Search

Deep learning models are defined in terms of a large number of hyperparameters, such as network architectures and optimiser settings. These hyperparameters must be determined separately from the model parameters such as network weights, and are often fixed by ad-hoc methods or by manual inspection of the results. An algorithmic, objective determination of hyperparameters demands the introduction of dedicated target metrics, different from those adopted for the model training. Here we present a new approach to the automated determination of hyperparameters in deep learning models based on statistical estimators constructed from an ensemble of models sampling the underlying probability distribution in model space. This strategy requires the simultaneous parallel training of up to several hundreds of models and can be effectively implemented by deploying hardware accelerators such as GPUs. As a proof-of-concept, we apply this method to the determination of the partonic substructure of the proton within the NNPDF framework and demonstrate the robustness of the resultant model uncertainty estimates. The new GPU-optimised NNPDF code results in a speed-up of up to two orders of magnitude, a stabilisation of the memory requirements, and a reduction in energy consumption of up to 90% as compared to sequential CPU-based model training. While focusing on proton structure, our method is fully general and is applicable to any deep learning problem relying on hyperparameter optimisation for an ensemble of models., Comment: 27 pages, 7 figures

Disordered systems theory provides powerful tools to analyze the generic behaviors of highdimensional systems, such as species-rich ecological communities or neural networks. By assuming randomness in their interactions, universality ensures that many microscopic details are irrelevant to system-wide dynamics; but the choice of a random ensemble still limits the generality of results. We show here, in the context of ecological dynamics, that these analytical tools do not require a specific choice of ensemble, and that solutions can be found based only on a fundamental rotational symmetry in the interactions, encoding the idea that traits can be recombined into new species without altering global features. Dynamical outcomes then depend on the spectrum of the interaction matrix as a free parameter, allowing us to bridge between results found in different models of interactions, and extend beyond them to previously unidentified behaviors. The distinctive feature of ecological models is the possibility of species extinctions, which leads to an increased universality of dynamics as the fraction of extinct species increases. We expect that these findings can inform new developments in theoretical ecology as well as for other families of complex systems., Comment: 12 pages, 3 figures

We present NNPDF4.0MC, a variant of the NNPDF4.0 set of parton distributions (PDFs) at LO, NLO and NNLO, with and without inclusion of the photon PDF, suitable for use with Monte Carlo (MC) event generators, which require PDFs to satisfy additional constraints in comparison to standard PDF sets. These requirements include PDF positivity down to a low scale $Q\sim 1$ GeV, smooth extrapolation in the very small and large $x$ regions, and numerically stable results even in extreme regions of phase space for all PDFs. We compare the NNPDF4.0MC PDFs to their baseline NNPDF4.0 counterparts, and to the NNPDF2.3LO set entering the Monash tune of the Pythia8 event generator. We briefly assess the phenomenological impact of these PDFs on the cross-sections for hard and soft QCD processes at the LHC., Comment: 14 pages, 9 figures, PDF sets available from https://nnpdf.mi.infn.it/nnpdf4-0-mc/ ; version to be published in jhep

The student-optimal stable mechanism (DA), the most popular mechanism in school choice, is the only one that is both stable and strategy-proof. However, when DA is implemented, a student can change the schools of others without changing her own. We show that this drawback is limited: a student cannot change her classmates without modifying her school. We refer to this new property as {\it{local non-bossiness}} and use it to provide a new characterization of DA that does not rely on stability. Furthermore, we show that local non-bossiness plays a crucial role in providing incentives to be truthful when students have preferences over their colleagues. As long as students first consider the school to which they are assigned and then their classmates, DA induces the only stable and strategy-proof mechanism. There is limited room to expand this preference domain without compromising the existence of a stable and strategy-proof mechanism.

We extend the existing leading (LO), next-to-leading (NLO), and next-to-next-to-leading order (NNLO) NNPDF4.0 sets of parton distribution functions (PDFs) to approximate next-to-next-to-next-to-leading order (aN$^3$LO). We construct an approximation to the N$^3$LO splitting functions that includes all available partial information from both fixed-order computations and from small and large $x$ resummation, and estimate the uncertainty on this approximation by varying the set of basis functions used to construct the approximation. We include known N$^3$LO corrections to deep-inelastic scattering structure functions and extend the FONLL general-mass scheme to $\mathcal{O}\left( \alpha_s^3\right)$ accuracy. We determine a set of aN$^3$LO PDFs by accounting both for the uncertainty on splitting functions due to the incomplete knowledge of N$^3$LO terms, and to the uncertainty related to missing higher corrections (MHOU), estimated by scale variation, through a theory covariance matrix formalism. We assess the perturbative stability of the resulting PDFs, we study the impact of MHOUs on them, and we compare our results to the aN$^3$LO PDFs from the MSHT group. We examine the phenomenological impact of aN$^3$LO corrections on parton luminosities at the LHC, and give a first assessment of the impact of aN$^3$LO PDFs on the Higgs and Drell-Yan total production cross-sections. We find that the aN$^3$LO NNPDF4.0 PDFs are consistent within uncertainties with their NNLO counterparts, that they improve the description of the global dataset and the perturbative convergence of Higgs and Drell-Yan cross-sections, and that MHOUs on PDFs decrease substantially with the increase of perturbative order., Comment: 63 pages, 41 figures. Final version, as published in EPJC. Various clarifications added. Fig 2.9 and Tab. 4.6, 5.1 and 5.2 added. In particular, comparison to approximate YR4 estimate of PDF uncertainties added (Tab 4.1, 4.2 and associate discussion in sect. 5)

