def grad_loop(self, miu):
"""Gradient function for optimization
Calculate gradient of objective function using finite differences.
Is called by get_ipopt_miu(). Calls loop(). Stores and returns result.
Args:
miu (nd.array): Array of values for miu, n = Dice().params.tmax
Returns:
nd.array: gradient of objective
"""
df = DiceDataMatrix(np.tile(self.vars, (61, 1, 1)).transpose(1, 2, 0))
for i in xrange(self.params.tmax):
df.miu[i, :] = miu[i]
df.miu[i, i] += self.eps
self.step(i, df, df.miu[i], deriv=True, opt=True)
self.set_opt_values(df)
return self.opt_grad_f
def obj_loop(self, miu):
"""Objective function for optimization
Calculate objective function. Is called by get_ipopt_miu().
Calls loop(). Stores and returns result.
Args:
miu (nd.array): Array of values for miu, n = Dice().params.tmax
Returns:
float: value of objective (utility)
"""
df = DiceDataMatrix(np.tile(self.vars, (61, 1, 1)).transpose(1, 2, 0))
for i in xrange(self.params.tmax):
df.miu[i, :] = miu[i]
df.miu[i, i] += self.eps
self.step(i, df, df.miu[i], deriv=True, opt=True)
self.set_opt_values(df)
return self.opt_obj
