def make_network(game, config, residuals=5, name="Residual CNN"):
"""
Make a full network.
Game is passed in to get the columns and rows.
"""
net = cx.Network(name)
net.add(cx.Layer("main_input", (game.v, game.h, 2),
colormap="Greys", minmax=(0,1)))
out_layer = add_conv_block(net, "main_input")
for i in range(residuals):
out_layer = add_residual_block(net, out_layer)
add_policy_block(net, out_layer)
add_value_block(net, out_layer)
net.compile(loss={'value_head': 'mean_squared_error',
'policy_head': softmax_cross_entropy_with_logits},
optimizer="sgd",
lr=config.LEARNING_RATE,
momentum=config.LEARNING_RATE,
loss_weights={'value_head': 0.5,
'policy_head': 0.5})
for layer in net.layers:
if layer.kind() == "hidden":
layer.visible = False
return net
import conx as cx
net = cx.Network("XOR2")
net.add(cx.Layer("input1", 2))
net.add(cx.Layer("input2", 2))
net.add(cx.Layer("hidden1", 2, activation="sigmoid"))
net.add(cx.Layer("hidden2", 2, activation="sigmoid"))
net.add(cx.Layer("shared-hidden", 2, activation="sigmoid"))
net.add(cx.Layer("output1", 2, activation="sigmoid"))
net.add(cx.Layer("output2", 2, activation="sigmoid"))
net.connect("input1", "hidden1")
net.connect("input2", "hidden2")
net.connect("hidden1", "shared-hidden")
net.connect("hidden2", "shared-hidden")
net.connect("shared-hidden", "output1")
net.connect("shared-hidden", "output2")
net.compile(loss='mean_squared_error',
optimizer=cx.SGD(lr=0.3, momentum=0.9))
ds = [
([[0, 0],[0, 0]], [[0, 0],[0, 0]], ["one", "one"]),
([[0, 0],[1, 1]], [[1, 1],[1, 1]], ["two", "two"]),
([[1, 1],[0, 0]], [[1, 1],[1, 1]], ["three", "three"]),
([[1, 1],[1, 1]], [[0, 0],[0, 0]], ["four", "four"])
]
net.dataset.load(ds)
net.train(2000, report_rate=10, accuracy=1)
net.test()
#
# Erstellung und Visualisierung eines Modelles mit ConX
#
import conx as cx
dataset = [[[0, 0], [0]], [[0, 1], [1]], [[1, 0], [1]], [[1, 1], [0]]]
net = cx.Network("XOR", 2, 5, 1, activation="sigmoid")
net.dataset.load(dataset)
net.compile(error='mean_squared_error', optimizer="sgd", lr=0.3, momentum=0.9)
net.train(2000, report_rate=10, accuracy=1.0)
net.picture()
import conx as cx
ds = [[[0, 0], [0, 0.5]], [[0, 1], [1, 0.5]], [[1, 0], [1, 0.5]],
[[1, 1], [0, 0.5]]]
net = cx.Network("XOR")
net.add(cx.Layer("input", 2))
net.add(cx.Layer("hidden1", 3, activation="relu"))
net.add(cx.Layer("hidden2", 4, activation="relu"))
net.add(cx.Layer("output", 2, activation="softmax"))
net.connect("input", "hidden1")
net.connect("hidden1", "hidden2")
net.connect("hidden2", "output")
net.compile(error='mean_squared_error', optimizer=cx.SGD(lr=0.3, momentum=0.9))
# NOTE:
# net = Network("XOR", 2, 3, 4, 1, activation="sigmoid")
# is the same as:
# net = Network("XOR")
# net.add(Layer("input", shape=2))
# net.add(Layer("hidden1", shape=3, activation="sigmoid"))
# net.add(Layer("hidden2", shape=4, activation="sigmoid"))
# net.add(Layer("output", shape=1, activation="sigmoid"))
# net.connect("input", "hidden1")
# net.connect("hidden1", "hidden2")
# net.connect("hidden2", "output")
net.dataset.load(ds)
net.train(2000, report_rate=10, accuracy=1)
net.test()
while True:
print('image %d' % k)
plt.imshow(images[k], cmap=cmap, interpolation=interpolation)
plt.draw()
k += 1
answer = raw_input('RETURN to continue, q to quit...')
if answer == 'q': break
# to save:
#np.savez_compressed('mydataset.npz', images=images, labels=labels)
plt.ion()
#---------------------------------------------------------------------------
net = cx.Network("MNIST")
net.add(cx.Layer("input", shape=784, vshape=(28, 28)))
net.add(
cx.Layer("hidden1",
shape=512,
vshape=(16, 32),
activation='relu',
dropout=0.2))
net.add(
cx.Layer("hidden2",
shape=512,
vshape=(16, 32),
activation='relu',
dropout=0.2))
net.add(cx.Layer("output", shape=10, activation='softmax'))
import conx as cx
net = cx.Network("Image")
net.add(cx.ImageLayer("input", (28, 28), 1))
net.add(cx.Conv2DLayer("conv1", 10, (5, 5), activation="relu"))
net.add(cx.Conv2DLayer("conv2", 10, (5, 5), activation="relu"))
net.add(cx.MaxPool2DLayer("pool1", pool_size=(2, 2)))
net.add(cx.FlattenLayer("flatten"))
net.add(cx.Layer("hidden1", 20, activation="relu"))
net.add(cx.Layer("output", 10, activation="softmax"))
net.connect()
net.compile(error='categorical_crossentropy', optimizer="adam")
net.dataset.get("mnist")
# NOTE:
# net = Network("XOR", 2, 3, 4, 1, activation="sigmoid")
# is the same as:
# net = Network("XOR")
# net.add(Layer("input", shape=2))
# net.add(Layer("hidden1", shape=3, activation="sigmoid"))
# net.add(Layer("hidden2", shape=4, activation="sigmoid"))
# net.add(Layer("output", shape=1, activation="sigmoid"))
# net.connect("input", "hidden1")
# net.connect("hidden1", "hidden2")
# net.connect("hidden2", "output")
#net.dataset.load(ds)
#net.train(2000, report_rate=10, accuracy=1)
import conx as cx
net = cx.Network("Add")
net.debug = True
net.add(cx.Layer("one", 1))
net.add(cx.Layer("two", 1))
net.add(cx.AddLayer("add"))
net.add(cx.Layer("hidden", 2))
net.add(cx.Layer("output", 1))
net.connect("one", "add")
net.connect("two", "add")
net.connect("add", "hidden")
net.connect("hidden", "output")
net.compile(error="mse", optimizer="adam")
