def test_simpleChain(self):
inp = [[1]]
out = [[0.412]]
n = mlp(sizes=[1, 1, 1],
weightGenerator=np.random.random,
biasGenerator=np.random.random,
activationFunction=LogSigmoid(1, -1))
train(net=n, inputs=inp, outputs=out,
numEpochs=200, learningRate=0.8)
assert_allclose([n.forward(i) for i in inp], out, atol=0.1)
def test_batchSimpleFork(self):
inp = [[1]]
out = [[0.412, 0.9]]
n = mlp(sizes=[1, 2],
weightGenerator=np.random.random,
biasGenerator=np.random.random,
activationFunction=LogSigmoid(-1, 1))
batchTrain(net=n,
inputs=inp,
outputs=out,
numEpochs=100,
learningRate=0.8)
assert_allclose([n.forward(i) for i in inp], out, atol=0.1)
def test_simpleFork(self):
inp = [[1]]
out = [[0.44, 0.77, 0.33]]
n = mlp(sizes=[len(inp[0]), len(out[0])],
weightGenerator=np.random.random,
biasGenerator=np.random.random,
activationFunction=LogSigmoid(0, 1))
train(net=n,
inputs=inp,
outputs=out,
numEpochs=100,
learningRate=0.7)
assert_allclose([n.forward(i) for i in inp], out, atol=0.1)
def test_xor(self):
inp = [[0, 0], [1, 0], [0, 1], [1, 1]]
out = [[0], [1], [1], [0]]
n = mlp(sizes=[2, 2, 1],
weightGenerator=np.random.random,
biasGenerator=np.random.random,
activationFunction=LogSigmoid(0, 1))
miniBatchTrain(net=n,
inputs=inp,
outputs=out,
numEpochs=4000,
learningRate=0.3,
momentum=0.8,
batchSize=4)
# pos = radial_tree_layout(n.g, n.g.vertex(0))
# graph_draw(n.g, pos=pos, vertex_text=n.g.vertex_index,
# vertex_font_size=18, output_size=(200, 200))
assert_allclose([n.forward(i) for i in inp], out, atol=0.1)
