#coding=utf8
from?neuralnetworktool?import?*
import?random
import?traceback
class?Neuron（）:
	def?__init__（selfnInput?=?0studyspeed?=?1thresholdFunc?=?sigmoidX）:
		self.nInput?=?nInput
		#init?the?weight?in?range?-1?to?1
		randomweights?=?[random.uniform（-1?1）?for?i?in?range（nInput）]
		self.weights?=?randomweights
		self.thresholdFunc?=?thresholdFunc
		self.studyspeed?=?studyspeed
		self.inputs?=?[]

	def?setInputs（selfinputs）:
		self.inputs?=?inputs

	def?getInputs（self）:
		return?self.inputs

	def?getOutput（selfinputs）:
		self.inputs?=?inputs
		outputs?=?np.squeeze（np.asarray（self.weights?*?inputs））
		#for?i?in?range（len（self.weights））:
		#	outputs?+=?self.weights[i]?*?self.inputs[i]
		return?round（self.thresholdFunc（outputs）3）


	def?adjustWeights（selfinputsexceptOutputrealOutput?=?None）:
		if?realOutput?==?None:
			f?=?self.getOutput（inputs）
		else:
			f?=?realOutput
		d?=?exceptOutput
		r?=?（d?-?f）?*?f?*?（1?-?f）
		self.adjustWeightsWithR（r）
		return?r

	#R反映出網絡輸出的平方差對相應的sigmoid函數的輸入中的變化的敏感度
	def?adjustWeightsWithR（selfR）:
		self.weights?+=?self.studyspeed?*?R?*??np.transpose（self.inputs）

	def?getWeights（self）:
		return?np.squeeze（np.asarray（self.weights））

	def?getWeightsMatrix（self）:
		return?self.weights

	def?setWeights（selfweights）:
		self.weights?=?np.matrix（weights）

	def?setStudySpeed（selfstudyspeed）:
		self.studyspeed?=?studyspeed

	def?printInfo（self）:
		print?self.weights


class?NeuralNetWork（）:
	def?__init__（selfnInput=0nHide=0nOutput=0studyspeed=0）:
		self.nInput?=?nInput
		self.nHide?=?nHide
		self.nOutput?=?nOutput
		self.studyspeed?=?studyspeed
		self.inputNeurons?=?[0?for?i?in?range（nInput）]
		self.hideNeurons?=?[
		Neuron（nInput?=?nInputstudyspeed?=?studyspeed）?for?i?in?range（nHide）]
		self.outputNeurons?=?[
		Neuron（nInput?=?nHidestudyspeed?=?studyspeed）?for?i?in?range（nOutput）]


	def?backpropagation（selftrainings=[]examples=[]）:
		for?i?in?range（len（trainings））:
			self.backpropagationSingle（
				training?=?trainings[i]
				example?=?examples[i]
				）

	def?backpropagationSingle（selftrainingexample）:
		#d為訓練集目標輸出
		d?=?example
		#輸入轉化為單列矩陣
		self.inputNeurons?=?np.transpose（np.matrix（training））
		#用于保存隱藏層輸出
		hideoutputs?=?[]
		#對輸入層作為輸入，得到隱藏層輸出結果
		for?hideNeuron?in?self.hideNeurons:
			hideoutputs.append（hideNeuron.getOutput（self.inputNeurons））
		

		#將隱藏層結果作為輸出層的輸入，得到輸出層的結果
		outputlayeroutputs?=?[]
		hideOutPutMatrixs?=??np.transpose（np.matrix（hideoutputs））
		for?outputNeuron?in?self.outputNeurons:
			outputlayeroutputs.append（outputNeuron.getOutput（hideOutPutMatrixs））
		#輸出層結果與期望結果做比較，并且反向傳播調節
		#保存調節前輸出層權值
		beforeweights?=?[]

		#對輸出層進行調節
		#保存輸出層各個神經元的調節因子
		outputRs?=?[]
		for?i?in?range（self.nOutput）:
			#f為實際輸出結果
			f?=?outputlayeroutputs[i]
			outputR?=?（d[i]?-?f）?*?f?*?（1?-?f）
			outputRs.append（outputR）
			#保存第i個輸出神經元的調節前權值
			beforeweights.append（self.outputNeurons[i].getWeights（））
			#對每個輸出層神經元進行調節
			self.outputNeurons[i].adjustWeightsWithR（outputR）
		#對隱藏層進行調節
		for?i?in?range（s

?屬性????????????大小?????日期????時間???名稱
-----------?---------??----------?-----??----
?????目錄???????????0??2016-05-05?03:06??NeuralNetwork-master\
?????文件?????????378??2016-05-05?03:06??NeuralNetwork-master\.gitattributes
?????文件?????????655??2016-05-05?03:06??NeuralNetwork-master\.gitignore
?????文件?????????744??2016-05-05?03:06??NeuralNetwork-master\README.md
?????文件???????????0??2016-05-05?03:06??NeuralNetwork-master\TestConfig.txt
?????目錄???????????0??2016-05-05?03:06??NeuralNetwork-master\data\
?????目錄???????????0??2016-05-05?03:06??NeuralNetwork-master\data\faces\
?????文件?????????131??2016-05-05?03:06??NeuralNetwork-master\data\faces\.anonr
?????目錄???????????0??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\
?????文件???????42648??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_angry_open.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_angry_open_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_angry_open_4.pgm
?????文件???????42480??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_angry_sunglasses.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_angry_sunglasses_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_angry_sunglasses_4.pgm
?????文件???????42802??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_happy_open.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_happy_open_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_happy_open_4.pgm
?????文件???????42944??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_happy_sunglasses.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_happy_sunglasses_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_happy_sunglasses_4.pgm
?????文件???????43292??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_neutral_open.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_neutral_open_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_neutral_open_4.pgm
?????文件???????42974??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_neutral_sunglasses.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_neutral_sunglasses_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_neutral_sunglasses_4.pgm
?????文件???????42693??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_sad_open.pgm
?????文件????????3853??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_sad_open_2.pgm
?????文件?????????973??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_sad_open_4.pgm
?????文件???????42967??2016-05-05?03:06??NeuralNetwork-master\data\faces\an2i\an2i_left_sad_sunglasses.pgm
............此處省略1896個文件信息