With the expansion of the Internet market,the traditional software development method has been difficult to meet the market demand due to the problems of long development cycle,tedious work,and difficult system mainte...With the expansion of the Internet market,the traditional software development method has been difficult to meet the market demand due to the problems of long development cycle,tedious work,and difficult system maintenance.Therefore,to improve software development efficiency,this study uses residual networks and bidirectional long short-term memory(BLSTM)networks to improve the Pix2code model.The experiment results show that after improving the visual module of the Pix2code model using residual networks,the accuracy of the training set improves from 0.92 to 0.96,and the convergence time is shortened from 3 hours to 2 hours.After using a BLSTM network to improve the language module and decoding layer,the accuracy and convergence speed of the model have also been improved.The accuracy of the training set grew from 0.88 to 0.92,and the convergence time was shortened by 0.5 hours.However,models improved by BLSTM networks might exhibit overfitting,and thus this study uses Dropout and Xavier normal distribution to improve the memory network.The results validate that the training set accuracy of the optimized BLSTM network remains around 0.92,but the accuracy of the test set has improved to a maximum of 85%.Dropout and Xavier normal distributions can effectively improve the overfitting problem of BLSTM networks.Although they can also decrease the model’s stability,their gain is higher.The training and testing accuracy of the Pix2code improved by residual network and BLSTM network are 0.95 and 0.82,respectively,while the code generation accuracy of the original Pix2code is only 0.77.The above data indicate that the improved Pix2code model has improved the accuracy and stability of code automatic generation.展开更多
为了提高深度卷积神经网络(DCNN)的图像并行处理能力,提高其图像识别的准确率和运行效率,研究过程以MapReduce并行计算框架和从图像到矩阵(Image to Column,Im2col)算法,分别进行原始图像特征并行提取和筛选、模型并行训练和参数并行更...为了提高深度卷积神经网络(DCNN)的图像并行处理能力,提高其图像识别的准确率和运行效率,研究过程以MapReduce并行计算框架和从图像到矩阵(Image to Column,Im2col)算法,分别进行原始图像特征并行提取和筛选、模型并行训练和参数并行更新,构建了并行DCNN优化算法。在性能检测阶段,将全连接神经网络和基于特征图和并行计算熵的深度卷积神经网络算法作为对照组,对比TOP⁃1准确率、浮点运算量、损失函数振荡性、运算时长四项指标,结果显示,此次提出的并行DCNN优化算法性能最佳。展开更多
The dispersion of K\-2CO\-3 on \%γ\%\|Al\-2O\-3 and the adsorption performance of K\-2CO\-3/\%γ\%\|Al\-2O\-3 to SO\-2 are investigated.The results show that K\-2CO\-3 can disperse onto the surface of \%γ\%\|Al\-2O\...The dispersion of K\-2CO\-3 on \%γ\%\|Al\-2O\-3 and the adsorption performance of K\-2CO\-3/\%γ\%\|Al\-2O\-3 to SO\-2 are investigated.The results show that K\-2CO\-3 can disperse onto the surface of \%γ\%\|Al\-2O\-3 as a monolayer and the dispersion threshold is 0.31\[\%m\%(K\-2CO\-3)/\%m\%(\%γ\%\|Al\-2O\-3), \%m\%/g\], which is close to the theoretical value calculated by assuming a bidentate vertical dispersion model of CO\-2 on the \%γ\%\|Al\-2O\-3 surface . The SO\-2 adsorption\|capacity on K\-2CO\-3/\%γ\%\|Al\-2O\-3 sample increases with the K\-2CO\-3 loading and reaches an extremum at its threshold. The adsorbent conversion of K\-2CO\-3/\%γ\%\|Al\-2O\-3 at the threshold is up to 72%. When the loading is higher than the threshold, the SO\-2 adsorption capacity decreases at first, then increases again. This phenomenon is caused by the reaction between SO\-2 and the bulk phase of K\-2CO\-3 crystallites. The sample decreases with the loading, and the sample with \{0.10\}\[\%m\%(K\-2CO\-3)/\%m(γ\%\|Al\-2O\-3), \%m\%/g\] loading shows the highest regeneration percentage of 63%. Compared with Na\-2CO\-3/\%γ\%\|Al\-2O\-3, K\-2CO\-3/\%γ\%\|Al\-2O\-3 might have some advantages.展开更多
基金supported by National Natural Science Foundation of China(No.62062063)the Science and Technology Research Project of Jiangxi Provincial Department of Education,China(No.GJJ202310)the Jiangxi Provincial Natural Science Foundation,China(No.20224BAB202022).