We include uncertainties due to missing higher order corrections to QCD computations (MHOU) used in the determination of parton distributions (PDFs) in the recent NNPDF4.0 set of PDFs. We use our previously published methodology, based on the treatment of MHOUs and their full correlations through a theory covariance matrix determined by scale variation, now fully incorporated in the new NNPDF theory pipeline. We assess the impact of the inclusion of MHOUs on the NNPDF4.0 central values and uncertainties, and specifically show that they lead to improved consistency of the PDF determination with an ensuing moderate reduction of PDF uncertainties at NNLO., Comment: 23 pages, 6 figures. Final version, as published in EPJC. Several small improvements; expanded discussion of the phi indicator in sect 4.2

We construct a set of parton distribution functions (PDFs), based on the recent NNPDF4.0 PDF set, that also include a photon PDF. The photon PDF is constructed using the LuxQED formalism, while QED evolution accounting for O(alpha), O(alpha alphas) and O(alpha^2) corrections is implemented and benchmarked by means of the EKO code. We investigate the impact of QED effects on NNPDF4.0, and compare our results both to our previous NNPDF3.1QED PDF set and to other recent PDF sets that include the photon. We assess the impact of photon-initiated processes and electroweak corrections on a variety of representative LHC processes, and find that they can reach the 5% level in vector boson pair production at large invariant mass., Comment: 32 pages, 19 figures. Final; version, as published on EPJC. Pheno discussion in Sect 4.2 improved, a few typos corrected

We provide a first quantitative indication that the wave function of the proton contains unequal distributions of charm quarks and antiquarks, i.e. a nonvanishing intrinsic valence charm distribution. A significant nonvanishing valence component cannot be perturbatively generated, hence our results reinforce previous evidence that the proton contains an intrinsic (i.e., not radiatively generated) charm quark component. We establish our result through a determination of the parton distribution functions (PDFs) of charm quarks and antiquarks in the proton. We propose two novel experimental probes of this intrinsic charm valence component: D-meson asymmetries in Z+c-jet production at the LHCb experiment, and flavor-tagged structure functions at the Electron-Ion Collider., Comment: 11 pages, 11 figures. PDF files available for download in LHAPDF format from https://nnpdf.mi.infn.it/nnpdf4-0-charm-asymmetry-study/ Final version, to be published in PRD (letters). Figures 2, 6 and 7 added. Fit without imposing the charm valence sum rule added. Various small clarifications

We formulate a one-center non-orthogonal configuration interaction singles (1C-NOCIS) theory for the computation of core excited states of an initial singlet state with two unpaired electrons. This model, which we refer to as 1C-NOCIS two-electron open-shell (2eOS), is appropriate for computing the K-edge near-edge X-ray absorption spectra (NEXAS) of the valence excited states of closed-shell molecules relevant to pump-probe time-resolved (TR) NEXAS experiments. With inclusion of core hole relaxation effects and explicit spin adaptation, 1C-NOCIS 2eOS requires mild shifts to match experiment, is free of artifacts due to spin contamination, and can capture the high-energy region of the spectrum beyond the transitions into the singly occupied molecular orbitals (SOMO). Calculations on water and thymine illustrate the different key features of excited-state NEXAS, namely the core-to-SOMO transition as well as shifts and spin-splittings in the transitions analogous to those of the ground state. Finally, simulations of the TR-NEXAS of acetylacetone after excitation onto its pi to pi-star singlet excited state at the carbon K-edge - an experiment carried out recently - showcases the ability of 1C-NOCIS 2eOS to efficiently simulate NEXAS based on non-adiabatic molecular dynamics simulations., Comment: Manuscript: 42 pages, 6 figures. SI: 48 pages, 10 figures

Wind speed and direction variations across the rotor affect power production. As utility-scale turbines extend higher into the atmospheric boundary layer (ABL) with larger rotor diameters and hub heights, they increasingly encounter more complex wind speed and direction variations. We assess three models for power production that account for wind speed and direction shear. Two are based on actuator disc representations and the third is a blade element representation. We also evaluate the predictions from a standard power curve model that has no knowledge of wind shear. The predictions from each model, driven by wind profile measurements from a profiling LiDAR, are compared to concurrent power measurements from an adjacent utility-scale wind turbine. In the field measurements of the utility-scale turbine, discrete combinations of speed and direction shear induce changes in power production of -19% to +34% relative to the turbine power curve for a given hub height wind speed. Positive speed shear generally corresponds to over-performance and positive direction shear to under-performance, relative to the power curve. Overall, the blade element model produces both higher correlation and lower error relative to the other models, but its quantitative accuracy depends on induction and controller sub-models. To further assess the influence of complex, non-monotonic wind profiles, we also drive the models with best-fit power law wind speed profiles and linear wind direction profiles. These idealized inputs produce qualitative and quantitative differences in power predictions from each model, demonstrating that time-varying, non-monotonic wind shear affects wind power production., Comment: 28 pages, 14 figures in main body, 1 figure in Appendix A, to be published in Wiley Wind Energy

