arXiv subjects

Antonio Rago

Publications and source records attributed to Antonio Rago.

At least 73 records · Page 4Linked to original sources

arXiv2023

Gauge generation and dissemination in OpenLat

In this contribution we report the status and plans of the open lattice initiative to generate and share new gauge ensembles using the stabilised Wilson fermion framework. The production strategy is presented in terms of a three stage plan alongside summaries of the data management as well as access policies. Current progress in completing the first stage of generating ensembles at four lattice spacings at the flavor symmetric point is given.

hep-lat↗

arXiv2023

Benchmark Continuum Limit Results for Spectroscopy with Stabilized Wilson Fermions

The OpenLat initiative presents its results of lattice QCD simulations using Stabilized Wilson Fermions (SWF) using 2+1 quark flavors. Focusing on the $\mathrm{SU}(3)$ flavor symmetric point $m_π=m_K=412$ MeV, four different lattice spacings ($a=0.064,0.077,0.094,0.12$ fm) are used to perform the continuum limit to study cutoff effects. We present results on light hadron masses; for the determination we use a Bayesian analysis framework with constraints and model averaging to minimize the bias in the analysis.

hep-lat↗

arXiv2023

Hadronic observables from master-field simulations

Substantial progress has been made recently in the generation of master-field ensembles. This has to be paired with efficient techniques to compute observables on gauge field configurations with a large volume. Here we present the results of the computation of hadronic observables, including hadron masses and meson decay constants, on large-volume and master-field ensembles with physical volumes of up to $(18\,\mathrm{fm})^4$ and $m_πL$ up to $25$, simulated using $N_{\mathrm{f}}=2+1$ stabilized Wilson fermions. We obtain sub-percent determinations from single gauge configurations with the combined use of position-space techniques, volume averages and master-field error estimation.

hep-lat↗

arXiv2022

Formalising the Robustness of Counterfactual Explanations for Neural Networks

The use of counterfactual explanations (CFXs) is an increasingly popular explanation strategy for machine learning models. However, recent studies have shown that these explanations may not be robust to changes in the underlying model (e.g., following retraining), which raises questions about their reliability in real-world applications. Existing attempts towards solving this problem are heuristic, and the robustness to model changes of the resulting CFXs is evaluated with only a small number of retrained models, failing to provide exhaustive guarantees. To remedy this, we propose Δ-robustness, the first notion to formally and deterministically assess the robustness (to model changes) of CFXs for neural networks. We introduce an abstraction framework based on interval neural networks to verify the Δ-robustness of CFXs against a possibly infinite set of changes to the model parameters, i.e., weights and biases. We then demonstrate the utility of this approach in two distinct ways. First, we analyse the Δ-robustness of a number of CFX generation methods from the literature and show that they unanimously host significant deficiencies in this regard. Second, we demonstrate how embedding Δ-robustness within existing methods can provide CFXs which are provably robust.

cs.LG↗

arXiv2022

Explaining Causal Models with Argumentation: the Case of Bi-variate Reinforcement

Causal models are playing an increasingly important role in machine learning, particularly in the realm of explainable AI. We introduce a conceptualisation for generating argumentation frameworks (AFs) from causal models for the purpose of forging explanations for the models' outputs. The conceptualisation is based on reinterpreting desirable properties of semantics of AFs as explanation moulds, which are means for characterising the relations in the causal model argumentatively. We demonstrate our methodology by reinterpreting the property of bi-variate reinforcement as an explanation mould to forge bipolar AFs as explanations for the outputs of causal models. We perform a theoretical evaluation of these argumentative explanations, examining whether they satisfy a range of desirable explanatory and argumentative properties.

cs.AI↗

arXiv2022

Forecasting Argumentation Frameworks

We introduce Forecasting Argumentation Frameworks (FAFs), a novel argumentation-based methodology for forecasting informed by recent judgmental forecasting research. FAFs comprise update frameworks which empower (human or artificial) agents to argue over time about the probability of outcomes, e.g. the winner of a political election or a fluctuation in inflation rates, whilst flagging perceived irrationality in the agents' behaviour with a view to improving their forecasting accuracy. FAFs include five argument types, amounting to standard pro/con arguments, as in bipolar argumentation, as well as novel proposal arguments and increase/decrease amendment arguments. We adapt an existing gradual semantics for bipolar argumentation to determine the aggregated dialectical strength of proposal arguments and define irrational behaviour. We then give a simple aggregation function which produces a final group forecast from rational agents' individual forecasts. We identify and study properties of FAFs and conduct an empirical evaluation which signals FAFs' potential to increase the forecasting accuracy of participants.

cs.AI↗

arXiv2022

Singlet channel scattering in a Composite Higgs model on the lattice

We present the first calculation of the scattering amplitude in the singlet channel beyond QCD. The calculation is performed in $SU(2)$ gauge theory with $N_f=2$ fundamental Dirac fermions and based on a finite-volume scattering formalism. The theory exhibits a $SU(4) \to Sp(4)$ chiral symmetry breaking pattern that is used to design minimal composite Higgs models currently tested at the LHC. Our results show that, for the range of underlying fermion mass considered, the lowest flavour singlet state is stable.

