra. The CAMS-RS algorithm is highly scalable with increasing number of spectra and our implementation allows clustering of up to a million spectra within minutes.

High dimensional model representation (HDMR) coupled intelligent sampling strategy for nonlinear problems

High-dimensional model representation (HDMR) is a general set of metamodel assessment and analysis tools to improve the efficiency of high dimensional underlying system behavior. Compared with the current popular modeling methods, such as Kriging (KG), radial basis function (RBF), and the moving least square approximation method (MLS), the distinctive characteristic of the HDMR is to decouple the input variables. Therefore, a high dimensional problem can be transformed as a low, middle or combination of middle dimensional function. Although the HDMR is a feasible method for high dimensional problems, the computational cost is still a bottleneck for complex engineering problems. To improve the efficiency of the HDMR method further, the purpose of this study is to use an intelligent sampling method for the HDMR. Because the HDMR cannot be integrated with the sampling method directly, a projection-based intelligent method is suggested. Compared with the popular HDMR methods, the construction procedure for the HDMR-based model is optimized. To validate the performance of the suggested method, multiple mathematical test functions are given to illustrate the modeling principles, procedures, and the efficiency and accuracy of HDMR models with problems of a wide scope of dimensionalities.

Intelligent sampling-based adaptive space-mapping response surface method for sheet metal forming optimisation

In this study, an adaptive space mapping method is introduced. A Coarse Model (CM) uses quadratic response surfaces for the objective and constraint functions. A fine simulation model is not only used for the correction of the coarse simulation model and validation of the final solution, but is also applied for construction of the Space Mapping (SM) expressions. The intelligent sampling method is implemented for the generation of initial samples and the Particle Swarm Optimisation (PSO) method is applied to optimise the response surface based on the CM in this work. The main purpose is to combine the SM algorithm and the modification of the distribution of the samples strategy. It promises that the response surface based on the CM should be updated stepwise in the right search direction. To demonstrate the application with the proposed optimisation method, it is successfully applied to the optimisation of geometric parameters of the addendum surface and Blank Holder Forces (BHFs) in a sheet-forming problem in a remarkably short computational time.

The least square support vector regression coupled with parallel sampling scheme metamodeling technique and application in sheet forming optimization

The least square support vector regression (LS-SVR) metamodel technique is proposed for sheet metal forming optimization. The major advantage of proposed approach is to build metamodel by consideration of empirical risk minimization (ERM) and structure risk minimization (SRM). In order to construct robust and accurate metamodel based LS-SVR, suitable quantity and intervals of samples are recommended. Thus, a parallel intelligent sampling scheme based on a boundary and best neighbor searching method (BBNS) is proposed to improve the efficiency and accuracy of metamodel. The BBNS was suggested and corresponding practical engineer problems were successfully solved by Hu and Li [Hu W, Li GY, Zhong ZH. Optimization of sheet metal forming processes by adaptive response surface based on intelligent sampling method. J Mater Process Technol 2008;197(1–3):77–88; Hu Wang, Li GY, Li Enying, Zhong ZH. Development of metamodeling based optimization system for high nonlinear engineering problems. Adv Eng Software 2008;39(8)629–45]. To increase the efficiency of metamodel based optimization method, the parallel architecture is implemented for BBNS due to its drawbacks. For validation of developed method, both of serial and parallel BBNS scheme are applied for the nonlinear function. The parallel BBNS is also verified to be an accuracy and efficiency scheme. Finally, the practical nonlinear engineering problems are optimized by suggested methodology and satisfied results are also obtained.

An intelligent sampling method for inspecting free-form surfaces

The two most important factors in sampling during coordinate verification of a surface are the number of measuring points that accurately represent the surface and the distribution of those points. Three algorithms are developed in this paper for sampling of free-form surfaces at a patch scale (not the entire surface). The three developed algorithms take each patch on the surface as a separate unit and distribute the points within the patch in two steps, namely, sample the most critical points that depend on maximum and minimum Gaussian curvature of each patch and add more points depending on the overall distribution of the most critical points distributed on the surface. In each step of sampling, the maximum difference between the CAD model and the surface created using those sampled points is obtained, and an optimum number of measuring points is determined. This minimizes sampling effort and (non-value-added) measurement time. A comparison is performed between the three proposed algorithms and two well known sampling techniques: the equiparametric approach and the patch-size-based sampling method. The proposed algorithms distribute the points in each patch based on a specific plan. The results show that the developed algorithms help select an efficient number of sampling points.

Development of metamodeling based optimization system for high nonlinear engineering problems

A metamodeling optimization system for nonlinear problems was developed in this study. Boundaries and best neighbors searching (BBNS) intelligent sampling method and fuzzy based progressive metamodeling for space reduction were integrated and applied for this system. The BBNS scheme generates new samples derived from information of boundaries and the best samples of initial sparse distributed samples. It is easy to obtain better samples and avoid local convergence due to boundary information. In order to construct accuracy metamodel, the fuzzy based progressive metamodeling method was implemented to cluster samples generated by BBNS several patches in optimization domain. Only better sets of them are involved in construction of metamodels in each patch by response surface and kriging method. The nonlinear problems with multi-humps as test functions were used for proving accuracy and efficiency of developed system. The practical nonlinear engineering problems were also successfully optimized by this system.

Optimization of sheet metal forming processes by adaptive response surface based on intelligent sampling method

Abstract In this study, the previously developed adaptive response surface method (ARSM) is suggested for construction of metamodel for highly non-linear responses. In order to develop the accuracy and efficiency of metamodel, the particle swarm optimization intelligent sampling (PSOIS) scheme is developed. This kind of intelligent method can guarantee the sampling search in right direction and constraint the bounds of design variables in feasible region. For validation of developed method, the Rosenbrock function is successfully approximated by proposed method; corresponding metamodel appropriateness can be well predicted by analysis of variance (ANOVA). Metamodel by ARSM with PSOIS are employed for optimization of initial blank shape and blank hold force (BHF) in sheet forming process, with validations by finite element simulations using LSDYNA970 commercial code. The results show that developed method is able to produce remarkable metamodels for highly non-linear problems with multi-parameter.