Proton-proton collisions at the LHC generate a high-intensity collimated beam of neutrinos in the forward (beam) direction, characterised by energies of up to several TeV. The recent observation of LHC neutrinos by FASER$\nu$ and SND@LHC signals that this hitherto ignored particle beam is now available for scientific inquiry. Here we quantify the impact that neutrino deep-inelastic scattering (DIS) measurements at the LHC would have on the parton distributions (PDFs) of protons and heavy nuclei. We generate projections for DIS structure functions for FASER$\nu$ and SND@LHC at Run III, as well as for the FASER$\nu$2, AdvSND, and FLArE experiments to be hosted at the proposed Forward Physics Facility (FPF) operating concurrently with the High-Luminosity LHC (HL-LHC). We determine that up to one million electron- and muon-neutrino DIS interactions within detector acceptance can be expected by the end of the HL-LHC, covering a kinematic region in $x$ and $Q^2$ overlapping with that of the Electron-Ion Collider. Including these DIS projections into global (n)PDF analyses, specifically PDF4LHC21, NNPDF4.0, and EPPS21, reveals a significant reduction of PDF uncertainties, in particular for strangeness and the up and down valence PDFs. We show that LHC neutrino data enables improved theoretical predictions for core processes at the HL-LHC, such as Higgs and weak gauge boson production. Our analysis demonstrates that exploiting the LHC neutrino beam effectively provides CERN with a "Neutrino-Ion Collider" without requiring modifications in its accelerator infrastructure., Comment: 42 pages, 22 figures. Added detailed studies on fluxes uncertainties

The quantum chemistry community has developed analytic forces for approximate electronic excited states to enable walking on excited state potential energy surfaces (PES). One can thereby computationally characterize excited state minima and saddle points. Always implicit in using this machinery is the fact that an excited state PES only exists within the realm of the Born-Oppenheimer approximation, where the nuclear and electronic degrees of freedom separate. This work demonstrates through ab initio calculations and simple nonadiabatic dynamics that some excited state minimum structures are fantastical: they appear to exist as stable configurations only as a consequence of the PES construct, rather than being physically observable. One such case is the S2 excited state of phosphine and a second case are local minima of a number of states of tris(bipyridine)ruthenium(II). Each fantastical structure exhibits an unphysically high predicted harmonic frequency and associated force constant. This fact can serve as a valuable diagnostic of when an optimized excited state structure is non-observable. Their origin lies in the coupling between different electronic states, and the resulting avoided crossings. The upper state may exhibit a minimum very close to the crossing, where the force constant relates to the strength of the electronic coupling rather than to any characteristic excited state vibration. Nonadiabatic dynamics results using a Landau-Zener model illustrate that fantastical excited state structures have extremely short lifetimes on the order of a few femtoseconds. Their appearance in a calculation signals the presence of a nearby avoided crossing or conical intersection through which the system will rapidly cross to a lower surface., Comment: 7 pages, 4 figures

In this work we present a novel strategy to evaluate multi-variable integrals with quantum circuits. The procedure first encodes the integration variables into a parametric circuit. The obtained circuit is then derived with respect to the integration variables using the parameter shift rule technique. The observable representing the derivative is then used as the predictor of the target integrand function following a quantum machine learning approach. The integral is then estimated using the fundamental theorem of integral calculus by evaluating the original circuit. Embedding data according to a reuploading strategy, multi-dimensional variables can be easily encoded into the circuit's gates and then individually taken as targets while deriving the circuit. These techniques can be exploited to partially integrate a function or to quickly compute parametric integrands within the training hyperspace., Comment: Code available at https://github.com/qiboteam/QiNNtegrate, version that appeared in journal

Wet chemistry C-S-H precipitation experiments were performed under controlled conditions of solution supersaturation in the presence and absence of gluconate and three hexitol molecules. Characterization of the precipitates with SAXS and cryo-TEM experiments confirmed the presence of a multi-step nucleation pathway. Induction times for the formation of the amorphous C-S-H spheroids were determined from light transmittance. Analysis of those data with the classical nucleation theory revealed a significant increase of the kinetic prefactor in the same order as the complexation constants of calcium and silicate with each of the organics. Finally, two distinct precipitation regimes of the C-S-H amorphous precursor were identified: i) a nucleation regime at low saturation indexes (SI) and ii) a spinodal nucleation regime at high SI where the free energy barrier to the phase transition is found to be of the order of the kinetic energy or less., Comment: Accepted in Cement and Concrete Research. 30 pages plus supplementary materials. arXiv admin note: substantial text overlap with arXiv:2111.02743

Let $G$ be a finite group, let $\pi$ be a set of primes and let $p$ be the smallest prime in $\pi$. In this work, we prove that $G$ possesses a normal and abelian Hall $\pi$-subgroup if and only if the probability that two random $\pi$-elements of $G$ commute is larger than $\frac{p^2+p-1}{p^3}$. We also prove that if $x$ is a $\pi$-element not lying in $O_{\pi}(G)$, then the proportion of $\pi$-elements commuting with $x$ is at most $1/p$., Comment: 18 pages, accepted in Mathematische Nachrichten, Theorem B has been improved