hep-lat↗

arXiv2022

Properties, ensembles and hadron spectra with Stabilised Wilson Fermions

In this joint contribution we announce the formation of the "OPEN LATtice initiative", https://openlat1.gitlab.io, to study Stabilised Wilson Fermions (SWF). They are a new avenue for QCD calculations with Wilson-type fermions and we report results on our continued study of this framework: Tuning the clover improvement coefficient, and extending the reach of lattice spacings to $a=0.12$ fm. We fix the flavor symmetric points $m_π=m_K=412$ MeV at $a=0.055,0.064, 0.077, 0.094, 0.12$ fm and define the trajectories to the physical point by fixing the trace of the quark mass matrix. Currently our pion mass range extends down to $m_π\sim200$ MeV. We outline our tuning goals and strategy as well as our future planned ensembles. First scaling studies are performed on $f_π$ and $m_π$. Additionally results of a preliminary continuum extrapolation of $m_N$ at the flavor symmetric point are presented. Going further a first determination of the light and strange hadron spectrum chiral dependence is shown, which serves to check the quality of the action for precision measurements. We also investigate other quantities such as flowed gauge observables to study how the continuum limit is approached. Taken together we observe the SWF enable us to perform stable lattice simulations across a large range of parameters in mass, volume and lattice spacing. Pooling resources our new initiative has made our reported progress possible and through it we will share generated gauge ensembles under an open science philosophy.

hep-lat↗

arXiv2021

Efficient computations of continuous action densities of states for lattice models

The Logarithmic Linear Relaxation (LLR) algorithm is an efficient method for computing densities of states for systems with a continuous spectrum. A key feature of this method is exponential error reduction, which allows us to evaluate the density of states of a system over hundreds of thousands of orders of magnitude with a fixed level of relative accuracy. As a consequence of exponential error reduction, the LLR method provides a robust alternative to traditional Monte Carlo calculations in cases in which states suppressed by the Boltzmann weight play nevertheless a relevant role, e.g., as transition regions between dominant configuration sets. After reviewing the algorithm, we will show an application in U(1) Lattice Gauge Theory that has enabled us to obtain the most accurate estimate of the critical coupling with modest computational resources, defeating exponential tunneling times between metastable vacua. As a further showcase, we will then present an application of the LLR method to the decorrelation of the topological charge in SU(3) Lattice Gauge Theory near the continuum limit. Finally, we will review in general applications of the LLR algorithm to systems affected by a strong sign problem and discuss the case of the Bose gas at finite chemical potential.

hep-lat↗

arXiv2021

Approaching the master-field: Hadronic observables in large volumes

The master-field approach to lattice QCD envisions performing calculations on a small number of large-volume gauge-field configurations. Substantial progress has been made recently in the generation of such fields, and this must be joined with measurement strategies that take advantage of the large volume. In these proceedings, we describe how to compute simple hadronic quantities efficiently and estimate their errors in the master-field approach, i.e. by studying cross-correlations of observables on a single configuration. We discuss the scaling of the uncertainty with the volume and compare extractions based on momentum-projected and position-space two-point functions. The latter show promising results, already at intermediate volumes, but come with additional technical complexities such as a more complicated manifestation of boundary effects, which we also address.

hep-lat↗

arXiv2021

Deep Argumentative Explanations

Despite the recent, widespread focus on eXplainable AI (XAI), explanations computed by XAI methods tend to provide little insight into the functioning of Neural Networks (NNs). We propose a novel framework for obtaining (local) explanations from NNs while providing transparency about their inner workings, and show how to deploy it for various neural architectures and tasks. We refer to our novel explanations collectively as Deep Argumentative eXplanations (DAXs in short), given that they reflect the deep structure of the underlying NNs and that they are defined in terms of notions from computational argumentation, a form of symbolic AI offering useful reasoning abstractions for explanation. We evaluate DAXs empirically showing that they exhibit deep fidelity and low computational cost. We also conduct human experiments indicating that DAXs are comprehensible to humans and align with their judgement, while also being competitive, in terms of user acceptance, with some existing approaches to XAI that also have an argumentative spirit.

cs.AI↗

arXiv2021

Argumentative XAI: A Survey

Explainable AI (XAI) has been investigated for decades and, together with AI itself, has witnessed unprecedented growth in recent years. Among various approaches to XAI, argumentative models have been advocated in both the AI and social science literature, as their dialectical nature appears to match some basic desirable features of the explanation activity. In this survey we overview XAI approaches built using methods from the field of computational argumentation, leveraging its wide array of reasoning abstractions and explanation delivery methods. We overview the literature focusing on different types of explanation (intrinsic and post-hoc), different models with which argumentation-based explanations are deployed, different forms of delivery, and different argumentation frameworks they use. We also lay out a roadmap for future work.

