def iterate(self):
direction = linear_system_solution_llt(self._g_f(self._x), self._grad)
self._g_f_count += 1
def g(a: float):
return self._f(add_vector(self._x, scale_vector(a, direction)))
step_max = 0.
f_step_max = g(step_max)
self._f_count += 1
while True:
next_f = g(step_max + 1.)
self._f_count += 1
step_max += 1.0
if f_step_max < next_f:
break
f_step_max = next_f
step_min = 0.
if (step_max - 2.) > 0.:
step_min = step_max - 1.
gs = GoldenRatioSearch(g, step_min, step_max, self._tol)
while not gs.test_result():
gs.iterate()
res_1, res_2, f_count = gs.get_result()
step = (res_1 + res_2) / 2
self._f_count += f_count
self._x = add_vector(self._x, scale_vector(step, direction))
self._grad = scale_vector(-1., self._df(self._x))
self._df_count += 1
def iterate(self):
direction = linear_system_solution_llt(self._g_f(self._x), self._grad)
self._g_f_count += 1
step = 1.0
self._f_count += 2
f_x_k = self._f(self._x)
grad_inner_p = self._delta * inner_product(self._grad, direction)
while (self._f(add_vector(self._x, scale_vector(step, direction))) - f_x_k) > \
step * grad_inner_p:
step *= self._mul
self._x = add_vector(self._x, scale_vector(step, direction))
self._grad = scale_vector(-1., self._df(self._x))
self._df_count += 1
def __init__(self, dfunc: Callable, step: float, tol: float, x0: List[float]):
self._df = dfunc
self._step = step
self._tol = tol
self._x0 = x0
self._x = x0
self._grad = scale_vector(-1., dfunc(x0))
self._df_count = 1
def _next_vector(self, u):
i_k = self._n_k[index_of_positive_element([u[i] for i in self._n_k])]
coefficient_k = self._x[i_k] / u[i_k]
for i in self._n_k:
if u[i] > 0:
temp1 = self._x[i] / u[i]
if temp1 < coefficient_k:
i_k = i
coefficient_k = temp1
self._x = sub_vector(self._x, scale_vector(coefficient_k, u))
return i_k
def __init__(self, func: Callable, dfunc: Callable, gesse_func: Callable, tol: float, x0: List[float]):
self._g_f = gesse_func
self._df = dfunc
self._f = func
self._tol = tol
self._x0 = x0
self._x = x0
self._grad = scale_vector(-1., dfunc(x0))
self._df_count = 1
self._f_count = 0
self._g_f_count = 0
def restart(self):
self._x = self._x0
self._grad = scale_vector(-1., self._df(self._x0))
self._df_count = 1
def iterate(self):
self._x = add_vector(self._x, scale_vector(self._step, self._grad))
self._grad = scale_vector(-1., self._df(self._x))
self._df_count += 1
def g(a: float):
return self._f(add_vector(self._x, scale_vector(a, direction)))