search

dfujim / bILT

Inverse Laplace Transform objects, optimized for BNMR data
GNU General Public License v3.0
0 stars 0 forks source link

No need to multiply by transpose? #1

Closed dfujim closed 4 years ago

dfujim commented 4 years ago

In these lines:

L = np.matmul(self.K.T, np.matmul(self.S, self.K))
q = np.matmul(self.K.T, np.matmul(self.S, self.y))

it may be superfluous to multiply by K.T. In the docs it is shown that this is not needed. The minimization function is actually slightly different if we do this.

rmlmcfadden commented 4 years ago

I've finally gone through it too and I think you are correct - the extra multiplication seems unnecessary. Removing the K.T multiplication also makes the resulting L-curve look much more like the ones in the literature, which I guess is further (indirect) confirmation that we are on the right path.