Pixel method parameters

The simplest crystal calculation consists of taking the atomic coordinates, cell data and space group for a crystal structure from an X-ray determination, building a static cluster of molecules to represent the crystal, and calculating all the interaction energies by some potential. This can be done by force field methods, Eq. (1.1.25) (time for one calculation a few milliseconds) or by Pixel, Eq. (1.1.34) (time for one calculation a few minutes). Figure 1.1.1 shows that the results are of comparable accuracy (remember that experimental values themselves are no more accurate than 5-10 kj mol-1). The UNI empirical force field [9] uses however about 100 parameters while the Pixel method uses only a few. 

Pattern recognition can be classified according to several parameters. Below we discuss only the supervised/unsupervised dichotomy because it represents two different ways of analyzing hyperspectral data cubes. Unsupervised methods (cluster analysis) classify image pixels without calibration and with spectra only, in contrast to supervised classifications. Feature extraction methods [21] such as PCA or wavelet compression are often applied before cluster analysis. 

Gull 1989) addresses these problems classical or quantified MaxEnt (in contrast to the previously described method, now dubbed historical ). It has been implemented in the MEMSYS5 package. Imaging is treated as a Bayesian parameter estimation problem for the pixel intensities. The prior probability has the form P(f) a ), the posterior probability for data D is given by P(f ) oc with the notation of Section 3. a is... 

---