基于Shapley additive explanations(SHAP)分析乳腺X线检查假阴性乳腺癌的特征

目的:基于Shapley additive explanations(SHAP)分析乳腺X线摄影(MG)检查假阴性乳腺癌的特征。方法:回顾性收集我院2018年1月—2022年5月首诊乳腺癌患者,评估临床病理和影像学表现。通过logistic回归分析确定MG假阴性乳腺癌的独立危险因素。应用SHAP分析各独立危险因素的贡献价值。结果:1505个病灶纳入研究,MG假阴性乳腺癌373个。较小的病灶、致密或不均匀致密的MG密度、无可疑钙化、非肿块样强化(NME)、原位癌和腋窝淋巴结阴性与MG假阴性乳腺癌独立相关,上述参数构建的联合模型具有良好的预测效能(AUC=0.871)。根据SHAP分析,病灶大小、MG密度和病灶MRI强化类型是导致MG假阴性的主要因素;钙化状态、病理类型和腋窝淋巴结状态是重要补充。结论:MG假阴性乳腺癌具有一定的特征,此类患者需补充乳腺MRI检查以提高癌灶的检出率。

Visualization for Explanation of Deep Learning-Based Defect Detection Model Using Class Activation Map

Recently,convolutional neural network(CNN)-based visual inspec-tion has been developed to detect defects on building surfaces automatically.The CNN model demonstrates remarkable accuracy in image data analysis;however,the predicted results have uncertainty in providing accurate informa-tion to users because of the“black box”problem in the deep learning model.Therefore,this study proposes a visual explanation method to overcome the uncertainty limitation of CNN-based defect identification.The visual repre-sentative gradient-weights class activation mapping(Grad-CAM)method is adopted to provide visually explainable information.A visualizing evaluation index is proposed to quantitatively analyze visual representations;this index reflects a rough estimate of the concordance rate between the visualized heat map and intended defects.In addition,an ablation study,adopting three-branch combinations with the VGG16,is implemented to identify perfor-mance variations by visualizing predicted results.Experiments reveal that the proposed model,combined with hybrid pooling,batch normalization,and multi-attention modules,achieves the best performance with an accuracy of 97.77%,corresponding to an improvement of 2.49%compared with the baseline model.Consequently,this study demonstrates that reliable results from an automatic defect classification model can be provided to an inspector through the visual representation of the predicted results using CNN models.

On fine-grained visual explanation in convolutional neural networks

Existing explanation methods for Convolutional Neural Networks(CNNs)lack the pixel-level visualization explanations to generate the reliable fine-grained decision features.Since there are inconsistencies between the explanation and the actual behavior of the model to be interpreted,we propose a Fine-Grained Visual Explanation for CNN,namely F-GVE,which produces a fine-grained explanation with higher consistency to the decision of the original model.The exact backward class-specific gradients with respect to the input image is obtained to highlight the object-related pixels the model used to make prediction.In addition,for better visualization and less noise,F-GVE selects an appropriate threshold to filter the gradient during the calculation and the explanation map is obtained by element-wise multiplying the gradient and the input image to show fine-grained classification decision features.Experimental results demonstrate that F-GVE has good visual performances and highlights the importance of fine-grained decision features.Moreover,the faithfulness of the explanation in this paper is high and it is effective and practical on troubleshooting and debugging detection.

Gas liquid cylindrical cyclone flow regime identification using machine learning combined with experimental mechanism explanation

The flow regimes of GLCC with horizon inlet and a vertical pipe are investigated in experiments,and the velocities and pressure drops data labeled by the corresponding flow regimes are collected.Combined with the flow regimes data of other GLCC positions from other literatures in existence,the gas and liquid superficial velocities and pressure drops are used as the input of the machine learning algorithms respectively which are applied to identify the flow regimes.The choosing of input data types takes the availability of data for practical industry fields into consideration,and the twelve machine learning algorithms are chosen from the classical and popular algorithms in the area of classification,including the typical ensemble models,SVM,KNN,Bayesian Model and MLP.The results of flow regimes identification show that gas and liquid superficial velocities are the ideal type of input data for the flow regimes identification by machine learning.Most of the ensemble models can identify the flow regimes of GLCC by gas and liquid velocities with the accuracy of 0.99 and more.For the pressure drops as the input of each algorithm,it is not the suitable as gas and liquid velocities,and only XGBoost and Bagging Tree can identify the GLCC flow regimes accurately.The success and confusion of each algorithm are analyzed and explained based on the experimental phenomena of flow regimes evolution processes,the flow regimes map,and the principles of algorithms.The applicability and feasibility of each algorithm according to different types of data for GLCC flow regimes identification are proposed.

