def fixNegsConstrainedOpt(self, to_minimize=norm_diff, badness=1e6):
"""Uses constrained minimization to find approx q matrix.
to_minimize: metric for comparing orig result and new result.
badness: scale factor for penalizing negative off-diagonal values.
"""
if not sum_neg_off_diags(self._data):
return self
q = ravel(without_diag(self._data))
p = expm(self._data)(t=1)
def err_f(q):
new_q = reshape(array(q), (4, 3))
new_q = with_diag(new_q, -sum(new_q, 1))
p_new = expm(new_q)(t=1)
result = to_minimize(ravel(p), ravel(p_new))
if q.min() < 0:
result += -q.min() * badness
return result
a = array(q)
xmin = fmin(func=err_f, x0=a, disp=0)
r = reshape(xmin, (4, 3))
new_q = with_diag(r, -sum(r, 1))
return self.__class__(new_q, self.Alphabet)
def fixNegsFmin(self, method=fmin, to_minimize=norm_diff, debug=False):
"""Uses an fmin method to find a good approximate q matrix.
Possible values for method:
fmin: simplex method (the default)
fmin_bfgs: bfgs optimizer #always produces negative elements!
fmin_cg: cg optimizer #doesn't work!
fmin_powell: powell method #doesn't work!
"""
q = self._data
#bail out if q is already ok to start with
if not sum_neg_off_diags(q):
return self
err_f = self._make_error_f(to_minimize)
initial_guess = q.copy()
xmin = method(err_f, initial_guess.flat, disp=0)
#disp=0 turns off messages
new_q = reshape(xmin, self.Alphabet.Shape)[:]
if debug:
if sum_neg_off_diags(new_q):
raise Exception, 'Made invalid Q matrix: %s' % q
return self.__class__(new_q, self.Alphabet)
def fixNegsFmin(self, method=fmin, to_minimize=norm_diff, debug=False):
"""Uses an fmin method to find a good approximate q matrix.
Possible values for method:
fmin: simplex method (the default)
fmin_bfgs: bfgs optimizer #always produces negative elements!
fmin_cg: cg optimizer #doesn't work!
fmin_powell: powell method #doesn't work!
"""
q = self._data
#bail out if q is already ok to start with
if not sum_neg_off_diags(q):
return self
err_f = self._make_error_f(to_minimize)
initial_guess = q.copy()
xmin = method(err_f, initial_guess.flat, disp=0)
#disp=0 turns off messages
new_q = reshape(xmin, self.Alphabet.Shape)[:]
if debug:
if sum_neg_off_diags(new_q):
raise Exception, 'Made invalid Q matrix: %s' % q
return self.__class__(new_q, self.Alphabet)
def fixNegsConstrainedOpt(self, to_minimize=norm_diff, badness=1e6):
"""Uses constrained minimization to find approx q matrix.
to_minimize: metric for comparing orig result and new result.
badness: scale factor for penalizing negative off-diagonal values.
"""
if not sum_neg_off_diags(self._data):
return self
q = ravel(without_diag(self._data))
p = expm(self._data)(t=1)
def err_f(q):
new_q = reshape(array(q), (4,3))
new_q = with_diag(new_q, -sum(new_q, 1))
p_new = expm(new_q)(t=1)
result = to_minimize(ravel(p), ravel(p_new))
if q.min() < 0:
result += -q.min() * badness
return result
a = array(q)
xmin = fmin(func=err_f, x0=a, disp=0)
r = reshape(xmin, (4,3))
new_q = with_diag(r, -sum(r, 1))
return self.__class__(new_q, self.Alphabet)
def result(q):
new_q = reshape(q, (4,4))
neg_sum = sum_neg_off_diags(new_q)
p_new = expm(new_q)(t=1)
return to_minimize(ravel(p), ravel(p_new)) - (BIG * neg_sum) \
+ (BIG * sum(abs(sum(new_q,1))))
def sumNegOffDiags(self):
"""Returns sum of negative off-diagonal elements."""
return sum_neg_off_diags(self._data)
def result(q):
new_q = reshape(q, (4, 4))
neg_sum = sum_neg_off_diags(new_q)
p_new = expm(new_q)(t=1)
return to_minimize(ravel(p), ravel(p_new)) - (BIG * neg_sum) \
+ (BIG * sum(abs(sum(new_q,1))))
def sumNegOffDiags(self):
"""Returns sum of negative off-diagonal elements."""
return sum_neg_off_diags(self._data)