def FoN(inputs, outputs):
seed(tmpObject.i)
tmpObject.i = tmpObject.i + 1
m = MultilayerPerceptron(inputs, nbHidden, outputs, learning_rate=lrate, momentum=momentum, grid=mode)
m.init_weights_randomly(-1, 1)
tmpObject.net = m
return m
def SoN(inputs, outputs):
n = MultilayerPerceptron(nbHidden,
20,
2,
learning_rate=0.15,
momentum=0.5,
grid=mode)
n.init_weights_randomly(-1, 1)
return n
def SoN(inputs, outputs):
return MultilayerPerceptron(inputs // 4,
inputs // 2,
inputs + inputs // 4 + outputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
def SoN(inputs, outputs):
return MultilayerPerceptron(nbr_hidden,
nbr_hidden * 2,
inputs + nbr_hidden + outputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
def SoN(inputs, outputs):
return MultilayerPerceptron(nbHidden,
20,
2,
learning_rate=0.1,
momentum=0.,
grid=mode)
def FoN2(inputs, outputs):
return MultilayerPerceptron(inputs,
inputs // 4,
outputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
def SoN(inputs, outputs):
tmpObject.last2 = MultilayerPerceptron(inputs // 4,
inputs // 2,
inputs + inputs // 4 + outputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
return tmpObject.last2
def control(inputs, outputs):
seed(tmpObject.i)
tmpObject.i = tmpObject.i + 1
tmpObject.control = MultilayerPerceptron(inputs,
nbHidden,
outputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
tmpObject.control.init_weights_randomly(-1, 1)
return tmpObject.control
from data import DataFile
if __name__ == '__main__':
mode = MultilayerPerceptron.R0to1
nbr_network = 5
momentum = 0.5
nbEpoch = 201
nbTry = 50
display_interval = range(nbEpoch)[3::5]
#create all networks
networks = [{} for _ in range(nbr_network)]
for i in range(nbr_network):
seed(i)
first_order = MultilayerPerceptron(16 * 16, 100, 10, learning_rate=0.15, momentum=momentum, grid=mode)
first_order.init_weights_randomly(-1, 1)
high_order_h = PRenforcement(100, temperature=1.)
networks[i] = {'first_order' : first_order,
'high_order_h' : high_order_h}
#create example
examples = DataFile("digit_handwritten_16.txt", mode)
#3 curves
y_plot = {'first_order' : [] ,
'high_order_h' : [],
'high_order_l': []}
y_perfo = {'first_order' : [] ,
return (True, random() < 0.7)
else:
return (False, random() < 0.3)
if __name__ == '__main__':
plt.title("low")
grid = MultilayerPerceptron.R0to1
nbr_network = 30
nbr_epoch = 200
nbr_trial = 10
first_order = []
high_order = []
for _ in range(nbr_network):
fo = MultilayerPerceptron(5, 40, 4, grid, 0.002, 0., 1., False, True)
fo.init_weights_randomly(-1, 1)
first_order.append(fo)
#0.0003 or 0.015
ho = MultilayerPerceptron(40, 40, 2, grid, 0.0003, 0., 1., False, True)
ho.init_weights_randomly(-1, 1)
high_order.append(ho)
#trials
wins = []
wagers = []
last_output = [ [0 for _ in range(5)] for _ in range(nbr_network)]
for epoch in range(nbr_epoch):
def pattern_to_list(pat):
res = []
for i in range(8):
res.append(index_max(pat[i * 6:i * 6 + 6]))
return res
if __name__ == '__main__':
ptrain_pattern = [random_pattern() for _ in range(80)]
# print(ptrain_pattern[0][0])
# print(pattern_to_list(ptrain_pattern[0][0]))
# exit()
first_order = MultilayerPerceptron(48, 40, 48, 0, 0.4, 0.5, 1., False,
True)
first_order.init_weights_randomly(-1, 1)
high_order = [
PerceptronR0to1(48, 0.4, 0.5, 1., Perceptron.OUTPUT, False, True)
for _ in range(2)
]
# high_order = [PerceptronN0to1(48, 0.4, 0.5, False) for _ in range(2)]
high_order[0].init_weights_randomly(0., 0.1)
high_order[1].init_weights_randomly(0., 0.1)
print("pre-training")
err = 0.
the = 0.
