Extrema finder

This package computes the extrema of a 3D field.

Requirements

Requires:

• BLAS
• LAPACK
• FFTW3 (will be removed eventually)
• numpy
• python 2 or 3 (of course!)
• gfortran (for other compilers, see below)

If you're using ifort or other, you have to clone this repository and modify the Makefile accordingly.

Install

Once you've met all dependencies, just run:

pip install -e git+https://github.com/cphyc/extrema.git#egg=extrema

Example

from extrema import extrema
import numpy as np

# Generate a random field
field = np.random.rand(32**3).reshape(32, 32, 32)

# Mandatory: set parameters of the run
npeaks = 32**3  # max number of peaks
nbins = 32**3   # number of bins of the field
nthreads = 2    # number of threads to use
extrema.set(npeaks, nbins, nthreads)

# Find the peaks
extrema.compute(field)

# Get the data back
data = dict(
     pos=extrema.peak_pos,     # pos of the critical point
     val=extrema.peak_values,  # value at the point
     eigval=extrema.eigval,    # eigen values of the hessian
     eigvec=extrema.eigvec,    # eigen vectors of the hessian
     type=extrema.peak_types,  # type of the extrema (4: maximum, 3,2:saddle point, 1:minimum)
     index=extrema.peak_index  # unravelled index of the cell containing the critical point
)

# Unravel indices
indices = np.unravel_index(data['index'], field.shape, order='F')

Copyright

The Fortran Routines have been borrowed from Dmitri Pogosyan. Any work using this package should name him explicitly.

License

This work is licensed under the CC-BY-SA license. You are allowed to copy, modify and distribute it as long as you keed the license. See more in the LICENSE file.