A reliable determination of probability density functions from data samples is still a relevant topic in scientific applications. In this work we investigate the possibility of defining an algorithm for density function estimation using adiabatic quantum computing. Starting from a sample of a one-dimensional distribution, we define a classical-to-quantum data embedding procedure which maps the empirical cumulative distribution function of the sample into time dependent Hamiltonian using adiabatic quantum evolution. The obtained Hamiltonian is then projected into a quantum circuit using the time evolution operator. Finally, the probability density function of the sample is obtained using quantum hardware differentiation through the parameter shift rule algorithm. We present successful numerical results for predefined known distributions and high-energy physics Monte Carlo simulation samples., Comment: 7 pages, 3 figures

Continuously comparing theory predictions to experimental data is a common task in analysis of particle physics such as fitting parton distribution functions (PDFs). However, typically, both the computation of scattering amplitudes and the evolution of candidate PDFs from the fitting scale to the process scale are non-trivial, computing intesive tasks. We develop a new stack of software tools that aim to facilitate the theory predictions by computing FastKernel (FK) tables that reduce the theory computation to a linear algebra operation. Specifically, I present PineAPPL, our workhorse for grid operations, EKO, a new DGLAP solver, and yadism, a new DIS library. Alongside, I review several projects that become available with the new tools., Comment: 5 pages, 2 figures; submitted to ACAT 2022 proceedings

We present a collection of tools automating the efficient computation of large sets of theory predictions for high-energy physics. Calculating predictions for different processes often require dedicated programs. These programs, however, accept inputs and produce outputs that are usually very different from each other. The industrialization of theory predictions is achieved by a framework which harmonizes inputs (runcard, parameter settings), standardizes outputs (in the form of grids), produces reusable intermediate objects, and carefully tracks all meta data required to reproduce the computation. Parameter searches and fitting of non-perturbative objects are exemplary use cases that require a full or partial re-computation of theory predictions and will thus benefit of such a toolset. As an example application we present a study of the impact of replacing NNLO QCD K-factors in a PDF fit with the exact NNLO predictions., Comment: Documentation and code at https://nnpdf.github.io/pineline ; updated to published version

Let $p$ be a prime and let $G$ be a finite group such that the smallest prime that divides $|G|$ is $p$. We find sharp bounds, depending on $p$, for the commuting probability and the average character degree to guarantee that $G$ is nilpotent or supersolvable., Comment: 15 pages

We use physics-informed neural networks (PINNs) to compute the first quasi-normal modes of the Kerr geometry via the Teukolsky equation. This technique allows us to extract the complex frequencies and separation constants of the equation without the need for sophisticated numerical techniques, and with an almost immediate implementation under the \texttt{PyTorch} framework. We are able to compute the oscillation frequencies and damping times for arbitrary black hole spins and masses, with accuracy typically below the percentual level as compared to the accepted values in the literature. We find that PINN-computed quasi-normal modes are indistinguishable from those obtained through existing methods at signal-to-noise ratios (SNRs) larger than 100, making the former reliable for gravitational-wave data analysis in the mid term, before the arrival of third-generation detectors like LISA or the Einstein Telescope, where SNRs of ${\cal O}(1000)$ might be achieved., Comment: 12 pages, 7 figures. v2: Matches published version, minor typos corrected

In this proceedings we demonstrate some advantages of a top-bottom approach in the development of hardware-accelerated code. We start with an autogenerated hardware-agnostic Monte Carlo generator, which is parallelized in the event axis. This allow us to take advantage of the parallelizable nature of Monte Carlo integrals even if we don't have control of the hardware in which the computation will run (i.e., an external cluster). The generic nature of such an implementation can introduce spurious bottlenecks or overheads. Fortunately, said bottlenecks are usually restricted to a subset of operations and not to the whole vectorized program. By identifying the more critical parts of the calculation one can get very efficient code and at the same time minimize the amount of hardware-specific code that needs to be written. We show benchmarks demonstrating how simply reducing the memory footprint of the calculation can increase the performance of a $2 \to 4$ process., Comment: Proceedings ICHEP 2022, 6 pages

We respond to the criticism raised by Courtoy et al., in which the faithfulness of the NNPDF4.0 sampling is questioned and an under-estimate of the NNPDF4.0 PDF uncertainties is implied. We list, correct, and clarify in detail a number of inaccurate or misleading claims that are made in this Reference. Specifically, we explain and explicitly demonstrate why the central value of the PDF distribution does not generally coincide with the absolute minimum of the $\chi^{2}$ to the data. We examine some PDFs that have been constructed in the above study and claimed to be "good solutions": we show that similar PDFs are found with the NNPDF methodology, but with very low probability., Comment: 18 pages, 12 figures

Fitting PDFs requires the integration of a broad range of datasets, both from data and theory side, into a unique framework. While for data the integration mainly consists in the standardization of the data format, for the theory predictions there are multiple ingredients involved. Different providers are developed by separate groups for different processes, with a variety of inputs (runcards) and outputs (interpolation grids). Moreover, since processes are measured at different scales, DGLAP evolution has to be provided for the PDF candidate, or precomputed into the grids. We are working towards the automation of all these steps in a unique framework, that will be useful for any PDF fitting groups, and possibly also for phenomenological studies., Comment: 6 pages, 1 figure. Contribution to the Proceedings of the ICHEP 2022 Conference