cs.AI↗

arXiv2021

Influence-Driven Explanations for Bayesian Network Classifiers

One of the most pressing issues in AI in recent years has been the need to address the lack of explainability of many of its models. We focus on explanations for discrete Bayesian network classifiers (BCs), targeting greater transparency of their inner workings by including intermediate variables in explanations, rather than just the input and output variables as is standard practice. The proposed influence-driven explanations (IDXs) for BCs are systematically generated using the causal relationships between variables within the BC, called influences, which are then categorised by logical requirements, called relation properties, according to their behaviour. These relation properties both provide guarantees beyond heuristic explanation methods and allow the information underpinning an explanation to be tailored to a particular context's and user's requirements, e.g., IDXs may be dialectical or counterfactual. We demonstrate IDXs' capability to explain various forms of BCs, e.g., naive or multi-label, binary or categorical, and also integrate recent approaches to explanations for BCs from the literature. We evaluate IDXs with theoretical and empirical analyses, demonstrating their considerable advantages when compared with existing explanation methods.

cs.AI↗

arXiv2021

Ergodic sampling of the topological charge using the density of states

In lattice calculations, the approach to the continuum limit is hindered by the severe freezing of the topological charge, which prevents ergodic sampling in configuration space. In order to significantly reduce the autocorrelation time of the topological charge, we develop a density of states approach with a smooth constraint and use it to study SU(3) pure Yang Mills gauge theory near the continuum limit. Our algorithm relies on simulated tempering across a range of couplings, which guarantees the decorrelation of the topological charge and ergodic sampling of topological sectors. Particular emphasis is placed on testing the accuracy, efficiency and scaling properties of the method. In their most conservative interpretation, our results provide firm evidence of a sizeable reduction of the exponent z related to the growth of the autocorrelation time as a function of the inverse lattice spacing.

hep-lat↗

arXiv2020

Master-field simulations of O(a)-improved lattice QCD: Algorithms, stability and exactness

In master-field simulations of lattice QCD, the expectation values of interest are obtained from a single or at most a few representative gauge-field configurations on very large lattices. If the light quarks are included, the generation of these fields using standard techniques is however challenging in view of various algorithmic instabilities and precision issues. Ways to overcome these problems are described here for the case of the O(a)-improved Wilson formulation of lattice QCD and the viability of the proposed measures is then checked in extensive simulations of the theory with 2+1 flavours of quarks.

hep-lat↗

arXiv2019

Computing general observables in lattice models with complex actions

The study of QFTs at finite density is hindered by the presence of the so-called sign problem. The action definition of such systems is, in fact, complex-valued making standard importance sampling Monte Carlo methods ineffective. In this work, we shall review the generalized density of states method for complex action systems and the Linear Logarithmic Relaxation algorithm (LLR). We will focus on the recent developments regarding the bias control of the LLR method and the evaluation of general observables in the DoS+LLR framework. Recent results on the well-known relativistic Bose gas will be presented, proving that in our approach the phase factor can be consistently evaluated over hundreds of orders of magnitude. A first exploratory study on the Thirring model in the DoS formalism will be presented as well.

hep-lat↗

arXiv2019

Free energy of the self-interacting relativistic lattice Bose gas at finite density

The density of state approach has recently been proposed as a potential route to circumvent the sign problem in systems at finite density. In this study, using the Linear Logarithmic Relaxation (LLR) algorithm, we extract the generalised density of states, which is defined in terms of the imaginary part of the action, for the self-interacting relativistic lattice Bose gas at finite density. After discussing the implementation and testing the reliability of our approach, we focus on the determination of the free energy difference between the full system and its phase-quenched counterpart. Using a set of lattices ranging from $4^4$ to $16^4$ , we show that in the low density phase, this overlap free energy can be reliably extrapolated to the thermodynamic limit. The numerical precision we obtain with the LLR method allows us to determine with sufficient accuracy the expectation value of the phase factor, which is used in the calculation of the overlap free energy, down to values of ${\cal O}(10^{-480})$. When phase factor measurements are extended to the dense phase, a change of behaviour of the overlap free energy is clearly visible as the chemical potential crosses a critical value. Using fits inspired by the approximate validity of mean-field theory, which is confirmed by our simulations, we extract the critical chemical potential as the non-analyticity point in the overlap free energy, obtaining a value that is in agreement with other determinations. Implications of our findings and potential improvements of our methodology are also discussed.

hep-lat↗

arXiv2019

The density of states approach to the sign problem

Approaches to the sign problem based on the density of states have been recently revived by the introduction of the LLR algorithm, which allows us to compute the density of states itself with exponential error reduction. In this work, after a review of the generalities of the method, we show recent results for the Bose gas in four dimensions, focussing on the identification of possible systematic errors and on methods of controlling the bias they can introduce in the calculation.

hep-lat↗