def __init__(self, frame):
self.frame = frame
self.model = nets.NetEditor.newNet(100, 10, Neurons.Sigmoid(),
nets.Net.leastSquaredDerivative,
(Neurons.Sigmoid(), 55))
self.buttons = list()
self.font = pygame.font.SysFont("arial", 14)
def __init__(self, n, weight_lenght, type):
neurons = []
for i in range(n):
weight = []
bias = random.uniform(-2, 2)
for w in range(weight_lenght):
weight.append(random.uniform(-2, 2))
if type == "sigmoid":
neuron = Neurons.Sigmoid(weight, bias)
else:
neuron = Neurons.Perceptron(weight, bias)
neurons.append(neuron)
self.neurons = neurons
self.next_layer = None
self.previous_layer = None
def createSimple(numInputs, numOutputs):
synapse = Synapses.createRootSynapse()
seed = Neurons.createRootNeuron([synapse], [synapse])
inputSynapses = [Synapses.createRootSynapse() for i in range(numInputs)]
outputSynapses = [Synapses.createRootSynapse() for i in range(numOutputs)]
"Connect seed to inputs and outputs."
inputNeurons = [InputNeuron(None, [], [inputSynapses[i]], zeros(Neurons.inputDivisionDataSize), \
inputTree()) for i in range(numInputs)]
outputNeurons = [OutputNeuron(None, [outputSynapses[i]], [], zeros(Neurons.divisionDataSize)) \
for i in range(numOutputs)]
inputNeuron = inputNeurons[-1]
outputNeuron = outputNeurons[-1]
inputSynapse = inputNeuron.getSynapse()
outputSynapse = outputNeuron.getSynapse()
"edit contents"
inputNeuron.data[Neurons.fireRateFunEnd - 1] = -2
inputNeuron.data[Neurons.fireRateFun] = 10
inputNeuron.data[Neurons.evolveRateFunEnd - 1] = 4
inputNeuron.data[Neurons.sensitivityFunEnd - 1] = 1
inputNeuron.data[Neurons.inputEvolveTransform + Neurons.dataSize + Neurons.inputSize] = -2
inputNeuron.data[Neurons.inputEvolveTransform] = 4
inputNeuron.data[Neurons.inputEvolveTransform + \
Neurons.dataSize * (Neurons.dataSize + 2)] = 0.2
seed.data[Neurons.fireRateFunEnd - 1] = -2
seed.data[Neurons.fireRateFun] = 10
seed.data[Neurons.evolveRateFunEnd - 1] = 4
seed.data[Neurons.sensitivityFunEnd - 1] = 1
seed.data[Neurons.evolveTransform + Neurons.dataSize] = -2
seed.data[Neurons.evolveTransform + \
Neurons.dataSize * (Neurons.dataSize + 1)] = 0.2
outputNeuron.data[Neurons.fireRateFunEnd - 1] = -2
outputNeuron.data[Neurons.fireRateFun] = 10
outputNeuron.data[Neurons.evolveRateFunEnd - 1] = 4
outputNeuron.data[Neurons.sensitivityFunEnd - 1] = 1
outputNeuron.data[Neurons.evolveTransform + Neurons.dataSize] = -2
outputNeuron.data[Neurons.evolveTransform + \
Neurons.dataSize * (Neurons.dataSize + 1)] = 0.2
inputSynapse.data[Synapses.weightFunEnd - 1] = 1
inputSynapse.data[Synapses.activationFunEnd - 1] = 1
outputSynapse.data[Synapses.weightFunEnd - 1] = 1
outputSynapse.data[Synapses.activationFunEnd - 1] = 1
"done editing contents"
brain = Brain(seed, numInputs, numOutputs, inputNeurons, outputNeurons)
brain._startTime()
return brain
def build(self):
# check if we have the legal dimension
if self.x == 0:
print('ERROR! You have to assign a cluster size.')