The theory of the strong force, quantum chromodynamics, describes the proton in terms of quarks and gluons. The proton is a state of two up quarks and one down quark bound by gluons, but quantum theory predicts that in addition there is an infinite number of quark-antiquark pairs. Both light and heavy quarks, whose mass is respectively smaller or bigger than the mass of the proton, are revealed inside the proton in high-energy collisions. However, it is unclear whether heavy quarks also exist as a part of the proton wavefunction, which is determined by non-perturbative dynamics and accordingly unknown: so-called intrinsic heavy quarks. It has been argued for a long time that the proton could have a sizable intrinsic component of the lightest heavy quark, the charm quark. Innumerable efforts to establish intrinsic charm in the proton have remained inconclusive. Here we provide evidence for intrinsic charm by exploiting a high-precision determination of the quark-gluon content of the nucleon based on machine learning and a large experimental dataset. We disentangle the intrinsic charm component from charm-anticharm pairs arising from high-energy radiation. We establish the existence of intrinsic charm at the 3-standard-deviation level, with a momentum distribution in remarkable agreement with model predictions. We confirm these findings by comparing to very recent data on Z-boson production with charm jets from the LHCb experiment., Comment: 10 pages, 4 figures + Supplementary Information, published version available in https://www.nature.com/articles/s41586-022-04998-2. v2: updated comparison with the MBC model in Fig. 1 right (thanks to Tim Hobbs for the suggestion)

Hybridization of multi-junction concentrator photovoltaics with single-junction flat plate solar cells (CPV/PV) can deliver the highest power output per module area of any PV technology. Conversion efficiencies up to 34.2% have been published under the AM1.5g spectrum at standard test conditions for the EyeCon module which combines Fresnel lenses and III-V four-junction solar cells with bifacial c-Si. We investigate here its energy yield and compare it to conventional CPV as well as flat plate PV. The advantage of the hybrid CPV/PV module is that it converts direct sunlight with the most advanced multi-junction cell technology, while accessing diffuse, lens-scattered and back side irradiance with a Si cell that also serves as the heat distributor for the concentrator cells. This article quantifies that hybrid bifacial CPV/PV modules are expected to generate a 25 - 35% higher energy yield with respect to their closest competitor in regions with a diffuse irradiance fraction around 50%. Additionally, the relative cost of electricity generated by hybrid CPV/PV technology was calculated worldwide under certain economic assumptions. Therefore, this article gives clear guidance towards establishing competitive business cases for the technology.

The particle-hole interaction problem is longstanding within time-dependent density functional theory (TDDFT) and leads to extreme errors in the prediction of K-edge X-ray absorption spectra (XAS). We derive a linear-response formalism that uses optimized orbitals of the n-1-electron system as reference, building orbital relaxation and a proper hole into the initial density. Our approach is an exact generalization of the static-exchange approximation that ameliorates particle-hole interaction error associated with the adiabatic approximation and reduces errors in TDDFT XAS by orders of magnitude. With a statistical performance of just 0.5 eV root-mean-square error and the same computational scaling as TDDFT under the core-valence separation approximation, we anticipate that this approach will be of great utility in XAS calculations of large systems., Comment: 11 pages (main text), 12 pages (supporting information), 9 figures

In this work, we classify all finite groups such that for every field extension F of \mathbb{Q}, F is the field of values of at most 3 irreducible characters., Comment: 18 pages

We investigate the use of orbital-optimized references in conjunction with single-reference coupled-cluster theory with single and double substitutions (CCSD) for the study of core excitations and ionizations of 18 small organic molecules, without any use of response theory or equation-of-motion formalisms. Three schemes are employed to successfully address the convergence difficulties associated with the coupled-cluster equations, and the spin contamination resulting from the use of a spin symmetry-broken reference, in the case of excitations. In order to gauge the inherent potential of the methods studied, an effort is made to provide reasonable basis set limit estimates for the transition energies. Overall, we find that the two best-performing schemes studied here for Delta-CCSD are capable of predicting excitation and ionization energies with errors comparable to experimental accuracies. The proposed Delta-CCSD schemes seem to fare better than the widely used equation-of-motion CCSD (EOM-CCSD) with core-valence separation protocol, with statistical errors being reduced by more than a factor of two when compared to FC-CVS-EOM-CCSD.

Wind turbines located in wind farms are operated to maximize only their own power production. Individual operation results in wake losses that reduce farm energy. In this study, we operate a wind turbine array collectively to maximize total array production through wake steering. The selection of the farm control strategy relies on the optimization of computationally efficient flow models. We develop a physics-based, data-assisted flow control model to predict the optimal control strategy. In contrast to previous studies, we first design and implement a multi-month field experiment at a utility-scale wind farm to validate the model over a range of control strategies, most of which are suboptimal. The flow control model is able to predict the optimal yaw misalignment angles for the array within +/- 5 degrees for most wind directions (11-32% power gains). Using the validated model, we design a control protocol which increases the energy production of the farm in a second multi-month experiment by 2.7% and 1.0%, for the wind directions of interest and for wind speeds between 6 and 8 m/s and all wind speeds, respectively. The developed and validated predictive model can enable a wider adoption of collective wind farm operation., Comment: 14 pages, 5 figures, 1 table

