def test_save(self):
net = Net.Network([2, 3, 4])
net.random()
os.chdir("..")
os.chdir("..")
net.save("networkSaveTest")
netLoad = Net.Network()
netLoad.load("networkSaveTest")
for l, lay in enumerate(net.layers):
for n, nd in enumerate(lay.nodes):
self.assertEqual(
nd.activation, netLoad.layers[l].nodes[n].activation,
"Value of a node should match after loading network")
self.assertEqual(
nd.zActivation, netLoad.layers[l].nodes[n].zActivation,
"Activation of a node should match after loading network")
self.assertEqual(
nd.bias, netLoad.layers[l].nodes[n].bias,
"Bias of a node should match after loading network")
for c, con in enumerate(nd.connections):
self.assertEqual(
con.weight,
netLoad.layers[l].nodes[n].connections[c].weight,
"Weight of a connection should match after loading network"
)
os.remove("saves/networkSaveTest.txt")
def test_init(self):
net = Net.Network()
self.assertTrue(
net.layers == [],
"With default network, layers should be empty list, was: " +
str(net.layers))
net = Net.Network([])
self.assertTrue(
net.layers == [],
"With empty list network, layers should be empty list,"
" was: " + str(net.layers))
net = Net.Network(None)
self.assertTrue(
net.layers == [],
"With none layers, layers should be empty list, was: " +
str(net.layers))
def test_feedInputs(self):
net = Net.Network([4, 2])
expect = [1, 2, 5, -3]
net.feedInputs(expect)
for i in range(4):
v = net.layers[0].nodes[i].activation
self.assertTrue(
v == expect[i], "Input node value should be " +
str(expect[i]) + ", was: " + str(v))
def test_getOutputs(self):
net = Net.Network([1, 4])
net.layers[-1].nodes[0].activation = 3
net.layers[-1].nodes[1].activation = -2
net.layers[-1].nodes[2].activation = 2
net.layers[-1].nodes[3].activation = 7
self.assertTrue(
net.getOutputs() == [3, -2, 2, 7],
"Outputs should be: [3, -2, 2, 7],"
" was: " + str(net.getOutputs()))
def getTestNet():
net = Net.Network([4, 6, 6, 1])
for i in range(len(net.layers)):
for j in range(len(net.layers[i].nodes)):
net.layers[i].nodes[j].bias = j * .1
if i > 0:
for j in range(len(net.layers[i].nodes)):
for k in range(len(net.layers[i - 1].nodes)):
net.layers[i].nodes[j].connections[k].weight = (i * .3 +
j) * .1
net.feedInputs([.1, .2, .3, .4])
net.calculate()
return net
def test_getText(self):
net = Net.Network([3, 4, 2])
s = net.getText()
expect = "Network:\n" \
"Layer:\n" \
"Node: value: 0, bias:0, \n" \
"Node: value: 0, bias:0, \n" \
"Node: value: 0, bias:0, \n" \
"Layer:\n" \
"Node: value: 0, bias:0, weight:0.0, weight:0.0, weight:0.0, \n" \
"Node: value: 0, bias:0, weight:0.0, weight:0.0, weight:0.0, \n" \
"Node: value: 0, bias:0, weight:0.0, weight:0.0, weight:0.0, \n" \
"Node: value: 0, bias:0, weight:0.0, weight:0.0, weight:0.0, \n" \
"Layer:\n" \
"Node: value: 0, bias:0, weight:0.0, weight:0.0, weight:0.0, weight:0.0, \n" \
"Node: value: 0, bias:0, weight:0.0, weight:0.0, weight:0.0, weight:0.0, \n"
self.assertTrue(
s == expect,
"Text from network should be:\n" + expect + "was:\n" + s)
def test_calculate(self):
layers = []
nodes = [Net.Node([]), Net.Node([])]
nodes[0].activation = 3.3
nodes[1].activation = -1.5
layers.append(Net.Layer(nodes))
nodes = [
Net.Node(bias=.2,
connections=[Net.Connection(1),
Net.Connection(-1)]),
Net.Node(bias=-1.3,
connections=[Net.Connection(.5),
Net.Connection(.2)])
]
layers.append(Net.Layer(nodes))
nodes = [
Net.Node(bias=-.8,
connections=[Net.Connection(1.5),
Net.Connection(.2)]),
Net.Node(bias=.45,
connections=[Net.Connection(4),
Net.Connection(3)])
]
layers.append(Net.Layer(nodes))
net = Net.Network(layers)
net.calculate()
expect = [0.688359339593154, 0.9974285761888321]
actual = net.getOutputs()
for i in range(2):
self.assertAlmostEqual(
expect[i], actual[i], 6, "Network output should be: " +
str(expect[i]) + " was: " + str(actual[i]))
def test_random(self):
net = Net.Network([2, 3, 4])
net.random()
for lay in net.layers:
TestLayer.testRandom(self, lay)