# 1-dimensional
elif self.y == 0:
for i in range(self.x):
self.cluster.append(Neurons.LIF())
# check whether the cluster size is correct
if len(self.cluster) != self.x:
print('ERROR! Size of cluster is not correct.')
self.dimension = 1
# 2-dimensional
else:
for i in range(self.x):
self.cluster.append([])
for j in range(self.y):
self.cluster[i].append(Neurons.LIF())
# check whether the cluster size is correct
if len(self.cluster) * len(self.cluster[0]) != self.x * self.y:
print('ERROR! Size of cluster is not correct.')
self.dimension = 2
def createEmpty(numInputs, numOutputs):
"Create isolated seed."
synapse = Synapses.createRootSynapse()
seed = Neurons.createRootNeuron([synapse], [synapse])
inputSynapses = [Synapses.createRootSynapse() for i in range(numInputs)]
outputSynapses = [Synapses.createRootSynapse() for i in range(numOutputs)]
"Connect seed to inputs and outputs."
inputNeurons = [InputNeuron(None, [], [inputSynapses[i]], zeros(Neurons.inputDivisionDataSize), \
inputTree()) for i in range(numInputs)]
outputNeurons = [OutputNeuron(None, [outputSynapses[i]], [], zeros(Neurons.divisionDataSize)) \
for i in range(numOutputs)]
"Make and start brain."
brain = Brain(seed, numInputs, numOutputs, inputNeurons, outputNeurons)
brain._startTime()
return brain
def __init__(self, previous_layer_size=1, size=1, values=[]):
self.size = size
self.previous_layer_size = previous_layer_size
self.biases = []
self.values = []
self.neurons = []
self.z = []
self.weights = []
self.error = []
self.grad = {"W": None, "b": None}
if len(values) != size and len(values) != 0:
print("Input size mismatch")
for i in range(self.size):
self.neurons.append(
Neurons.Neuron(previous_layer_size=self.previous_layer_size))
if len(values) == size:
self.neurons[i].value = values[i]
if self.previous_layer_size != 0:
self.initialize_parameters()
def populate_grid(self):
self.currentcount = 1
self.dictsize = self.xmax * self.ymax * self.zmax
self.neuraldictionary = {}
self.current_listpos = 1
self.current_xpos = 1
self.current_ypos = 1
self.current_zpos = 1
for neuron in range(1,(self.dictsize + 1)):
self.neuraldictionary.update({(self.current_xpos, self.current_ypos, self.current_zpos):
neur.DigiNeuron(x=self.current_xpos,y=self.current_ypos,
z=self.current_zpos)})
if self.current_ypos > self.ymax:
self.current_xpos = 1
self.current_ypos = 1
self.current_zpos = self.current_zpos + 1
elif self.current_xpos > self.xmax:
self.current_xpos = 1
self.current_ypos = self.current_ypos + 1
elif self.current_xpos < self.xmax:
self.current_xpos = self.current_xpos + 1
def testSoftmaxFire(self):
neuron = n.SoftMax()
arr = neuron.fire(self.arr)
self.assertEquals(len(arr), len(self.arr))
# @Email : [email protected] # @GitHub : https://github.com/hello-roderickwang import Neurons import Synapse import Cluster import numpy as np import matplotlib.pyplot as plt if __name__ == '__main__': x = np.linspace(-np.pi, np.pi, 201) sin = (np.sin(x)+1)/2 a = Neurons.LIF() b = Neurons.LIF() weight = 2 connection = Synapse.Synapse(a, b) connection.set_weight(weight) v_a = [] v_b = [] for step in range(0, len(sin)): a.stimulate(sin[step]) a.go() v_a.append(a.out) connection.go()
def onClick(self):
layer = nets.Layer(Neurons.Sigmoid(), 3, nets.Layer.HIDDEN)
nets.NetEditor.spliceIn(layer, self.upLayer, self.downLayer)
