Kim H. Parker edited 2_5_Proposed_algorithm_begin__.tex  about 8 years ago

Commit id: 098e7e954c0ccd56ec93b9dcd62e71107c9b134a

deletions | additions      

       

\item Estimate the probability density functions $\phi(dP)$, $\phi(\rho cdU)$, $\phi(dP_+)$ and $\phi(dP_-)$ using a kernel density estimation method.  \item Calculate $\phi(dP_+ \pm dP_-) = \phi(dP_+) \otimes \phi(\pm dP_-)$ by convolution.  \item Calculate the entropies $H(dP)$, $H(\rho cdU)$, $H(dP_+ + dP_-)$ and $H(dP_+ - dP_-)$  \item Calculate the mean pdfs $\phi(M_P) = \frac{1}{2}\big(\phi(dP) + \phi(dP_+ + dP_-)\big)$ and $\phi(M_U) = \frac{1}{2}\big(\phi(\rho cdU) - +  \phi(dP_+ - dP_-)\big)$. \item Calculate the entropies $H(M_P)$ and $H(M_U)$.  \item Calculate the Jensen-Shannon divergences $JS_P = H(M_P) - \frac{1}{2}\big(H(dP) + H(dP_+ + dP_-)\big)$ \\and $JS_U = H(M_U) - \frac{1}{2}\big(H(\rho cDU) + H(dP_+ - dP_-)\big)$.  \item Calculate the sum of the Jensen-Shannon distances $\Delta = \sqrt{JS_P} + \sqrt{JS_U}$.