#	Pagina
attuale pagina	/open-h2020/projects/210200/results.html

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - NNNPDF (Proton strucure for discovery at the Large Hadron Collider)

Teaser

Summary

\"This project addresses the issue of determining the structure of the proton, as probed in high-energy collisions such as the Large Hadron Collider of CERN (LHC). Its main novel aspect consists of systematically using Machine Learning towards this goal. Subsidiary innovations deal with the way to estimate theoretical uncertainties, specifically those related to partial knowledge of the underlying predictions in the theory of strong interactions. The importance of this project for society is both direct and indirect. The direct impact has to do with making progress in our understanding of the fundamental laws of nature. These are currently encoded in a theory, the so-called standard model of fundamental interactions, which is tested experimentally at particle accelerators such as the LHC. Whereas no deviation between experimental data and the predictions of this theory has ever been observed, we know that it cannot be complete because, for example, it cannot account for dark matter which makes up about 85% of matter in the universe. The structure of the proton is determined by the strong force, one of the four fundamental forces of Nature, which is described within the standard model by the theory of quantum chromodynamics. Understanding the structure of the proton simultaneously probes our understanding of this theory, and also it enables the subtle tests of the standard model at the LHC which are currently our best way of going beyond the current theory. Indeed, because the LHC is a proton accelerator, no discovery is possible at the LHC without a detailed understanding of proton structure. For instance, the then best understanding of proton structure was crucial in the discovery of the Higgs boson in 2012. The techniques developed in this project are aimed at making possible discoveries which go beyond this and which will be the focus of experimentation at the LHC over the next two decades. The indirect impact has to do with the project methodology, namely the use of machine learning methods. Machine learning techniques are becoming ubiquitous in a variety of applications which go from speech recognition to self-driving cars. In all these situations, machine learning tools are used to determine a true answer from fuzzy information. In the context of this project instead, what is being determined from fuzzy information is a statistical distribution of true answers. This is due to the quantum nature of the objects being studied, which can only be characterized in terms of probability distributions. These techniques are likely to be useful for situations in which instead of a unique correct answer there exists a distribution of possible answers. The overall objectives of the project are the development of a suite of machine learning tools which can achieve a full determination of the proton structure, while optimizing automatically the way information is extracted from the underlying data. This is to be coupled with the development of a set of theoretical results specific to quantum chromodynamics which will lead to accurate estimates of missing theoretical predictions based on optimal exploitation of the information contained in partial results, some of which will be obtained specifically in this project, and all of which will be systematized.
In a first phase the overall objectives consist of singling out the specific machine learning tools which allow for automatic optimization of the determination of proton structure, and of classifying mathematical properties of partial results (\"\"resummation\"\") which allow for the approximate determination of yet unknown corrections to higher order theoretical predictions.\"

Work performed

A significant fraction of the activity during the first reporting period involved setting up the research group. A project assistant was recruited since day 1 of the project on overhead in order to help with all administrative and organizational procedure as well as with the dissemination of the project\'s results. Two postdocs and two PhD students were recruited, all with starting date 01 October 2018. Note that the starting date for PhD students is fixed by University regulations and the job market for postdocs in theoretical high energy physics follows a fixed calendar (see http://insti.physics.sunysb.edu/itp/postdoc-agreement.html). An assistant professor was hired by the HI and joined the team, with significant time commitment to the project, with the same starting date.
A website (http://n3pdf.mi.infn.it/) and associated twitter account were set up. They are regularly maintained and list all scientific and outreach activities from the group. All papers (published in open access) and code (publicly available through public repositories) are linked on this site.
The team has met once a week since the start of the project, with two more weekly conference call meetings held jointly by the NNNPDF team and the NNPDF collaboration led by the PI (http://nnpdf.mi.infn.it/). Two meetings in person of the NNNPDF team and the NNPDF collaboration were held in September 2018 in Gargnano (Lake Garda, Italy) and in February 2019 at VU Amsterdam. All this follows the original plan.
In terms of scientific achievements, two major pieces of preliminary work were accomplished by the PI, with the publication of the NNPDF3.1 PDF (parton distribution function) set, which amounts to a state-of-the-art determination of proton structure with current artificial-intelligence based, but not yet machine-learning based methodology. This was used for a precision determination of the strong coupling constant which determines the strength of the force which binds the constituents of the proton. The development of the new methodology, which is the main goal of the project, has produced first relevant, published preliminary results, which pertain the investigation of the possibility of automatizing the methodology. First direct results on proton structure determined using this methodology are expected in the coming months, in full agreement with the expected road map of the project. Work related to combined resummation, the technique used to glean information on missing theoretical information, was published specifically in relation to heavy quarks and here too applications to proton structure determination are expected in the coming months in agreement with the expected roadmap. Studies on the social impact of high energy physics research were further pursued and published.

Final results

The main progress expected by the end of the project is a determination of proton structure to an accuracy better than 1%, and a set of automated tools allowing for the improvement of this determination as more experimental information becomes available. This is to be supplemented by systematic machinery allowing for the estimate of missing higher order corrections. In terms of methodology, the first goal will involve the development of machine learning tools in a context in which the quantity to be determined is a probability distribution of results, as opposed to an individual result. The second goal will involve a full systematization of the so-called resummation of quantum chromodynamics: the set of methods whereby partial predictions are obtained which become exact in specific kinematic limits (such as when the energy of a collision is very high, or very low).