英语翻译Artificial neural networks construction[ 5 ]We used the feed-forward multi-layer network for this work .The learning algorithm for our neural network is the widely used back-propagation algorithm.The fabric properties were the inputs to t
来源:学生作业帮助网 编辑:六六作业网 时间:2024/11/24 02:08:08
英语翻译Artificial neural networks construction[ 5 ]We used the feed-forward multi-layer network for this work .The learning algorithm for our neural network is the widely used back-propagation algorithm.The fabric properties were the inputs to t
英语翻译
Artificial neural networks construction[ 5 ]
We used the feed-forward multi-layer network for this work .The learning algorithm for our neural network is the widely used back-propagation algorithm.The fabric properties were the inputs to the ANN and the performance values of the fabrics were the outputs .The ANN was set to the training mode first ; in this mode ,the fabric prop erties were the training inputs and the actual fabric performance values were the training target .At first ,the ANN predicted the performance of the garments manufacture according to the fabric properties ,compared them to the target values ,and modified the next prediction procession on the basis of the errors .Repeated this prediction-modification cycle ,the training will be finished until the errors between predict values and target values meet the need.After being trained ,the ANN was set to the query mode ,in which it provided predicted garment manufacture performance values ( query output ) from fabric properties values (query input) .
The back-propagation procedure is :
(1) To initialize the unit weights to small random values.(2) To apply an inputvect or ( X) to the network.(3) To feed forward or propagate the inputvect or t o determine all unit outputs .
(4) To compare unit responses in the output layer ( Y) with the desired or target responses ( Y′)
where Y = f ∑n i =0 Wij x i j = 1 ,2 ,3 ,⋯,m (1)error f unction E :(公式略)
transmit function was adopted sigmoid function f ( x) :(公式略)
(5) To compute and propagate an error sensitivity measure backward (starting at the output layer )
through the network ,using this as the basis for weight correction.
(6) To minimize the overall error at each stage through unit weight adjustments :(公式略)
(7) Momentum learning is an improvement to the straight gradient descent.,The equation to update the weight becomes :(公式略)
(8) To output the result when calculation arrive the training cycles or accuracy of prediction.
Otherwise go back to step (2) .The BP artificial neural network structure is illust ratedin Fig.1.
Experimental methods
1 、Fabric performa nce inves tigation in the clothing manufacturing
A number of 64 fabrics have been collected from the apparel industry for the present study.These f abrics are mainly wool ,cot ton ,silk ,polyester ,cotton/ polyester blends ,wool / polyester blends of lightweights .The specification of the f abric group is shown in Table 1.
英语翻译Artificial neural networks construction[ 5 ]We used the feed-forward multi-layer network for this work .The learning algorithm for our neural network is the widely used back-propagation algorithm.The fabric properties were the inputs to t
人工神经网络建设[ 5 ]
我们使用前馈多层网络,这方面的工作.学习算法为我们的神经网络是广泛使用的反向传播算法.织物性能的投入,人工神经网络的性能和价值观念的织物的产出.人工神经网络建立的培训模式第一;在这种模式下,织物支柱erties的培训投入与实际织物性能值分别为培训目标.首先,人工神经网络预测的业绩成衣制造根据织物性能相比,他们的目标价值,并修改了游行,预测未来的基础上的错误.重复这一预测的修改周期,培训工作将完成,直到误差预测值和目标值满足需要.经过培训,建立人工神经网络的查询方式,其中服装提供预测制造的性能值(查询输出)由织物性能值(查询输入) .
反向传播的程序是:
( 1 )初始化的单位重量小的随机值.( 2 )要申请一个inputvect或(十)到网络上.( 3 )前馈或宣传inputvect或确定所有单位的产出.
( 4 )要比较单位答复的输出层( Y )的理想与目标或答复( Y表示)
其中Y = F的∑ni = 0 Wij xij = 1 ,2 ,3 ,⋯ ,男( 1 )误差函数电子邮件:(公式略)
传输功能是通过乙状结肠函数f ( x ) :(公式略)
( 5 )计算和传播错误的敏感性措施落后(从输出层)
通过网络,以此为基础的体重更正.
( 6 )为了尽量减少错误的整体在每个阶段通过单位重量的调整:(公式略)
( 7 )动量学习是一种改进的连续梯度下降.,方程来更新的重量会变成:(公式略)
( 8 )输出时的结果计算得出的训练周期或准确预测.
否则回到步骤( 2 ) .的BP人工神经网络的结构illust ratedin图.1 .
实验方法
1 ,织物performa只要inves tigation在服装制造业
一些64面料已收集到的服装行业本研究.这些f abrics主要是羊毛,摇篮吨,丝绸,涤纶,棉/涤纶混纺织物,毛/涤混纺的lightweights .该规范的F abric组列于表1 .
人工神经网络建设[ 5 ]
我们使用前馈多层网络,这方面的工作。学习算法为我们的神经网络是广泛使用的反向传播算法。织物性能的投入,人工神经网络的性能和价值观念的织物的产出。人工神经网络建立的培训模式第一;在这种模式下,织物支柱erties的培训投入与实际织物性能值分别为培训目标。首先,人工神经网络预测的业绩成衣制造根据织物性能相比,他们的目标价值,并修改了游行,预测未来的基础上的错误。重...
全部展开
人工神经网络建设[ 5 ]
我们使用前馈多层网络,这方面的工作。学习算法为我们的神经网络是广泛使用的反向传播算法。织物性能的投入,人工神经网络的性能和价值观念的织物的产出。人工神经网络建立的培训模式第一;在这种模式下,织物支柱erties的培训投入与实际织物性能值分别为培训目标。首先,人工神经网络预测的业绩成衣制造根据织物性能相比,他们的目标价值,并修改了游行,预测未来的基础上的错误。重复这一预测的修改周期,培训工作将完成,直到误差预测值和目标值满足需要。经过培训,建立人工神经网络的查询方式,其中服装提供预测制造的性能值(查询输出)由织物性能值(查询输入) 。
反向传播的程序是:
( 1 )初始化的单位重量小的随机值。 ( 2 )要申请一个inputvect或(十)到网络上。 ( 3 )前馈或宣传inputvect或确定所有单位的产出。
( 4 )要比较单位答复的输出层( Y )的理想与目标或答复( Y表示)
其中Y = F的∑ni = 0 Wij xij = 1 , 2 , 3 , ⋯ ,男( 1 )误差函数é
传输功能是通过乙状结肠函数f ( x )
( 5 )计算和传播错误的敏感性措施落后(从输出层)
通过网络,以此为基础的体重更正。
( 6 )为了尽量减少错误的整体在每个阶段通过单位重量调整
( 7 )动量学习是一种改进的连续梯度下降。 ,方程来更新的重量会变成:
( 8 )输出时的结果计算得出的训练周期或准确预测。
否则回到步骤( 2 ) 。的BP人工神经网络的结构illust ratedin图。 1 。
实验方法
1 ,织物performa只要inves tigation在服装制造业
一些64面料已收集到的服装行业本研究。这些f abrics主要是羊毛,摇篮吨,丝绸,涤纶,棉/涤纶混纺织物,毛/涤混纺的lightweights 。该规范的F abric组列于表1 。
收起