def get_trainer(trainer_name, net, ds, batchlearning):
if trainer_name == "bp":
return BackpropTrainer(net,
ds,
batchlearning=batchlearning,
verbose=True)
elif trainer_name == "dl":
return DeepBeliefTrainer(net, ds)
net.addConnection(FullConnection(bias, h1))
net.addConnection(FullConnection(bias, h2))
net.addConnection(FullConnection(bias, out))
net.sortModules()
return net
if __name__ == "__main__":
import GwData
data = GwData.GwData()
xs = get_binary_data(data)
ys = data.labels_for("50")
sdataset = SupervisedDataSet(xs.shape[1], 1)
udataset = UnsupervisedDataSet(xs.shape[1])
for i, x in enumerate(xs):
sdataset.addSample(x, ys[i])
udataset.addSample(x)
epochs = 100
layerDims = [xs.shape[1], 300, 100, 2]
#net = buildNetwork(*layerDims)
net = custom_build_network(layerDims)
trainer = DeepBeliefTrainer(net, dataset=udataset)
#trainer = DeepBeliefTrainer(net, dataset=sdataset)
trainer.trainEpochs(epochs)
tf = open('novelty_plant.txt','r')
first = True
for line in tf.readlines():
if (not first):
data = [float(x) for x in line.strip().split('\t') if x != '']
# indata = tuple(data[:6])
# outdata = tuple(data[6:])
ds.addSample(data)
first = False
n = buildNetwork(ds.dim,8,8,1,recurrent=True)
t = DeepBeliefTrainer(n,ds, epochs=50)
t.trainEpochs(1)
t.testOnData(ds, verbose= True)
ds.addSample((0, 0), (0,))
ds.addSample((0, 1), (1,))
ds.addSample((1, 0), (1,))
ds.addSample((1, 1), (0,))
for input, target in ds:
print(input, target)
#net = buildNetwork(2, 3, 1, bias=True, hiddenclass=TanhLayer)#1000
# net = buildNetwork(2, 6, 1, bias=True) # 3000
net = buildNetwork(2, 3, 1, bias=True)
__author__ = 'Justin S Bayer, [email protected]' __version__ = '$Id$' import scipy from pybrain.datasets import UnsupervisedDataSet from pybrain.unsupervised.trainers.deepbelief import DeepBeliefTrainer from pybrain.tools.shortcuts import buildNetwork from pybrainexamples.datasets.mnist import makeMnistDataSets net = buildNetwork(784, 500, 500, 2000, bias=True) train, test = makeMnistDataSets('/Users/bayerj/Desktop/MNIST/') trainer = DeepBeliefTrainer(net, train) trainer.train() print "RBM Phase finished. Now backprop." softmaxer = SoftmaxLayer(10) con = FullConnection(net.outmodules[0], softmaxer) net.addModule(softmaxer) net.outmodules = [softmaxer] trainer = BackpropTrainer(trainer, ds) for i in xrange(sys.maxint): error = trainer.train() print "%i: %.2f" % (i, error)
#! /usr/bin/env python2.5 # -*- coding: utf-8 -*- # Miniscule deep belief net example __author__ = 'Justin S Bayer, [email protected]' __version__ = '$Id$' from pybrain.datasets import UnsupervisedDataSet from pybrain.unsupervised.trainers.deepbelief import DeepBeliefTrainer from pybrain.tools.shortcuts import buildNetwork ds = UnsupervisedDataSet(6) ds.addSample([0, 1] * 3) ds.addSample([1, 0] * 3) net = buildNetwork(6, 2, 2, 2, bias=True) params = net.params.copy() trainer = DeepBeliefTrainer(net, ds) trainer.train() print params == net.params
net.addConnection(FullConnection(bias, h1))
net.addConnection(FullConnection(bias, h2))
net.addConnection(FullConnection(bias, out))
net.sortModules()
return net
if __name__ == "__main__":
import GwData
data = GwData.GwData()
xs = get_binary_data(data)
ys = data.labels_for("50")
sdataset = SupervisedDataSet(xs.shape[1], 1)
udataset = UnsupervisedDataSet(xs.shape[1])
for i,x in enumerate(xs):
sdataset.addSample(x, ys[i])
udataset.addSample(x)
epochs = 100
layerDims = [xs.shape[1], 300, 100, 2]
#net = buildNetwork(*layerDims)
net = custom_build_network(layerDims)
trainer = DeepBeliefTrainer(net, dataset=udataset)
#trainer = DeepBeliefTrainer(net, dataset=sdataset)
trainer.trainEpochs(epochs)
