import network2
import Reader
import numpy
def list2vec(x):
# this converts a list to a vector that the neural network code likes.
return numpy.array([x]).transpose()
def listset2vectorset(listset):
# takes in a list of (feature, result) tuples.
# turns the feature/result list into feature/result ndarrays of size (n,1).
# Note that the outputs are not vectors of size (n,)
return [(list2vec(x), list2vec(y)) for x, y in listset]
def Vanilla(layers, trainingData, evaluationData):
net = network2.Network(layers, cost=network2.CrossEntropyCost)
net.SGD(trainingData, \
epochs=300, mini_batch_size=int(0.05*len(trainingData)), eta=0.01, lmbda=.5, \
evaluation_data=evaluationData, monitor_evaluation_accuracy=True, \
monitor_training_accuracy=True)
if __name__ == "__main__":
data = Reader.ReadVectorSet("../stocksim/nn/pre/msftvec.csv")
(training, evaluation) = List.Split(data)
Vanilla([6, 6, 1], \
listset2vectorset(training), listset2vectorset(evaluation))