rms = []
rms2 = []
def SoN(inputs, outputs):
n = MultilayerPerceptron(nbHidden, 20, 2, learning_rate=0.15, momentum=0.5, grid=mode)
n.init_weights_randomly(-1, 1)
return n
else:
return (False, random() < 0.2)
if __name__ == '__main__':
grid = MultilayerPerceptron.R0to1
nbr_network = 50
nbr_epoch = 200
nbr_trial = 10
displays = range(nbr_epoch)[::5]
first_order = []
high_order = []
for _ in range(nbr_network):
fo = MultilayerPerceptron(5, 40, 4, grid, 0.002, 0., 1., False, True)
fo.init_weights_randomly(-1, 1)
first_order.append(fo)
#0.0003 or 0.015
ho = MultilayerPerceptron(40, 40, 2, grid, 0.015, 0., 1., False, True)
ho.init_weights_randomly(-1, 1)
high_order.append(ho)
#trials
wins = []
wagers = []
last_output = [[0 for _ in range(5)] for _ in range(nbr_network)]
for epoch in range(nbr_epoch):
nbr_win = 0
# examples = DataFileR("iris.txt")
momentum = 0.9
nbInputs = len(examples.inputs[0])
nbHidden = 10
nbOutputs = len(examples.outputs[0])
nbShape = nbOutputs
#create all networks
networks = [{} for _ in range(nbr_network)]
for i in range(nbr_network):
first_order = MultilayerPerceptron(nbInputs,
nbHidden,
nbOutputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
high_order_10 = MultilayerPerceptron(nbHidden,
nbHidden * 2,
nbInputs + nbHidden + nbOutputs,
learning_rate=lrate,
momentum=momentum,
grid=mode)
networks[i] = {
'first_order': first_order,
'high_order_10': high_order_10
}
learned = [[0 for _ in range(nbShape)] for _ in range(nbDiscre**nbHidden)]
def FoN(inputs, outputs):
tmpObject.last = MultilayerPerceptron(inputs, nbr_hidden, outputs, learning_rate=lrate, momentum=momentum, grid=mode)
return tmpObject.last
if __name__ == '__main__':
mode = MultilayerPerceptron.R0to1
nbr_network = 5
momentum = 0.9
lrate = 0.1
nbEpoch = 1000
display_interval = [0, 25, 50, 100, 200, 500, 999]
display_interval2 = range(nbEpoch)[::7]
#create all networks
networks = [{} for _ in range(nbr_network)]
for i in range(nbr_network):
first_order = MultilayerPerceptron(20,
5,
10,
learning_rate=lrate,
momentum=momentum,
grid=mode)
high_order_10 = MultilayerPerceptron(5,
10,
35,
learning_rate=lrate,
momentum=momentum,
grid=mode)
high_order_5 = MultilayerPerceptron(5,
5,
35,
learning_rate=lrate,
momentum=momentum,
grid=mode)
mode = MultilayerPerceptron.R0to1
nbr_network = 5
momentum = 0.5
nbEpoch = 201
nbTry = 50
display_interval = range(nbEpoch)[3::5]
seed(100)
#create all networks
networks = [{} for _ in range(nbr_network)]
for i in range(nbr_network):
seed(i)
control = MultilayerPerceptron(16 * 16,
100,
10,
learning_rate=0.15,
momentum=momentum,
grid=mode)
control.init_weights_randomly(-1, 1)
high_order_h = MultilayerPerceptron(100,
20,
2,
learning_rate=0.1,
momentum=0.,
grid=mode)
first_order = AdHock(control)
networks[i] = {
'first_order': first_order,
'high_order_h': high_order_h,
if __name__ == '__main__':
mode = MultilayerPerceptron.R0to1
nbr_network = 5
momentum = 0.5
nbEpoch = 201
nbTry = 50
display_interval = range(nbEpoch)[3::5]
#create all networks
networks = [{} for _ in range(nbr_network)]
for i in range(nbr_network):
seed(i)
control = MultilayerPerceptron(16 * 16, 100, 10, learning_rate=0.15, momentum=momentum, grid=mode,
temperature=1, random=False, enable_bias=True)
control.init_weights_randomly(-1, 1)
first_order = AdHock(control)
networks[i] = {'first_order' : first_order,
'control': control}
#create example
examples = DataFile("digit_handwritten_16.txt", mode)
#3 curves
y_perfo = {'first_order' : [] ,
'high_order_h' : [],
'wager_proportion': [],
'feedback' : [],
import matplotlib.pyplot as plt
from data import DataFile
if __name__ == '__main__':
mode = MultilayerPerceptron.R0to1
nbr_network = 5
momentum = 0.5
nbEpoch = 201
display_interval = range(nbEpoch)[3::5]
#create all networks
networks = [{} for _ in range(nbr_network)]
for i in range(nbr_network):
first_order = MultilayerPerceptron(7, 100, 10, learning_rate=0.15, momentum=momentum, grid=mode)
high_order_h = MultilayerPerceptron(100, 100, 2, learning_rate=0.1, momentum=0, grid=mode)
high_order_l = MultilayerPerceptron(100, 100, 2, learning_rate=10e-7, momentum=0, grid=mode)
networks[i] = {'first_order' : first_order,
'high_order_h' : high_order_h,
'high_order_l' : high_order_l}
for network in networks:
for k in network.keys():
if(k == 'first_order'):
network[k].init_weights_randomly(-1, 1)
#create example
examples = DataFile("digital_shape.txt", mode)
def pattern_to_list(pat):
res = []
for i in range(8):
res.append(index_max(pat[i * 6 : i * 6 + 6]))
return res
if __name__ == "__main__":
ptrain_pattern = [random_pattern() for _ in range(80)]
# print(ptrain_pattern[0][0])
# print(pattern_to_list(ptrain_pattern[0][0]))
# exit()
first_order = MultilayerPerceptron(48, 40, 48, 0, 0.4, 0.5, 1.0, False, True)
first_order.init_weights_randomly(-1, 1)
high_order = [PerceptronR0to1(48, 0.4, 0.5, 1.0, Perceptron.OUTPUT, False, True) for _ in range(2)]
# high_order = [PerceptronN0to1(48, 0.4, 0.5, False) for _ in range(2)]
high_order[0].init_weights_randomly(0.0, 0.1)
high_order[1].init_weights_randomly(0.0, 0.1)
print("pre-training")
err = 0.0
the = 0.0
rms = []
rms2 = []
nbEpoch = 60
