Taking into account noise which is always present in expermental data,
we have to rewrite (8) as
|
(9) |
In order to infer
, we apply
Bayesian probability theory for the calculation of the posterior probability
of a density distribution
given the absorbance data
. Using Bayes' Theorem we obtain:
|
(10) |
Assuming uncorrelated Gaussian noise with mean zero and variance
the likelihood is given by:
|
(11) |
As prior we choose the MaxEnt prior (see e.g. [Siv96] for
further discussion):
|
(12) |
with hyperparameter , default model
and entropy
defined by:
|
(13) |
The default model
is the MaxEnt-solution in case of strong
regularization
.
In order to obtain the maximum posterior or MAP solution for fixed
hyperparameter , we have to maximize
denoting the misfit by
and ignoring the denominator in
(12) since we assume that it does not vary much compared
to the exponential. The maximum is calculated numerically by Newton's method,
for the details we refer to [Lin01].
Finally we have to specify the optimal hyperparameter , which is fixed
by the relation:
|
(14) |
since we expect each data point to deviate by from its true
value on average. Starting with large (in order to ensure
convergence of the Newton iterations), the hyperparameter is determined by
interval bisection.
Danilo Neuber
2003-10-03