Welcome to the site of the NNPDFs collaboration
NNPDF approach to parton distribution functions
NNPDF NEWS
The LO and NNLO sets of the NNPDF2.1 release of Parton Distributions is now available for download here
The NNDPF collaboration places its main focus
in obtaining an unbiased global parton distribution functions (PDFs)
fit that represents faithfully statistical, systematic and
normalization errors associated to data.
The general strategy is made of several steps and sketched briefly as follows:
- Monte Carlo generation of replica
All errors as given by experimental collaborations are translated into
a Monte Carlo set of artificial data. This set does give back the
experimental covariance matrix.
- Construction and evolution of parton distributions
PDFs are constructed as Neural Networks in real space. Those neural pdfs
are then evolved and convoluted with Wilson coefficients to deliver
observables. The fitting of the neural network is done on each Monte Carlo
set using Genetic Algorithms.
- Statistical faithfulness
The final set of neural PDFs can then be used to reproduce observables,
including errors and their correlations.
More information on the NNPDF approach can be found in the
NNPDF @ Wikipedia pages.
|
