arXiv subjects

Michael Wilson

Publications and source records attributed to Michael Wilson.

27 records · Page 2Linked to original sources

arXiv2021

Machine Learning Trivializing Maps: A First Step Towards Understanding How Flow-Based Samplers Scale Up

A trivializing map is a field transformation whose Jacobian determinant exactly cancels the interaction terms in the action, providing a representation of the theory in terms of a deterministic transformation of a distribution from which sampling is trivial. Recently, a proof-of-principle study by Albergo, Kanwar and Shanahan [arXiv:1904.12072] demonstrated that approximations of trivializing maps can be `machine-learned' by a class of invertible, differentiable neural models called \textit{normalizing flows}. By ensuring that the Jacobian determinant can be computed efficiently, asymptotically exact sampling from the theory of interest can be performed by drawing samples from a simple distribution and passing them through the network. From a theoretical perspective, this approach has the potential to become more efficient than traditional Markov Chain Monte Carlo sampling techniques, where autocorrelations severely diminish the sampling efficiency as one approaches the continuum limit. A major caveat is that it is not yet understood how the size of models and the cost of training them is expected to scale. As a first step, we have conducted an exploratory scaling study using two-dimensional $ϕ^4$ with up to $20^2$ lattice sites. Although the scope of our study is limited to a particular model architecture and training algorithm, initial results paint an interesting picture in which training costs grow very quickly indeed. We describe a candidate explanation for the poor scaling, and outline our intentions to clarify the situation in future work.

hep-lat↗

arXiv2021

Bayesian Approach to Inverse Problems: an Application to NNPDF Closure Testing

We discuss the Bayesian approach to the solution of inverse problems and apply the formalism to analyse the closure tests performed by the NNPDF collaboration. Starting from a comparison with the approach that is currently used for the determination of parton distributions (PDFs) by the NNPDF collaboration, we discuss some analytical results that can be obtained for linear problems and use these results as a guidance for the more complicated non-linear problems. We show that, in the case of Gaussian distributions, the posterior probability density of the parametrized PDFs is fully determined by the results of the NNPDF fitting procedure. In the particular case that we consider, the fitting procedure and the Bayesian analysis yield exactly the same result. Building on the insight that we obtain from the analytical results, we introduce new estimators to assess the statistical faithfulness of the fit results in closure tests. These estimators are defined in data space, and can be studied analytically using the Bayesian formalism in a linear model in order to clarify their meaning. Finally we present numerical results from a number of closure tests performed with current NNPDF methodologies. These further tests allow us to validate the NNPDF4.0 methodology and provide a quantitative comparison of the NNPDF4.0 and NNPDF3.1 methodologies. As PDFs determinations move into precision territory, the need for a careful validation of the methodology becomes increasingly important: the error bar has become the focal point of contemporary PDFs determinations. In this perspective, theoretical assumptions and other sources of error are best formulated and analysed in the Bayesian framework, which provides an ideal language to address the precision and the accuracy of current fits.

hep-ph↗

arXiv2021

Efficient Modelling of Trivializing Maps for Lattice $ϕ^4$ Theory Using Normalizing Flows: A First Look at Scalability

General-purpose Markov Chain Monte Carlo sampling algorithms suffer from a dramatic reduction in efficiency as the system being studied is driven towards a critical point. Recently, a series of seminal studies suggested that normalizing flows - a class of deep generative models - can form the basis of a sampling strategy that does not suffer from this 'critical slowing down'. The central idea is to use machine learning techniques to build (approximate) trivializing maps, i.e. field transformations that map the theory of interest into a 'simpler' theory in which the degrees of freedom decouple, and where the statistical weight in the path integral is given by a distribution from which sampling is easy. No separate process is required to generate training data for such models, and convergence to the desired distribution is guaranteed through a reweighting procedure such as a Metropolis test. In a proof-of-principle demonstration on two-dimensional $ϕ^4$ theory, Albergo et al. (arXiv:1904.12072) modelled the trivializing map as a sequence of pointwise affine transformations. We pick up this thread, with the aim of quantifying how well we can expect this approach to scale as we increase the number of degrees of freedom in the system. We make several modifications to the original design that allow our models learn more efficient representations of trivializing maps using much smaller neural networks, which leads to a large reduction in the computational cost required to train models of equivalent quality. After making these changes, we find that sampling efficiency is almost entirely dictated by how extensively a model has been trained, while being unresponsive to further alterations that increase model flexibility. However, as we move towards the continuum limit the training costs scale extremely quickly, which urgently requires further work to fully understand and mitigate.

hep-lat↗

arXiv2021

An open-source machine learning framework for global analyses of parton distributions

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.

