def out(self, input, alpha=None, theta=None):
_input = input if isinstance(input, np.ndarray) else np.array(input)
_alpha = alpha if alpha else self._alpha
_theta = theta if theta else self._theta
logging.debug('PARAMETERS (alpha, theta): (%s, %s)' % (_alpha, _theta))
logging.debug('INPUT: (%s), SIGMOID: %s' % (','.join(
[str(i)
for i in _input]), sigmoid(self._w.dot(_input), _alpha, _theta)))
if np.random.rand(1, 1) < sigmoid(self._w.dot(_input), _alpha, _theta):
logging.debug('OUTPUT: %s' % 1)
return 1
else:
logging.debug('OUTPUT: %s' % 0)
return 0
def data_frame(self):
bin_values = []
for state in self._states:
bin_values.append(0 if sigmoid(np.round(state, 4)) < 0.5 else 1)
return pd.DataFrame(
data={"states": self._states, "values": self._values, "binaries": bin_values}, index=self.domain
)
def __init__(self, name, w, initial_state, alpha=1.0, theta=0.0, time_delta=0.001):
self._name = name
self._w = w if isinstance(w, np.ndarray) else np.array(w)
self._alpha = alpha
self._theta = theta
self._time_delta = time_delta
self._states = [initial_state]
self._values = [sigmoid(self._states[-1], self._alpha, self._theta)]
def activate(self, x):
assert len(x) == len(self.W[0][0]) - 1
self.Y = [np.append(x, [1.0, ]), ]
for l in range(len(self.W)):
self.Y.append(np.ones(len(self.W[l]) + (1 if l < len(self.W[l]) - 1 else 0)))
for i, w in enumerate(self.W[l]):
self.Y[-1][i] = sigmoid(w.dot(self.Y[-2]))
return self.Y[-1]
def data_frame(self):
bin_values = []
for state in self._states:
bin_values.append(0 if sigmoid(np.round(state, 4)) < 0.5 else 1)
return pd.DataFrame(data={
'states': self._states,
'values': self._values,
'binaries': bin_values
},
index=self.domain)
def out(self, input, alpha=None, theta=None):
_input = input if isinstance(input, np.ndarray) else np.array(input)
_alpha = alpha if alpha else self._alpha
_theta = theta if theta else self._theta
logging.debug('PARAMETERS (alpha, theta): (%s, %s)' % (_alpha, _theta))
logging.debug('INPUT: (%s), SIGMOID: %s' % (','.join([str(i) for i in _input]), sigmoid(self._w.dot(_input), _alpha, _theta)))
if np.random.rand(1, 1) < sigmoid(self._w.dot(_input), _alpha, _theta):
logging.debug('OUTPUT: %s' % 1)
return 1
else:
logging.debug('OUTPUT: %s' % 0)
return 0
def __init__(self,
name,
w,
initial_state,
alpha=1.0,
theta=0.0,
time_delta=0.001):
self._name = name
self._w = w if isinstance(w, np.ndarray) else np.array(w)
self._alpha = alpha
self._theta = theta
self._time_delta = time_delta
self._states = [
initial_state,
]
self._values = [
sigmoid(self._states[-1], self._alpha, self._theta),
]
def get_value(self, x):
return sigmoid(self._w.dot(np.append([
1.0,
], x)))
def update(self, input):
_input = input if isinstance(input, np.ndarray) else np.array(input)
self._states.append(self._states[-1] + self._time_delta * (-self._states[-1] + self._w.dot(_input)))
self._values.append(sigmoid(self._states[-1], alpha=self._alpha, theta=self._theta))
def get_value(self, x):
return sigmoid(self._w.dot(np.append([1.0, ], x)))
def update(self, input):
_input = input if isinstance(input, np.ndarray) else np.array(input)
self._states.append(self._states[-1] + self._time_delta *
(-self._states[-1] + self._w.dot(_input)))
self._values.append(
sigmoid(self._states[-1], alpha=self._alpha, theta=self._theta))
return 0
def __call__(self, input, alpha=None, theta=None):
return self.out(input, alpha, theta)
if __name__ == '__main__':
#logging.basicConfig(level=logging.DEBUG)
w = np.array([3, 2, 1])
alpha = 1
theta = 1
p_element = ProbabilisticElement(w, alpha, theta)
num_of_loops = 50000
for i in [
[1, 0, 1],
[1, -1, 0],
]:
num_of_ones = 0
for l in range(0, num_of_loops):
if p_element(i):
num_of_ones += 1
probability = num_of_ones / num_of_loops
print('Sigmoid: ', sigmoid(w.dot(i), theta=theta))
print('Probability of one: ', probability)
assert np.round(probability,
2) == np.round(sigmoid(w.dot(i), theta=theta), 2)
logging.debug('OUTPUT: %s' % 1)
return 1
else:
logging.debug('OUTPUT: %s' % 0)
return 0
def __call__(self, input, alpha=None, theta=None):
return self.out(input, alpha, theta)
if __name__ == '__main__':
#logging.basicConfig(level=logging.DEBUG)
w = np.array([3, 2, 1])
alpha = 1
theta = 1
p_element = ProbabilisticElement(w, alpha, theta)
num_of_loops = 50000
for i in [[1, 0, 1], [1, -1, 0], ]:
num_of_ones = 0
for l in range(0, num_of_loops):
if p_element(i):
num_of_ones += 1
probability = num_of_ones / num_of_loops
print('Sigmoid: ', sigmoid(w.dot(i), theta=theta))
print('Probability of one: ', probability)
assert np.round(probability, 2) == np.round(sigmoid(w.dot(i), theta=theta), 2)
def f(self, s):
return sigmoid(s)
