def _eval_sfun(self, ngen=1):
"""Run DE for a function that takes a single pv as an input and retuns a single value."""
popc, fitc = self._population, self._fitness
popt, fitt = self._trial_pop, self._trial_fit
for ipop in xrange(self.n_pop):
fitc[ipop] = self.m * self.minfun(popc[ipop,:])
for igen in xrange(ngen):
popt[:,:] = de_f.evolve_population(popc, self.F, self.C)
for ipop in xrange(self.n_pop):
fitt[ipop] = self.m * self.minfun(popt[ipop,:])
msk = fitt < fitc
popc[msk,:] = popt[msk,:]
fitc[msk] = fitt[msk]
self._minidx = np.argmin(fitc)
yield popc[self._minidx,:], fitc[self._minidx]
def _eval_sfun(self, ngen=1):
"""Run DE for a function that takes a single pv as an input and retuns a single value."""
popc, fitc = self._population, self._fitness
popt, fitt = self._trial_pop, self._trial_fit
for ipop in xrange(self.n_pop):
fitc[ipop] = self.m * self.minfun(popc[ipop, :])
for igen in xrange(ngen):
popt[:, :] = de_f.evolve_population(popc, self.F, self.C)
for ipop in xrange(self.n_pop):
fitt[ipop] = self.m * self.minfun(popt[ipop, :])
msk = fitt < fitc
popc[msk, :] = popt[msk, :]
fitc[msk] = fitt[msk]
self._minidx = np.argmin(fitc)
yield popc[self._minidx, :], fitc[self._minidx]
def _eval_vfun(self, ngen=1):
"""Run DE for a function that takes the whole population as an input and retuns a value for each pv."""
popc, fitc = self._population, self._fitness
popt, fitt = self._trial_pop, self._trial_fit
fitc[:] = self.m * self.minfun(self._population)
for igen in xrange(ngen):
x = float(ngen - igen) / float(ngen)
self.F = np.random.uniform(0.25, 0.75)
self.C = x * self.cmax + (1 - x) * self.cmin
popt[:, :] = de_f.evolve_population(popc, self.F, self.C)
fitt[:] = self.m * self.minfun(popt)
msk = fitt < fitc
popc[msk, :] = popt[msk, :]
fitc[msk] = fitt[msk]
self._minidx = np.argmin(fitc)
yield popc[self._minidx, :], fitc[self._minidx]
def _eval_vfun(self, ngen=1):
"""Run DE for a function that takes the whole population as an input and retuns a value for each pv."""
popc, fitc = self._population, self._fitness
popt, fitt = self._trial_pop, self._trial_fit
fitc[:] = self.m * self.minfun(self._population)
for igen in xrange(ngen):
x = float(ngen-igen)/float(ngen)
self.F = np.random.uniform(0.25,0.75)
self.C = x*self.cmax + (1-x)*self.cmin
popt[:,:] = de_f.evolve_population(popc, self.F, self.C)
fitt[:] = self.m * self.minfun(popt)
msk = fitt < fitc
popc[msk,:] = popt[msk,:]
fitc[msk] = fitt[msk]
self._minidx = np.argmin(fitc)
yield popc[self._minidx,:], fitc[self._minidx]