hep-ph↗

arXiv2021

A Topological Approach for Motion Track Discrimination

Detecting small targets at range is difficult because there is not enough spatial information present in an image sub-region containing the target to use correlation-based methods to differentiate it from dynamic confusers present in the scene. Moreover, this lack of spatial information also disqualifies the use of most state-of-the-art deep learning image-based classifiers. Here, we use characteristics of target tracks extracted from video sequences as data from which to derive distinguishing topological features that help robustly differentiate targets of interest from confusers. In particular, we calculate persistent homology from time-delayed embeddings of dynamic statistics calculated from motion tracks extracted from a wide field-of-view video stream. In short, we use topological methods to extract features related to target motion dynamics that are useful for classification and disambiguation and show that small targets can be detected at range with high probability.

cs.CV↗

arXiv2019

A First Determination of Parton Distributions with Theoretical Uncertainties

The parton distribution functions (PDFs) which characterize the structure of the proton are currently one of the dominant sources of uncertainty in the predictions for most processes measured at the Large Hadron Collider (LHC). Here we present the first extraction of the proton PDFs that accounts for the missing higher order uncertainty (MHOU) in the fixed-order QCD calculations used in PDF determinations. We demonstrate that the MHOU can be included as a contribution to the covariance matrix used for the PDF fit, and then introduce prescriptions for the computation of this covariance matrix using scale variations. We validate our results at next-to-leading order (NLO) by comparison to the known next order (NNLO) corrections. We then construct variants of the NNPDF3.1 NLO PDF set that include the effect of the MHOU, and assess their impact on the central values and uncertainties of the resulting PDFs.

hep-ph↗

arXiv2019

Parton Distributions with Theory Uncertainties: General Formalism and First Phenomenological Studies

We formulate a general approach to the inclusion of theoretical uncertainties, specifically those related to the missing higher order uncertainty (MHOU), in the determination of parton distribution functions (PDFs). We demonstrate how, under quite generic assumptions, theory uncertainties can be included as an extra contribution to the covariance matrix when determining PDFs from data. We then review, clarify, and systematize the use of renormalization and factorization scale variations as a means to estimate MHOUs consistently in deep inelastic and hadronic processes. We define a set of prescriptions for constructing a theory covariance matrix using scale variations, which can be used in global fits of data from a wide range of different processes, based on choosing a set of independent scale variations suitably correlated within and across processes. We set up an algebraic framework for the choice and validation of an optimal prescription by comparing the estimate of MHOU encoded in the next-to-leading order (NLO) theory covariance matrix to the observed shifts between NLO and NNLO predictions. We perform a NLO PDF determination which includes the MHOU, assess the impact of the inclusion of MHOUs on the PDF central values and uncertainties, and validate the results by comparison to the known shift between NLO and NNLO PDFs. We finally study the impact of the inclusion of MHOUs in a global PDF determination on LHC cross-sections, and provide guidelines for their use in precision phenomenology. In addition, we also compare the results based on the theory covariance matrix formalism to those obtained by performing PDF determinations based on different scale choices.

hep-ph↗

arXiv2018

Machine learning determination of dynamical parameters: The Ising model case

We train a set of Restricted Boltzmann Machines (RBMs) on one- and two-dimensional Ising spin configurations at various values of temperature, generated using Monte Carlo simulations. We validate the training procedure by monitoring several estimators, including measurements of the log-likelihood, with the corresponding partition functions estimated using annealed importance sampling. The effects of various choices of hyper-parameters on training the RBM are discussed in detail, with a generic prescription provided. Finally, we present a closed form expression for extracting the values of couplings, for every $n$-point interaction between the visible nodes of an RBM, in a binary system such as the Ising model. We aim at using this study as the foundation for further investigations of less well-known systems.

physics.comp-ph↗

arXiv2014

Accurate, fully-automated NMR spectral profiling for metabolomics

Many diseases cause significant changes to the concentrations of small molecules (aka metabolites) that appear in a person's biofluids, which means such diseases can often be readily detected from a person's "metabolic profile". This information can be extracted from a biofluid's NMR spectrum. Today, this is often done manually by trained human experts, which means this process is relatively slow, expensive and error-prone. This paper presents a tool, Bayesil, that can quickly, accurately and autonomously produce a complex biofluid's (e.g., serum or CSF) metabolic profile from a 1D1H NMR spectrum. This requires first performing several spectral processing steps then matching the resulting spectrum against a reference compound library, which contains the "signatures" of each relevant metabolite. Many of these steps are novel algorithms and our matching step views spectral matching as an inference problem within a probabilistic graphical model that rapidly approximates the most probable metabolic profile. Our extensive studies on a diverse set of complex mixtures, show that Bayesil can autonomously find the concentration of all NMR-detectable metabolites accurately (~90% correct identification and ~10% quantification error), in <5minutes on a single CPU. These results demonstrate that Bayesil is the first fully-automatic publicly-accessible system that provides quantitative NMR spectral profiling effectively -- with an accuracy that meets or exceeds the performance of trained experts. We anticipate this tool will usher in high-throughput metabolomics and enable a wealth of new applications of NMR in clinical settings. Available at http://www.bayesil.ca.

cs.AI↗