class Potential(object):
def __init__(self, amount_of_classes, threshold=50):
self.__amount_of_classes = amount_of_classes
self.__threshold = threshold
self.__result = Function(0, 0, 0, 0)
self.__correction = 1
def train(self, training_vectors):
isContinue = True
iteration_number = 0
isSuccess = True
while isContinue and (iteration_number < self.__threshold):
iteration_number += 1
isContinue = self.__iterate(training_vectors)
if iteration_number >= self.__threshold:
isSuccess = False
return self.__result, isSuccess
def __iterate(self, training_vectors):
isContinue = False
for vector_index, vector in enumerate(training_vectors):
for point_index, point in enumerate(vector):
self.__result += self.__get_separating_function(
point) * self.__correction
index = (point_index + 1) % len(vector)
next_class_number = (
vector_index + 1
) % self.__amount_of_classes if index == 0 else vector_index
next_point = training_vectors[next_class_number][index]
self.__correction = self.__calc_new_correction(
next_point, next_class_number)
if self.__correction != 0:
isContinue = True
return isContinue
def __get_separating_function(self, point):
return Function(point.x * 4, point.y * 4, point.x * point.y * 4 * 4, 1)
def __calc_new_correction(self, next_point, next_class_number):
function_value = self.__result.calc_value(next_point)
if (function_value <= 0) and (next_class_number == 0):
return 1
if (function_value > 0) and (next_class_number == 1):
return -1
return 0