专家系统工具C-ADVISOR的解释系统EXPLANATION

为了提高C—ADVISOR的实用性,我们设计了一个工具解释系统EXPLANATION。该解释系统采用预制文本法(prepared text)和执行追踪法(execution traces)对用户的提问给予解释,具有动态解释和静态解释两种功能。系统给出一种用户模型,能够针对用户的知识水平作出相应的解释。系统采用紧缩存贮技术,节省了大量的磁盘存贮空间。系统还设置一个数据库生成程序,为解释系统的建立、修改和维护提供方便的条件。

Literal Translation plus Explanation in Translating Political Words

Political words are closely related to the current situations and affairs of China, which is often dependent on China's culture and history. Therefore it is not uncommon that there are some words with background knowledge known to the source language readers, but unknown to the target language readers. Literal translation plus explanation is particularly useful in translating these words with particular background and cultural connotations.

Investigation of feature contribution to shield tunneling-induced settlement using Shapley additive explanations method

Accurate prediction of shield tunneling-induced settlement is a complex problem that requires consideration of many influential parameters.Recent studies reveal that machine learning(ML)algorithms can predict the settlement caused by tunneling.However,well-performing ML models are usually less interpretable.Irrelevant input features decrease the performance and interpretability of an ML model.Nonetheless,feature selection,a critical step in the ML pipeline,is usually ignored in most studies that focused on predicting tunneling-induced settlement.This study applies four techniques,i.e.Pearson correlation method,sequential forward selection(SFS),sequential backward selection(SBS)and Boruta algorithm,to investigate the effect of feature selection on the model’s performance when predicting the tunneling-induced maximum surface settlement(S_(max)).The data set used in this study was compiled from two metro tunnel projects excavated in Hangzhou,China using earth pressure balance(EPB)shields and consists of 14 input features and a single output(i.e.S_(max)).The ML model that is trained on features selected from the Boruta algorithm demonstrates the best performance in both the training and testing phases.The relevant features chosen from the Boruta algorithm further indicate that tunneling-induced settlement is affected by parameters related to tunnel geometry,geological conditions and shield operation.The recently proposed Shapley additive explanations(SHAP)method explores how the input features contribute to the output of a complex ML model.It is observed that the larger settlements are induced during shield tunneling in silty clay.Moreover,the SHAP analysis reveals that the low magnitudes of face pressure at the top of the shield increase the model’s output。

Analysis on Explanation Effect of the European Numerical Prediction on Temperature

[Objective] The research aimed to analyze explanation effect of the European numerical prediction on temperature. [Method] Based on CMSVM regression method, by using 850 hPa grid point data of the European numerical prediction from 2003 to 2009 and actual data of the maximum and minimum temperatures at 8 automatic stations in Qingyang City, prediction model of the temperature was established, and running effect of the business from 2008 to 2010 was tested and evaluated. [Result] The method had very good guidance role in real-time business running of the temperature prediction. Test and evaluation found that as forecast time prolonged, prediction accuracies of the maximum and minimum temperatures declined. When temperature anomaly was higher (actual temperature was higher than historical mean), prediction accuracy increased. Influence of the European numerical prediction was bigger. [Conclusion] Compared with other methods, operation of the prediction method was convenient, modeling was automatic, running time was short, system was stable, and prediction accuracy was high. It was suitable for implementing of the explanation work for numerical prediction product at meteorological station.

Theoretical Explanation and Improvement to the Flare Model of Lithography Based on the Kirk Test

The Kirk test has good precision for measuring stray light in optical lithography and is the usual method of measuring stray light.However,Kirk did not provide a theoretical explanation to his simulation model.We attempt to give Kirk's model a kind of theoretical explanation and a little improvement based on the model of point spread function of scattering and the theory of statistical optics.It is indicated by simulation that the improved model fits Kirk's measurement data better.

海登·怀特在《元史学》中混用interpretation与explanation的动机探究

有证据表明,海登?怀特的《元史学》导论(1973年11月)极有可能是根据《历史中的解释》一文(1973年初)修订而成。在《历史中的解释》中,怀特比较严格地区分了interpretation与explanation这两个概念,但在《元史学》导论中,他却混用二者,并在叙事层面偏好使用explanation。通过这一有意为之的举动,怀特消解了德国历史诠释学中interpretation与explanation的对立,拉低了interpretation的理论高度,清除了interpretation隐含的规训意味,同时赋予了explanation某种整体性、非认知性,从而超越了德国历史诠释学和战后英美学界的历史解释问题争论。

The Roles of Low-level Jets in “21·7” Henan Extremely Persistent Heavy Rainfall Event