文摘With the expansion of the Internet market,the traditional software development method has been difficult to meet the market demand due to the problems of long development cycle,tedious work,and difficult system maintenance.Therefore,to improve software development efficiency,this study uses residual networks and bidirectional long short-term memory(BLSTM)networks to improve the Pix2code model.The experiment results show that after improving the visual module of the Pix2code model using residual networks,the accuracy of the training set improves from 0.92 to 0.96,and the convergence time is shortened from 3 hours to 2 hours.After using a BLSTM network to improve the language module and decoding layer,the accuracy and convergence speed of the model have also been improved.The accuracy of the training set grew from 0.88 to 0.92,and the convergence time was shortened by 0.5 hours.However,models improved by BLSTM networks might exhibit overfitting,and thus this study uses Dropout and Xavier normal distribution to improve the memory network.The results validate that the training set accuracy of the optimized BLSTM network remains around 0.92,but the accuracy of the test set has improved to a maximum of 85%.Dropout and Xavier normal distributions can effectively improve the overfitting problem of BLSTM networks.Although they can also decrease the model’s stability,their gain is higher.The training and testing accuracy of the Pix2code improved by residual network and BLSTM network are 0.95 and 0.82,respectively,while the code generation accuracy of the original Pix2code is only 0.77.The above data indicate that the improved Pix2code model has improved the accuracy and stability of code automatic generation.
文摘为了提高深度卷积神经网络(DCNN)的图像并行处理能力,提高其图像识别的准确率和运行效率,研究过程以MapReduce并行计算框架和从图像到矩阵(Image to Column,Im2col)算法,分别进行原始图像特征并行提取和筛选、模型并行训练和参数并行更新,构建了并行DCNN优化算法。在性能检测阶段,将全连接神经网络和基于特征图和并行计算熵的深度卷积神经网络算法作为对照组,对比TOP⁃1准确率、浮点运算量、损失函数振荡性、运算时长四项指标,结果显示,此次提出的并行DCNN优化算法性能最佳。
文摘The dispersion of K\-2CO\-3 on \%γ\%\|Al\-2O\-3 and the adsorption performance of K\-2CO\-3/\%γ\%\|Al\-2O\-3 to SO\-2 are investigated.The results show that K\-2CO\-3 can disperse onto the surface of \%γ\%\|Al\-2O\-3 as a monolayer and the dispersion threshold is 0.31\[\%m\%(K\-2CO\-3)/\%m\%(\%γ\%\|Al\-2O\-3), \%m\%/g\], which is close to the theoretical value calculated by assuming a bidentate vertical dispersion model of CO\-2 on the \%γ\%\|Al\-2O\-3 surface . The SO\-2 adsorption\|capacity on K\-2CO\-3/\%γ\%\|Al\-2O\-3 sample increases with the K\-2CO\-3 loading and reaches an extremum at its threshold. The adsorbent conversion of K\-2CO\-3/\%γ\%\|Al\-2O\-3 at the threshold is up to 72%. When the loading is higher than the threshold, the SO\-2 adsorption capacity decreases at first, then increases again. This phenomenon is caused by the reaction between SO\-2 and the bulk phase of K\-2CO\-3 crystallites. The sample decreases with the loading, and the sample with \{0.10\}\[\%m\%(K\-2CO\-3)/\%m(γ\%\|Al\-2O\-3), \%m\%/g\] loading shows the highest regeneration percentage of 63%. Compared with Na\-2CO\-3/\%γ\%\|Al\-2O\-3, K\-2CO\-3/\%γ\%\|Al\-2O\-3 might have some advantages.
