NNAD-Interface

This repository includes a collection of applications of the NNAD library for the computation of analytic derivatives of a feed-forward neural network to minimisation problems. The examples relie on the ceres-solver solver minimiser that also provides numerical and automatic differentiations.

Prerequisites

The basic dependencies to run the example codes are:

• cmake
• pkg-config
• ceres-solver
• glog
• gflags
• eigen3
• yaml-cpp

Additional external libraries may be required depending on the specific examples. More details are found in the README files in the single example folders.

Examples

The following example codes are available so far:

• Legendre: a direct fit of a NN to pseudodata generated using an oscillating Legendre polynomial as an underlying law.
• PDFs: a realistic minimisation problem in which the functions to be determined appear inside a convolution integral. This problem closely resembles the extraction of collinear parton-distribution functions (PDFs) from experimental data.

Reference

• Rabah Abdul Khalek, Valerio Bertone, On the derivatives of feed-forward neural networks, arXiv:2005.07039

Contacts

• Rabah Abdul Khalek: rabah.khalek@gmail.com
• Valerio Bertone: valerio.bertone@cern.ch