xN.setSize(x.width() - 6, x.height() - 6) xN.setBpp(24) xN.setPixelAll((0,0,0)) for i in xE: xN.setPixel(i[0] - 3,i[1] - 3,(255,255,255)) xN.saveImage('bmp4.bmp') xN = x.edgeDetectSobel() xN.setBpp(24) xN.saveImage('bmp5.bmp') """ x = feedforwardNeuralNetwork.feedforwardNeuralNetwork(3, [5, 3, 2], "NN1") c = 0 while True: print "New epoch %d" % (c) c += 1 x.learn([1, 1, 0, 0, 1], [1, 0], 0.1) x.learn([0, 1, 1, 1, 0], [0, 1], 0.1) x.learn([0, 1, 1, 1, 1], [1, 1], 0.1) x.learn([0, 1, 0, 0, 0], [0, 0], 0.1) print x.feed([1, 1, 0, 0, 1]) print x.feed([0, 1, 1, 1, 0]) print x.feed([0, 1, 1, 1, 1])
print i fpO.write('0.01\n') for j in range(784): fpO.write('%lf ' % (float(int(fpI.read(1).encode('hex'),16)) / 256)) fpO.write('\n') c = int(fpL.read(1).encode('hex'),16) for j in range(10): if(j == c): fpO.write('1 ') else: fpO.write('0 ') fpO.write('\n') fpI.close() fpL.close() fpO.close() ''' tic = time.clock() fNN = feedforwardNeuralNetwork.feedforwardNeuralNetwork(3,[784,300,10],'fNN') fNN.learn('training-set.txt') fNN.exportWeight('fNN-Weight1.txt') toc = time.clock() print toc - tic
tic = time.clock() fNN = feedforwardNeuralNetwork.feedforwardNeuralNetwork(4,[30,20,5,1],'fNN-30-20-5-1-lim','fNN-30-20-5-1-Weight50-lim.txt') toc = time.clock() print toc - tic fNN.learn('train_short_lim_fine.txt','',50) toc = time.clock() print toc - tic fNN.exportWeight('fNN-30-20-5-1-Weight100-lim.txt') toc = time.clock() print toc - tic ''' import feedforwardNeuralNetwork import time tic = time.clock() fNN = feedforwardNeuralNetwork.feedforwardNeuralNetwork(4,[30,20,5,1],'fNN-30-20-5-1','fNN-30-20-5-1-Weight100-lim.txt') toc = time.clock() print toc - tic fNN.feed('rawtest.txt','testout.txt') toc = time.clock() print toc - tic