18.01.2015 Views

Karjalainen, Pasi A. Regularization and Bayesian methods for ...

Karjalainen, Pasi A. Regularization and Bayesian methods for ...

Karjalainen, Pasi A. Regularization and Bayesian methods for ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

24 2. Estimation theory<br />

2.5 Bayes cost method<br />

If we assume that θ is a r<strong>and</strong>om vector with a known joint density p(θ,z) with<br />

the observation z, we have made the so-called <strong>Bayesian</strong> assumption [142]. This<br />

assumption leads to the so-called Bayes cost method <strong>for</strong> solving the estimator<br />

ˆθ(z). For the cost method we define the function C(θ, ˆθ) that assigns to each<br />

combination of actual parameter value <strong>and</strong> estimate the unique real valued cost.<br />

We call C(θ, ˆθ) the cost function. The expected value of the cost is given by<br />

{<br />

B(ˆθ) = E C(θ, ˆθ(z))<br />

}<br />

=<br />

∫ ∞ ∫ ∞<br />

−∞<br />

−∞<br />

C(θ, ˆθ(z))p(θ,z)dθ dz (2.30)<br />

that is called the Bayes cost. From (2.15) we obtain p(θ,z) = p(z|θ)p(θ) <strong>and</strong> the<br />

expectation can be written in the <strong>for</strong>m<br />

∫ ∞<br />

(∫ ∞<br />

)<br />

B(ˆθ) = C(θ, ˆθ(z))p(z|θ)dz p(θ)dθ (2.31)<br />

−∞ −∞<br />

The inner integral is clearly the conditional expectation of the cost given θ <strong>and</strong> we<br />

write<br />

∫ ∞<br />

B(ˆθ|θ) = C(θ, ˆθ(z))p(z|θ)dz (2.32)<br />

−∞<br />

{<br />

= E C(θ, ˆθ)<br />

}<br />

∣<br />

∣θ<br />

(2.33)<br />

This can be called the conditional Bayes cost, given θ, <strong>and</strong> in terms of the conditional<br />

cost the Bayes cost of the estimator can be written as<br />

∫ ∞<br />

B(ˆθ) = B(ˆθ|θ)p(θ)dθ (2.34)<br />

−∞<br />

{<br />

= E θ<br />

{E C(θ, ˆθ)<br />

}}<br />

∣<br />

∣θ<br />

(2.35)<br />

}<br />

= E θ<br />

{B(ˆθ|θ)<br />

(2.36)<br />

Similarily using p(θ,z) = p(θ|z)p(z) we can write<br />

{<br />

B(ˆθ) = E C(θ, ˆθ(z))<br />

}<br />

{<br />

= E z<br />

{E C(θ, ˆθ)<br />

}}<br />

∣<br />

∣z<br />

}<br />

= E z<br />

{B(ˆθ|z)<br />

(2.37)<br />

(2.38)<br />

(2.39)<br />

where<br />

∫ ∞<br />

B(ˆθ|z) = C(θ, ˆθ(z))p(θ|z)dθ (2.40)<br />

−∞<br />

{<br />

= E C(θ, ˆθ)<br />

}<br />

∣<br />

∣z<br />

(2.41)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!