We construct locally Lindel\"of scattered P-spaces (LLSP spaces, in short) with prescribed widths and heights under different set-theoretic assumptions. We prove that there is an LLSP space of width $\omega_1$ and height $\omega_2$ and that it is relatively consistent with ZFC that there is an LLSP space of width $\omega_1$ and height $\omega_3$. Also, we prove a stepping up theorem that, for every cardinal $\lambda \geq \omega_2$, permits us to construct from an LLSP space of width $\omega_1$ and height $\lambda$ satisfying certain additional properties an LLSP space of width $\omega_1$ and height $\alpha$ for every ordinal $\alpha < \lambda^+$. Then, we obtain as consequences of the above results the following theorems: (1) For every ordinal $\alpha < \omega_3$ there is an LLSP space of width $\omega_1$ and height $\alpha$. (2) It is relatively consistent with ZFC that there is an LLSP space of width $\omega_1$ and height $\alpha$ for every ordinal $\alpha < \omega_4$., Comment: 14 pages

Since the first determination of a structure function many decades ago, all methodologies used to determine structure functions or parton distribution functions (PDFs) have employed a common prefactor as part of the parametrization. The NNPDF collaboration pioneered the use of neural networks to overcome the inherent bias of constraining the space of solution with a fixed functional form while still keeping the same common prefactor as a preprocessing. Over the years various, increasingly sophisticated, techniques have been introduced to counter the effect of the prefactor on the PDF determination. In this paper we present a methodology to remove the prefactor entirely, thereby significantly simplifying the methodology, without a loss of efficiency and finding good agreement with previous results., Comment: 10 pages, 7 figures, final version published in EPJC

The present paper investigates the influence of gluconate and hexitol additives on the precipitation mechanism and kinetics of C-S-H. To this end, wet chemistry C-S-H precipitation experiments were performed under controlled conditions of solution supersaturation, under varying silicate concentration, while the transmittance of the solution was followed. This allowed determining induction times for the formation of C-S-H precursors in the presence and absence of gluconate and three hexitol molecules. Characterization of the precipitates was performed via small angle X-ray scattering and cryo-transmission electron microscopy experiments, which allowed the identification of a multi-step nucleation pathway also in the presence of the organics. Analysis of the induction time data in the framework of the classical nucleation theory revealed a significant increase of the kinetic pre-factor, which is associated to physical rates of cluster collision and aggregation during the nucleation process. Values of the kinetic pre-factor increase in the same order as the complexation constants of calcium and silicate with the each of the organics. This points to a retarding mechanism of crystallization related to steric hindrance of the aggregation of the early formed clusters via adsorption of the organics at their surfaces.

We present the software framework underlying the NNPDF4.0 global determination of parton distribution functions (PDFs). The code is released under an open source licence and is accompanied by extensive documentation and examples. The code base is composed by a PDF fitting package, tools to handle experimental data and to efficiently compare it to theoretical predictions, and a versatile analysis framework. In addition to ensuring the reproducibility of the NNPDF4.0 (and subsequent) determination, the public release of the NNPDF fitting framework enables a number of phenomenological applications and the production of PDF fits under user-defined data and theory assumptions., Comment: 14 pages, 4 figures. A companion paper describes the NNPDF4.0 PDFs

We present a new set of parton distribution functions (PDFs) based on a fully global dataset and machine learning techniques: NNPDF4.0. We expand the NNPDF3.1 determination with 44 new datasets, mostly from the LHC. We derive a novel methodology through hyperparameter optimisation, leading to an efficient fitting algorithm built upon stochastic gradient descent. We use NNLO QCD calculations and account for NLO electroweak corrections and nuclear uncertainties. Theoretical improvements in the PDF description include a systematic implementation of positivity constraints and integrability of sum rules. We validate our methodology by means of closure tests and "future tests" (i.e. tests of backward and forward data compatibility), and assess its stability, specifically upon changes of PDF parametrization basis. We study the internal compatibility of our dataset, and investigate the dependence of results both upon the choice of input dataset and of fitting methodology. We perform a first study of the phenomenological implications of NNPDF4.0 on representative LHC processes. The software framework used to produce NNPDF4.0 is made available as an open-source package together with documentation and examples., Comment: 139 pages, more than 60 figures. A companion paper describes the open-source NNPDF code. Figs 9.1 and 9.3 replaced after correction of a plotting bug (thanks to J.Huston and G.Salam for spotting it). A few more typos corrected and a reference updated. Final published version. References updated, several typos corrected, 72 issues raised by the referee addressed

We present MadFlow, a first general multi-purpose framework for Monte Carlo (MC) event simulation of particle physics processes designed to take full advantage of hardware accelerators, in particular, graphics processing units (GPUs). The automation process of generating all the required components for MC simulation of a generic physics process and its deployment on hardware accelerator is still a big challenge nowadays. In order to solve this challenge, we design a workflow and code library which provides to the user the possibility to simulate custom processes through the MadGraph5_aMC@NLO framework and a plugin for the generation and exporting of specialized code in a GPU-like format. The exported code includes analytic expressions for matrix elements and phase space. The simulation is performed using the VegasFlow and PDFFlow libraries which deploy automatically the full simulation on systems with different hardware acceleration capabilities, such as multi-threading CPU, single-GPU and multi-GPU setups. The package also provides an asynchronous unweighted events procedure to store simulation results. Crucially, although only Leading Order is automatized, the library provides all ingredients necessary to build full complex Monte Carlo simulators in a modern, extensible and maintainable way. We show simulation results at leading-order for multiple processes on different hardware configurations., Comment: 7 pages, 4 figures, to appear at Off-Shell 2021, code available at https://github.com/N3PDF/madflow. arXiv admin note: text overlap with arXiv:2105.10529

