def my_minimize (model, params): return scipy.optimize.minimize(model, params, method='SLSQP').x nR=np.copy(map(float, params) ) rml.set_max_P(1.0) ret=my_results(0) ret.niter=0 while True: # print params if(rml.get_max_P()<0.75): rml.set_max_P(0.75) ret.jac=Model_jac(params) ret.hess=Model_hess(params) if (np.linalg.det(ret.hess)==0): for x in range(0, len(params) ): params[x]=params[x]/2. continue R=np.linalg.inv( np.matrix(ret.hess ) )*np.matrix( ret.jac ).transpose() R=np.array(R.transpose() )[0] for x in range(0, len(R) ): nR[x]=params[x]-R[x] params=np.copy(nR) if (max ( abs(R) ) < 10**-6): break ret.niter+=1 if (ret.niter>99): ret.success=False break ret.x=params ret.func=model(params) return params
def my_minimize(model, params):
return scipy.optimize.minimize(model, params, method='SLSQP').x
nR = np.copy(map(float, params))
rml.set_max_P(1.0)
ret = my_results(0)
ret.niter = 0
while True:
# print params
if (rml.get_max_P() < 0.75):
rml.set_max_P(0.75)
ret.jac = Model_jac(params)
ret.hess = Model_hess(params)
if (np.linalg.det(ret.hess) == 0):
for x in range(0, len(params)):
params[x] = params[x] / 2.
continue
R = np.linalg.inv(np.matrix(ret.hess)) * np.matrix(ret.jac).transpose()
R = np.array(R.transpose())[0]
for x in range(0, len(R)):
nR[x] = params[x] - R[x]
params = np.copy(nR)
if (max(abs(R)) < 10**-6):
break
ret.niter += 1
if (ret.niter > 99):
ret.success = False
break
ret.x = params
ret.func = model(params)
return params
def my_dll(model, params, x):
ret = my_results(model(params))
ret.x = params
return scipy.optimize.minimize(model_x[x], params,
method='SLSQP').fun - model(ret.x)
params = [0., 0., 0., 0., 0., 0., 0.]
nR = np.copy(map(float, params))
ret.niter = 0
P = rml.get_max_P()
while True:
if (rml.get_max_P() < 0.75):
rml.set_max_P(0.75)
ret.hess = Model_hess(params)
ret.jac = Model_jac(params)
ret.hess = np.delete(ret.hess, (x), axis=0)
ret.hess = np.delete(ret.hess, (x), axis=1)
ret.jac = np.delete(ret.jac, (x), axis=0)
if (np.linalg.det(ret.hess) == 0):
for x in range(0, len(params)):
params[x] = params[x] / 2.
continue
try:
R = np.linalg.inv(np.matrix(ret.hess)) * np.matrix(
ret.jac).transpose()
except:
print params
print ret.hess
print np.linalg.inv(ret.hess)
print ret.jac
quit()
R = np.array(R.transpose())[0]
for y in range(0, len(params)):
if y < x:
nR[y] = params[y] - R[y]
elif y > x:
nR[y] = params[y] - R[y - 1]
params = np.copy(nR)
if (max(abs(R)) < 10**-4):
break
ret.niter += 1
if (ret.niter > 99):
ret.success = False
break
ret.func = model(params) - model(ret.x)
ret.x = params
rml.set_max_P(P)
return ret.func
def my_dll (model, params, x): ret=my_results(model(params) ) ret.x=params return scipy.optimize.minimize(model_x[x], params, method='SLSQP').fun-model(ret.x) params=[0.,0.,0.,0.,0.,0.,0.] nR=np.copy(map(float, params) ) ret.niter=0 P=rml.get_max_P() while True: if(rml.get_max_P()<0.75): rml.set_max_P(0.75) ret.hess=Model_hess(params) ret.jac=Model_jac(params) ret.hess=np.delete(ret.hess, (x), axis=0) ret.hess=np.delete(ret.hess, (x), axis=1) ret.jac=np.delete(ret.jac, (x), axis=0) if (np.linalg.det(ret.hess)==0): for x in range(0, len(params) ): params[x]=params[x]/2. continue try: R=np.linalg.inv( np.matrix(ret.hess ) )*np.matrix( ret.jac ).transpose() except: print params print ret.hess print np.linalg.inv(ret.hess) print ret.jac quit() R=np.array(R.transpose() )[0] for y in range(0, len(params) ): if y<x: nR[y]=params[y]-R[y] elif y>x: nR[y]=params[y]-R[y-1] params=np.copy(nR) if (max ( abs(R) ) < 10**-4): break ret.niter+=1 if (ret.niter>99): ret.success=False break ret.func=model(params)-model(ret.x) ret.x=params rml.set_max_P(P) return ret.func