An extremely heavy rainfall event lasting from 17 to 22 July 2021 occurred in Henan Province of China, with accumulated precipitation of more than 1000 mm over a 6-day period that exceeded its mean annual precipitation. The present study examines the roles of persistent low-level jets(LLJs) in maintaining the precipitation using surface station observations and reanalysis datasets. The LLJs triggered strong ascending motions and carried moisture mainly from the outflow of Typhoon In-fa(2021). The varying directions of the LLJs well corresponded to the meridional shifts of the rainfall. The precipitation rate reached a maximum during 20-21 July as the LLJs strengthened and expanded vertically into double LLJs, including synoptic-weather-system-related LLJs(SLLJs) at 850–700 hPa and boundary-layer jets(BLJs)at ~950 hPa. The coupling of the SLLJ and BLJ provided strong mid-and low-level convergence on 20 July, whereas the SLLJ produced mid-level divergence at its entrance that coupled with low-level convergence at the terminus of the BLJ on21 July. The formation mechanisms of the two types of LLJs are further examined. The SLLJs and the low-pressure vortex(or inverted trough) varied synchronously as a whole and were affected by the southwestward movement of the WPSH in the rainiest period. The persistent large total pressure gradient force at low levels also maintained the strength of low-level geostrophic winds, thus sustaining the BLJs on the synoptic scale. The results based on a Du-Rotunno 1D model show that the Blackadar and Holton mechanisms jointly governed the BLJ dynamics on the diurnal scale.

A Probabilistic Rating Prediction and Explanation Inference Model for Recommender Systems

Collaborative Filtering(CF) is a leading approach to build recommender systems which has gained considerable development and popularity. A predominant approach to CF is rating prediction recommender algorithm, aiming to predict a user's rating for those items which were not rated yet by the user. However, with the increasing number of items and users, thedata is sparse.It is difficult to detectlatent closely relation among the items or users for predicting the user behaviors. In this paper,we enhance the rating prediction approach leading to substantial improvement of prediction accuracy by categorizing according to the genres of movies. Then the probabilities that users are interested in the genres are computed to integrate the prediction of each genre cluster. A novel probabilistic approach based on the sentiment analysis of the user reviews is also proposed to give intuitional explanations of why an item is recommended.To test the novel recommendation approach, a new corpus of user reviews on movies obtained from the Internet Movies Database(IMDB) has been generated. Experimental results show that the proposed framework is effective and achieves a better prediction performance.

Alternative kind of hydrogen atoms as a possible explanation for the latest puzzling observation of the 21 cm radio line from the early Universe

There is a puzzling astrophysical result concerning the latest observation of the absorption profile of the redshifted radio line 21 cm from the early Universe(as described in Bowman et al.). The amplitude of the profile was more than a factor of two greater than the largest predictions. This could mean that the primordial hydrogen gas was much cooler than expected. Some explanations in the literature suggested a possible cooling of baryons either by unspecified dark matter particles or by some exotic dark matter particles with a charge a million times smaller than the electron charge. Other explanations required an additional radio background. In the present paper, we entertain a possible different explanation for the above puzzling observational result: the explanation is based on the alternative kind of hydrogen atoms(AKHA),whose existence was previously demonstrated theoretically, as well as by the analysis of atomic experiments. Namely, the AKHA are expected to decouple from the cosmic microwave background(CMB) much earlier(in the course of the Universe expansion) than usual hydrogen atoms, so that the AKHA temperature is significantly lower than that of usual hydrogen atoms. This seems to lower the excitation(spin) temperature of the hyperfine doublet(responsible for the 21 cm line) sufficiently enough for explaining the above puzzling observational result. This possible explanation appears to be more specific and natural than the previous possible explanations. Further observational studies of the redshifted 21 cm radio line from the early Universe could help to verify which explanation is the most relevant.

Regulation on the Protection of Fossils Issued by the State Council of the People's Republic of China:Scientists' Researches Protected Legally—Further Explanation

With large-scale engineering projects being carried out in China, a large number of fossil localities have been discovered and excavated by responsible agencies, but still some important fossils of great value have been removed and smuggled into foreign countries. In the last three years, more than 1345 fossil specimens have been intercepted by Customs in Shenzhen, Shanghai, Tianjin, Beijing and elsewhere, and more than 5000 fossils, most of which are listed as key fossils,

Henderson Island’s Plastic: An Explanation

Plastic has been accumulating on the beaches of Henderson Island in vast quantities according to recent news reports. It is proposed that the plastic is brought to the island by a very broad permanent surface current flowing southeastward past the island. Other characteristics of the flow are that its temperature is relatively high, its depth is shallow (about 100 m), its speed is sluggish (10 - 20 cm/sec), and by broad is meant more than 5000 km along 28 S. Henderson Island is located at the east/west midpoint of this wide warm current (130 W). By knowing more definitely where the plastic is coming from, than the vague suggestions provided by the news sources, it may be possible in the future to slow down or stop the piling up of trash on what were pristine beaches of this World Heritage Site.

