wassname/simpeg
1
0
Fork 0
mirror of https://github.com/wassname/simpeg.git synced 2026-07-26 13:37:21 +08:00
Code Issues Packages Projects Releases Wiki Activity

comment G vec and Gt vec

Browse Source
This commit is contained in:
rowanc1
2014-04-27 21:49:00 -07:00
parent 45884ba865
commit d1f23081d6
1 changed files with 22 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
simpegEM/TDEM/TDEM_b.py
+22 -8
View File
@@ -74,29 +74,43 @@ class ProblemTDEM_b(BaseTDEMProblem):
:rtype: simpegEM.TDEM.FieldsTDEM
:return: f
Multiply G by a vector where
Multiply G by a vector
"""
if u is None:
u = self.fields(m)
# Note: Fields has shape (nF/E, nTx, nT+1)
# However, p will only really fill (:,:,1:nT+1)
# meaning the 'initial fields' are zero (:,:,0)
p = FieldsTDEM(self.mesh, self.survey)
p[:, 'b', :] = 0.0 #np.zeros((self.mesh.nF, self.survey.nTx, self.prob.nT))
p[:, 'e', 0] = 0.0 #np.zeros((self.mesh.nF, self.survey.nTx))
# p = FieldsTDEM(self.mesh, 1, self.nT, 'b')
p[:, 'b', :] = 0.0 # b at all times is zero.
p[:, 'e', 0] = 0.0 # fake initial fields
curModel = self.mapping.transform(m)
c = self.mesh.getEdgeInnerProductDeriv(curModel)*self.mapping.transformDeriv(m)*vec
for i in range(self.nT):
for i in range(1,self.nT+1):
# TODO: G[1] may be dependent on the model
# for a galvanic source (deriv of the dc problem)
for tx in self.survey.txList:
p[tx, 'e', i+1] = -u[tx,'e',i+1]*c
p[tx, 'e', i] = -u[tx,'e',i]*c # - diag(e) * MsigDeriv * v
return p
def Gtvec(self, m, v, u=None):
def Gtvec(self, m, vec, u=None):
"""
:param numpy.array m: Conductivity model
:param numpy.array vec: vector (like a fields)
:param simpegEM.TDEM.FieldsTDEM u: Fields resulting from m
:rtype: np.ndarray (like a model)
:return: p
Multiply G.T by a vector
"""
if u is None:
u = self.fields(m)
nTx, nE = self.survey.nTx, self.mesh.nE
tmp = np.zeros(nE if nTx == 1 else (nE,nTx))
# Here we can do internal multiplications of Gt*v and then multiply by MsigDeriv.T in one go.
for i in range(1,self.nT+1):
tmp += v[:,'e',i]*u[:,'e',i]
tmp += vec[:,'e',i]*u[:,'e',i]
curModel = self.mapping.transform(m)
p = -mkvc(self.mapping.transformDeriv(m).T*self.mesh.getEdgeInnerProductDeriv(curModel).T*tmp)