def trainData(setObj,classes,trainingset,class_bags,hidden_neurons=10, alpha=1, epochs=50000,dropout_percent=0.5):
numpy.random.seed(1)
#randomly select weights
last_mean_error = 1
# randomly initialize our weights with mean 0
synapse_0 = 2 * numpy.random.random((len(trainingset[0]), hidden_neurons)) - 1
#matrix with 16 rows(number of words) 10 colomns (middle layer nuerons)
synapse_1 = 2 * numpy.random.random((hidden_neurons, len(classes))) - 1
prev_synapse_0_weight_update = numpy.zeros_like(synapse_0)
prev_synapse_1_weight_update = numpy.zeros_like(synapse_1)
#contain all 0s
synapse_0_direction_count = numpy.zeros_like(synapse_0)
synapse_1_direction_count = numpy.zeros_like(synapse_1)
#contain all 0s
synapse_0_direction_count = numpy.zeros_like(synapse_0)
synapse_1_direction_count = numpy.zeros_like(synapse_1)
for j in iter(range(epochs + 1)):
layer_0 = numpy.array(trainingset)
layer_1 = sigmoid(numpy.dot(layer_0, synapse_0))
layer_2 = sigmoid(numpy.dot(layer_1, synapse_1))
layer_2_error = numpy.array(class_bags) - layer_2
#get the error
if numpy.mean(numpy.abs(layer_2_error)) < last_mean_error:
#print("delta after " + str(j) + " iterations:" + str(np.mean(np.abs(layer_2_error))))
last_mean_error = numpy.mean(numpy.abs(layer_2_error))
else:
#print("break:", np.mean(np.abs(layer_2_error)), ">", last_mean_error)
break
# in what direction is the target value?
# were we really sure? if so, don't change too much.
layer_2_delta = layer_2_error * derivative(layer_2)
# how much did each l1 value contribute to the l2 error (according to the weights)?
layer_1_error = layer_2_delta.dot(synapse_1.transpose())
# in what direction is the target l1?
# were we really sure? if so, don't change too much.
layer_1_delta = layer_1_error * derivative(layer_1)
synapse_1_weight_update = (layer_1.T.dot(layer_2_delta))
synapse_0_weight_update = (layer_0.T.dot(layer_1_delta))
if (j > 0):
synapse_0_direction_count += numpy.abs(
((synapse_0_weight_update > 0) + 0) - ((prev_synapse_0_weight_update > 0) + 0))
synapse_1_direction_count += numpy.abs(
((synapse_1_weight_update > 0) + 0) - ((prev_synapse_1_weight_update > 0) + 0))
synapse_1 += alpha * synapse_1_weight_update
synapse_0 += alpha * synapse_0_weight_update
prev_synapse_0_weight_update = synapse_0_weight_update
prev_synapse_1_weight_update = synapse_1_weight_update
synapse0=synapse_0.tolist()
synapse1=synapse_1.tolist()
return synapse0,synapse1
def think(sentence, layer, parent, synapse_0, synapse_1):
y = give_word_bag(sentence.lower(), (storeSynapes.read(layer, parent))[2])
x = y[0]
l0 = numpy.array(x)
# matrix multiplication of input and hidden layer
l1 = sigmoid(numpy.dot(x, synapse_0))
# output layer
l2 = sigmoid(numpy.dot(l1, synapse_1))
return l2