Fitting code: n3fit

• n3fit is the next generation fitting code for NNPDF developed by the N3PDF team [CCM19]

• n3fit is responsible for fitting PDFs from NNPDF4.0 onwards.

• The code is implemented in python using Keras and can run with Tensorflow (default) or pytorch (with the environment variable KERAS_BACKEND=torch).

• The sections below are an overview of the n3fit design.

Important

If you just want to know how to run a fit using n3fit, head to How to run a PDF fit.

• Methodology overview
  • Introduction
  • Neural network architecture
  • Preprocessing
  • Optimizer
  • Stopping algorithm
  • Positivity
  • Integrability
  • Feature Scaling
  • Diagonal basis
• n3fit runcard detailed guide
  • Dataset selection
  • Monte Carlo replica data
  • Preprocessing
  • Network Architecture
  • Optimizer
  • Positivity
  • Integrability
  • Regularized covariance matrices
  • Inspecting and profiling the code
  • Running fits in parallel
  • Other options
• Hyperoptimization algorithm
  • Motivation
  • K-folding cross-validation
  • Creating partitions
  • Interpretation of results
  • Implementation in n3fit
  • Implementation in validphys
  • Positivity and integrability
  • Practical Usage
  • Changing the hyperoptimization target
  • Running hyperoptimizations in parallel with MongoDB