A theoretical explanation of the decays of Majorana oscillations

Majorana zero modes in the hybrid semiconductor-superconductornanowire is one of the promising candidates for topologicalquantum computing. Recently, in nanowires with a superconductingisland, the signature of Majorana zero modescan be revealed as a subgap state whose energy oscillatesaround zero in magnetic field. This oscillation was interpretedas overlapping Majoranas. However, the oscillation amplitudeeither dies away after an overshoot or decays, sharply oppositeto the theoretically predicted enhanced oscillations for Majoranabound states, as the magnetic field increases. Several theoreticalstudies have tried to address this discrepancy, but arepartially successful. This discrepancy has raised the concernson the conclusive identification of Majorana bound states, andhas even endangered the scheme of Majorana qubits basedon the nanowires.

A novel method for gamma spectrum analysis of low-level and intermediate-level radioactive waste

The uncertainty of nuclide libraries in the analysis of the gamma spectra of low-and intermediate-level radioactive waste(LILW)using existing methods produces unstable results.To address this problem,a novel spectral analysis method is proposed in this study.In this method,overlapping peaks are located using a continuous wavelet transform.An improved quadratic convolution method is proposed to calculate the widths of the peaks and establish a fourth-order filter model to estimate the Compton edge baseline with the overlapping peaks.Combined with the adaptive sensitive nonlinear iterative peak,this method can effectively subtracts the background.Finally,a function describing the peak shape as a filter is used to deconvolve the energy spectrum to achieve accurate qualitative and quantitative analyses of the nuclide without the aid of a nuclide library.Gamma spectrum acquisition experiments for standard point sources of Cs-137 and Eu-152,a segmented gamma scanning experiment for a 200 L standard drum,and a Monte Carlo simulation experiment for triple overlapping peaks using the closest energy of three typical LILW nuclides(Sb-125,Sb-124,and Cs-134)are conducted.The results of the experiments indicate that(1)the novel method and gamma vision(GV)with an accurate nuclide library have the same spectral analysis capability,and the peak area calculation error is less than 4%;(2)compared with the GV,the analysis results of the novel method are more stable;(3)the novel method can be applied to the activity measurement of LILW,and the error of the activity reconstruction at the equivalent radius is 2.4%;and(4)The proposed novel method can quantitatively analyze all nuclides in LILW without a nuclide library.This novel method can improve the accuracy and precision of LILW measurements,provide key technical support for the reasonable disposal of LILW,and ensure the safety of humans and the environment.

Axial length shortening in myopic children with Stickler syndrome after repeated low-level red-light therapy

AIM:To report the myopia-controlling effect of repeated low-level red-light(RLRL)therapy in patients with Stickler syndrome(STL),an inherited collagenic disease typically presenting with early onset myopia.METHODS:Three STL children,aged 3,7,and 11y,received RLRL therapy throughout the follow-up period of 17,3,and 6mo,respectively after exclusion of fundus anomalies.Data on best-corrected visual acuity(BCVA),intraocular pressure,cycloplegic subjective refraction,ocular biometrics,scanning laser ophthalmoscope,optical coherence tomography,genetic testing,systemic disease history,and family history were recorded.RESULTS:At the initiation of the RLRL therapy,the spherical equivalent(SE)of 6 eyes from 3 patients ranged from-3.75 to-20.38 D,axial length(AL)were from 23.88 to 30.68 mm,and BCVA were from 0.4 to 1.0(decimal notation).Myopia progression of all six eyes slowed down after RLRL therapy.AL in five out of the six eyes shortened-0.07 to-0.63 mm.No side effects were observed.CONCLUSION:Three cases of STL whose progression of myopic shift and AL elongation are successfully reduced and even reversed after RLRL therapy.

Explanation vs Performance in Data Mining: A Case Study with Predicting Runaway Projects

Often, the explanatory power of a learned model must be traded off against model performance. In the case of predict-ing runaway software projects, we show that the twin goals of high performance and good explanatory power are achievable after applying a variety of data mining techniques (discrimination, feature subset selection, rule covering algorithms). This result is a new high water mark in predicting runaway projects. Measured in terms of precision, this new model is as good as can be expected for our data. Other methods might out-perform our result (e.g. by generating a smaller, more explainable model) but no other method could out-perform the precision of our learned model.

THEORETICAL EXPLANATION OF SPIN REORIENTATION IN DyTiFe_(11) COMPOUND

DyTiFe_(11) compound is a ferromagnetic substance.It has tetragonal body-centered ThMn_(12)-type crystallographic structure.At room temperature,the easy magnetization direction is the c-axis.A spin reorientation begins to appear at about 175K.The contribution of Fe sublattice to magnetocrystalline anisotropy was determined by experiments and that of Dy sublattice was obtained by using single ion model calculation.Results show that the spin reorientation arises from the competition of anisotropy between Fe and Dy sublattices.