Exemple #1
0
    res = eval_sub(x)
    # we clear the memory of the recurrent layer between input
    # sequences because otherwise the network starts in an unknown
    # state.
    rn.clear()
    return res

test_sub = theano.function([x, y], cost.output)
def test(x, y):
    res = test_sub(x, y)
    # clear here too
    rn.clear()
    return res

train_sub = theano.function([], cost.output,
                            updates=get_updates(cost.params,cost.output,0.05),
                            givens={x: trainx, y: trainy})

def train():
    res = train_sub()
    rn.clear()
    return res

# Since we didn't do any training (yet) the net has poor performance
print "Test at start:", test(testx, testy)

# Now to do some training
for _ in range(100):
    train()

print "Test after 100:", test(testx, testy)
#%% Build ANN.
sx = theano.tensor.matrix('x')
sy = theano.tensor.matrix('y')

h = SimpleNode(sx, 2, 4)
h2 = SimpleNode(h, 4, 2)
out = SimpleNode(h2, 2, 1, nlin=pynnet.nlins.sigmoid)
cost = errors.mse(out, sy)

theano.config.blas.ldflags = ''
eval = theano.function([sx], out.output)
test = theano.function([sx, sy], cost.output)
train = theano.function([sx, sy],
                        cost.output,
                        updates=get_updates(cost.params, cost.output, 0.01))

print("Error at start:", test(X, Y))

for i in range(200000):
    train(X, Y)
print("Error after 200000:", test(X, Y))


def pfunc(x):
    return 1. - eval(x)


clp = probaproxy(pfunc)

plt.figure()
Exemple #3
0
from pynnet import *
from pynnet.nodes import errors
from pynnet.training import get_updates

sx = theano.tensor.matrix("x")
sy = theano.tensor.matrix("y")

# We initialize an MLP with one hidden layer of two units.
h = SimpleNode(sx, 2, 2)
out = SimpleNode(h, 2, 1)
cost = errors.mse(out, sy)

# We can build functions from expressions to use our network
eval = theano.function([sx], out.output)
test = theano.function([sx, sy], cost.output)
train = theano.function([sx, sy], cost.output, updates=get_updates(cost.params, cost.output, 0.01))


x = [[0, 0], [0, 1], [1, 0], [1, 1]]
y = [[0], [1], [1], [0]]

# At the start the error is terrible
print "Error at start:", test(x, y)

# So we train a little
for i in range(10):
    train(x, y)

# Now the error should be a bit better
print "Error after 10:", test(x, y)