In this proceedings we present MadFlow, a new framework for the automation of Monte Carlo (MC) simulation on graphics processing units (GPU) for particle physics processes. In order to automate MC simulation for a generic number of processes, we design a program which provides to the user the possibility to simulate custom processes through the MadGraph5_aMC@NLO framework. The pipeline includes a first stage where the analytic expressions for matrix elements and phase space are generated and exported in a GPU-like format. The simulation is then performed using the VegasFlow and PDFFlow libraries which deploy automatically the full simulation on systems with different hardware acceleration capabilities, such as multi-threading CPU, single-GPU and multi-GPU setups. We show some preliminary results for leading-order simulations on different hardware configurations., Comment: 6 pages, 3 figures, to appear in the vCHEP 2021 conference proceedings

We present a compression algorithm for parton densities using synthetic replicas generated from the training of a Generative Adversarial Network (GAN). The generated replicas are used to further enhance the statistics of a given Monte Carlo PDF set prior to compression. This results in a compression methodology that is able to provide a compressed set with smaller number of replicas and a more adequate representation of the original probability distribution. We also address the question of whether the GAN could be used as an alternative mechanism to avoid the fitting of large number of replicas., Comment: 16 pages, code at https://github.com/N3PDF/pycompressor, v2: Final version to be published in EPJC. Modified Fig. 9 and fixed typos in Sec. 6

We discuss a test of the generalization power of the methodology used in the determination of parton distribution functions (PDFs). The "future test" checks whether the uncertainty on PDFs, in regions in which they are not constrained by current data, are compatible with future data. The test is performed by using the current optimized methodology for PDF determination, but with a limited dataset, as available in the past, and by checking whether results are compatible within uncertainty with the result found using a current more extensive dataset. We use the future test to assess the generalization power of the NNPDF4.0 unpolarized PDF and the NNPDFpol1.1 polarized PDF methodology. Specifically, we investigate whether the former would predict the rise of the unpolarized proton structure function $F_2$ at small $x$ using only pre HERA data, and whether the latter would predict the so-called "proton spin crisis" using only pre-EMC data., Comment: 29 pages, 7 figures. Prepared for the 60th anniversary of the Cracow School of Theoretical Physics, to be published in the proceedings

We present PDFFlow, a new software for fast evaluation of parton distribution functions (PDFs) designed for platforms with hardware accelerators. PDFs are essential for the calculation of particle physics observables through Monte Carlo simulation techniques. The evaluation of a generic set of PDFs for quarks and gluons at a given momentum fraction and energy scale requires the implementation of interpolation algorithms as introduced for the first time by the LHAPDF project. PDFFlow extends and implements these interpolation algorithms using Google's TensorFlow library providing the possibility to perform PDF evaluations taking fully advantage of multi-threading CPU and GPU setups. We benchmark the performance of this library on multiple scenarios relevant for the particle physics community., Comment: 6 pages, 6 figures. Code available at "https://github.com/N3PDF/pdfflow". Refer also to arXiv:2009.06635

We present a first attempt to design a quantum circuit for the determination of the parton content of the proton through the estimation of parton distribution functions (PDFs), in the context of high energy physics (HEP). The growing interest in quantum computing and the recent developments of new algorithms and quantum hardware devices motivates the study of methodologies applied to HEP. In this work we identify architectures of variational quantum circuits suitable for PDFs representation (qPDFs). We show experiments about the deployment of qPDFs on real quantum devices, taking into consideration current experimental limitations. Finally, we perform a global qPDF determination from collider data using quantum computer simulation on classical hardware and we compare the obtained partons and related phenomenological predictions involving hadronic processes to modern PDFs., Comment: 13 pages, 14 figures, 4 tables, accepted for publication in PRD, code available at https://github.com/Quantum-TII/qibo

In this work we demonstrate the usage of the VegasFlow library on multidevice situations: multi-GPU in one single node and multi-node in a cluster. VegasFlow is a new software for fast evaluation of highly parallelizable integrals based on Monte Carlo integration. It is inspired by the Vegas algorithm, very often used as the driver of cross section integrations and based on Google's powerful TensorFlow library. In this proceedings we consider a typical multi-GPU configuration to benchmark how different batch sizes can increase (or decrease) the performance on a Leading Order example integration., Comment: 7 pages, ICHEP2020 proceedings, updated text

In this proceedings we demonstrate how to implement and construct the PineAPPL grids, designed for fast-interpolation of Monte Carlo simulation with electroweak and QCD corrections, using the VegasFlow framework for Monte Carlo simulation on hardware accelerators. We provide an example of synchronous and asynchronous filling operations of PineAPPL grids from Monte Carlo events generated by VegasFlow. We compare the performance of this procedure on multithreading CPU and GPU., Comment: 5 pages, 2 figures, LHCP2020 proceedings

