To run the programme type run dsk$21:[public.leeds.nmg.exe]maxent.exe. Alternatively, type @dsk$21: [public.leeds.nmg.com]login.com and this will define the symbol maxent which you can use to run the programme.
CAUTION (1) : The exponential in the basic iteration formula (shown on the first page) makes the algorithm unstable for
large values of . Although the programme will scale the input data and give
a sensible initial value
(
=0.1), it is
still possible that the iteration will get out of control and the programme crash. If this happens, you must decrease the
as described in the section ``Changing the defaults''. It is also possible that the MAXENT algorithm can not
converge to the expected value of
with a given
. In this case, you again have to decrease
.
CAUTION (2) : The map files must be calculated with the most recent MTZ version of the CCP4 suite.
Note well (1) : In the case of crystallographic applications, the input image must be one unit cell (not the asymmetric unit).
Note well (2) : The programme uses the Danielson & Lanczos FFT algorithm (from Numerical recipes, originally written from N.M.Brenner). This means : (i) the number of grid points in each dimension of the input map(s) must be a power of 2, and, (ii) assuming constant convergence properties (which is not true), MAXENT is a O(Nlog2N) process, where N is the number of grid points in the image. Calculating the MAXENT map of a 128x32x64 image (262144 grid points) took 2 hours and 15 minutes of CPU time on a DEC 4000. The same calculation for a 128x128 2D image took 10 minutes of CPU time on the same machine.