We present PDFFlow, a new software for fast evaluation of parton distribution functions (PDFs) designed for platforms with hardware accelerators. PDFs are essential for the calculation of particle physics observables through Monte Carlo simulation techniques. The evaluation of a generic set of PDFs for quarks and gluon at a given momentum fraction and energy scale requires the implementation of interpolation algorithms as introduced for the first time by the LHAPDF project. PDFFlow extends and implements these interpolation algorithms using Google's TensorFlow library providing the capabilities to perform PDF evaluations taking fully advantage of multi-threading CPU and GPU setups. We benchmark the performance of this library on multiple scenarios relevant for the particle physics community., Comment: 8 pages, 7 figures, 2 tables. Code available at https://github.com/N3PDF/pdfflow

We investigate the rate of orbital orientation dilution of young stellar clusters in the vicinity of supermassive black holes. Within the framework of vector resonant relaxation, we predict the time evolution of the two-point correlation function of the stellar orbital plane orientations as a function of their initial angular separation and diversity in orbital parameters (semi-major axis, eccentricity). As expected, the larger the spread in initial orientations and orbital parameters, the more efficient the dilution of a given set of co-eval stars, with a characteristic timescale set up by the coherence time of the background potential fluctuations. A Markovian prescription which matches numerical simulations allows us to efficiently probe the underlying kinematic properties of the unresolved nucleus when requesting consistency with a given dilution efficiency, imposed by the observed stellar disc within the one arcsecond of Sgr A*. As a proof of concept, we compute maps of constant dilution times as a function of the semi major axis cusp index and fraction of intermediate mass black holes in the old background stellar cluster. This computation suggests that vector resonant relaxation should prove useful in this context since it impacts orientations on timescales comparable to the stars' age., Comment: 20 pages, 16 figures, submitted to MNRAS

The intentional yaw misalignment of leading, upwind turbines in a wind farm, termed wake steering, has demonstrated potential as a collective control approach for wind farm power maximization. The optimal control strategy, and resulting effect of wake steering on wind farm power production, are in part dictated by the power degradation of the upwind yaw misaligned wind turbines. In the atmospheric boundary layer, the wind speed and direction may vary significantly over the wind turbine rotor area, depending on atmospheric conditions and stability, resulting in freestream turbine power production which is asymmetric as a function of the direction of yaw misalignment and which varies during the diurnal cycle. In this study, we propose a model for the power production of a wind turbine in yaw misalignment based on aerodynamic blade elements which incorporates the effects of wind speed and direction changes over the turbine rotor area in yaw misalignment. A field experiment is performed using multiple utility-scale wind turbines to characterize the power production of yawed freestream operating turbines depending on the wind conditions, and the model is validated using the experimental data. The resulting power production of a yaw misaligned variable speed wind turbine depends on a nonlinear interaction between the yaw misalignment, the atmospheric conditions, and the wind turbine control system., Comment: 37 pages, 15 figures

In this paper, an exact algorithm in polynomial time is developed to solve unrestricted binary quadratic programs. The computational complexity is $O\left( n^{\frac{15}{2}}\right) $, although very conservative, it is sufficient to prove that this minimization problem is in the complexity class $P$. The implementation aspects are also described in detail with a special emphasis on the transformation of the quadratic program into a linear program that can be solved in polynomial time. The algorithm was implemented in MATLAB and checked by generating five million matrices of arbitrary dimensions up to 30 with random entries in the range $\left[ -50,50\right] $. All the experiments carried out have revealed that the method works correctly.

We present VegasFlow, a new software for fast evaluation of high dimensional integrals based on Monte Carlo integration techniques designed for platforms with hardware accelerators. The growing complexity of calculations and simulations in many areas of science have been accompanied by advances in the computational tools which have helped their developments. VegasFlow enables developers to delegate all complicated aspects of hardware or platform implementation to the library so they can focus on the problem at hand. This software is inspired on the Vegas algorithm, ubiquitous in the particle physics community as the driver of cross section integration, and based on Google's powerful TensorFlow library. We benchmark the performance of this library on many different consumer and professional grade GPUs and CPUs., Comment: 6 pages, 5 figures, final version published in CPC

Parton Distribution Functions (PDFs) model the parton content of the proton. Among the many collaborations which focus on PDF determination, NNPDF pioneered the use of Neural Networks to model the probability of finding partons (quarks and gluons) inside the proton with a given energy and momentum. In this proceedings we make use of state of the art techniques to modernize the NNPDF methodology and study different models and optimizers in order to improve the quality of the PDF: improving both the quality and efficiency of the fits. We also present the evolutionary_keras library, a Keras implementation of the Evolutionary Algorithms used by NNPDF., Comment: Proceedings for Artificial Intelligence for Science, Industry and Society 2019

In this proceedings we describe the computational challenges associated to the determination of parton distribution functions (PDFs). We compare the performance of the convolution of the parton distributions with matrix elements using different hardware instructions. We quantify and identify the most promising data-model configurations to increase PDF fitting performance in adapting the current code frameworks to hardware accelerators such as graphics processing units., Comment: 6 pages, 2 figures, 3 tables, in proceedings of PHOTON 2019