Journal of Statistical Mechanics: Theory and Experiment

We propose a simple method to extract the community structure of large networks. Our method is a heuristic method that is based on modularity optimization. It is shown to outperform all other known community detection methods in terms of computation time. Moreover, the quality of the communities detected is very good, as measured by the so-called modularity. This is shown first by identifying language communities in a Belgian mobile phone network of 2 million customers and by analysing a web graph of 118 million nodes and more than one billion links. The accuracy of our algorithm is also verified on ad hoc modular networks.

Outbreaks of repeated pandemics and heavy epidemics are daunting threats to human life. This study aims at investigating the dynamics of disease conferring temporary or waning immunity with several forced-control policies aided by vaccination game theory. Considering an infinite and well-mixed homogenous population, our proposed model further illustrates the significance of introducing two well-known forced control techniques, namely, quarantine and isolation, in order to model the dynamics of an infectious disease that spreads within a human population where pre-emptive vaccination has partially been taken before the epidemic season begins. Moreover, we carefully examine the combined effects of these two types (pre-emptive and forced) of protecting measures using the SEIR-type epidemic model. An in-depth investigation based on evolutionary game theory numerically quantifies the weighing impact of individuals’ vaccinating decisions to improve the efficacy of forced control policies leading up to the relaxation of the epidemic spreading severity. A deterministic SVEIR model, including vaccinated ( V) and exposed ( E) states, is proposed having no spatial structure while implementing these intervention techniques. This study uses a mixed control strategy relying on quarantine and isolation policies to quantify the optimum requirement of vaccines for eradicating disease prevalence completely from human societies. Furthermore, our theoretical study justifies the fact that adopting forced control policies significantly reduces the required level of vaccination to suppress emerging disease prevalence, and it also confirms that the joint policy works even better when the epidemic outbreak takes place at a higher transmission rate. Research reveals that the isolation policy is a better disease attenuation tool than the quarantine policy, especially in endemic regions where the disease progression rate is relatively higher. However, a meager progression rate gradually weakens the speed of an epidemic outbreak and, therefore, applying a moderate level of control policies is sufficient to restore the disease-free state. Essentially, positive measures (pre-emptive vaccination) regulate the position of the critical line between two phases, whereas exposed provisions (quarantine or isolation) are rather dedicated to mitigating the disease spreading in endemic regions. Thus, an optimal interplay between these two types of intervention techniques works remarkably well in attenuating the epidemic size. Despite having advanced on the development of new vaccines and control strategies to mitigate epidemics, many diseases like measles, tuberculosis, Ebola, and flu are still persistent. Here, we present a dynamic analysis of the SVEIR model using mean-field theory to develop a simple but efficient strategy for epidemic control based on the simultaneous application of the quarantine and isolation policies.

Highlights

• This model incorporates the elements of mathematical epidemiology and a vaccination game into a single framework.

• A dynamical analysis of the SEIR/V epidemic model equipped with quarantine and isolation policies is introduced.

• The proposed game-theoretic framework rigorously addresses real situations when control policies are adopted simultaneously as well as separately.

• Adopting control policies can significantly reduce the required level of vaccination to suppress disease prevalence.

• A joint policy seems impressive when an epidemic outbreak takes place at a higher transmission rate.

• For a smaller basic reproduction number, the quarantine policy outperforms the isolation policy.

We examine a class of stochastic deep learning models with a tractable method to compute information-theoretic quantities. Our contributions are three-fold: (i) we show how entropies and mutual informations can be derived from heuristic statistical physics methods, under the assumption that weight matrices are independent and orthogonally-invariant. (ii) We extend particular cases in which this result is known to be rigorously exact by providing a proof for two-layers networks with Gaussian random weights, using the recently introduced adaptive interpolation method. (iii) We propose an experiment framework with generative models of synthetic datasets, on which we train deep neural networks with a weight constraint designed so that the assumption in (i) is verified during learning. We study the behavior of entropies and mutual informations throughout learning and conclude that, in the proposed setting, the relationship between compression and generalization remains elusive.

Supervised deep learning involves the training of neural networks with a large number N of parameters. For large enough N, in the so-called over-parametrized regime, one can essentially fit the training data points. Sparsity-based arguments would suggest that the generalization error increases as N grows past a certain threshold N ^*. Instead, empirical studies have shown that in the over-parametrized regime, generalization error keeps decreasing with N. We resolve this paradox through a new framework. We rely on the so-called Neural Tangent Kernel, which connects large neural nets to kernel methods, to show that the initialization causes finite-size random fluctuations of the neural net output function f  _N around its expectation . These affect the generalization error for classification: under natural assumptions, it decays to a plateau value in a power-law fashion  ∼ N ^−1/2. This description breaks down at a so-called jamming transition N  =   N ^*. At this threshold, we argue that diverges. This result leads to a plausible explanation for the cusp in test error known to occur at N ^*. Our results are confirmed by extensive empirical observations on the MNIST and CIFAR image datasets. Our analysis finally suggests that, given a computational envelope, the smallest generalization error is obtained using several networks of intermediate sizes, just beyond N ^*, and averaging their outputs.

The practical successes of deep neural networks have not been matched by theoretical progress that satisfyingly explains their behavior. In this work, we study the information bottleneck (IB) theory of deep learning, which makes three specific claims: first, that deep networks undergo two distinct phases consisting of an initial fitting phase and a subsequent compression phase; second, that the compression phase is causally related to the excellent generalization performance of deep networks; and third, that the compression phase occurs due to the diffusion-like behavior of stochastic gradient descent. Here we show that none of these claims hold true in the general case, and instead reflect assumptions made to compute a finite mutual information metric in deterministic networks. When computed using simple binning, we demonstrate through a combination of analytical results and simulation that the information plane trajectory observed in prior work is predominantly a function of the neural nonlinearity employed: double-sided saturating nonlinearities like yield a compression phase as neural activations enter the saturation regime, but linear activation functions and single-sided saturating nonlinearities like the widely used ReLU in fact do not. Moreover, we find that there is no evident causal connection between compression and generalization: networks that do not compress are still capable of generalization, and vice versa. Next, we show that the compression phase, when it exists, does not arise from stochasticity in training by demonstrating that we can replicate the IB findings using full batch gradient descent rather than stochastic gradient descent. Finally, we show that when an input domain consists of a subset of task-relevant and task-irrelevant information, hidden representations do compress the task-irrelevant information, although the overall information about the input may monotonically increase with training time, and that this compression happens concurrently with the fitting process rather than during a subsequent compression period.

We study temporally persistent and spatially extended extreme events of temperature anomalies, i.e. heat waves and cold spells, using large deviation theory. To this end, we consider a simplified yet Earth-like general circulation model of the atmosphere and numerically estimate large deviation rate functions of near-surface temperature in the mid-latitudes. We find that, after a re-normalisation based on the integrated auto-correlation, the rate function one obtains at a given latitude by looking locally in space at long time averages agrees with what is obtained, instead, by looking locally in time at large spatial averages along the latitude. This is a result of scale symmetry in the spatio-temporal turbulence and of the fact that advection is primarily zonal. This agreement hints at the universality of large deviations of the temperature field. Furthermore, we discover that the obtained rate function is able to describe the statistics of temporal averages of spatial averages performed over large spatial scales, thus allowing one to look into spatio-temporal large deviations. Finally, we find out that, as a result of a modification in the rate function, large deviations are relatively more likely to occur when looking at spatial averages performed over intermediate scales. This is due to the existence of weather patterns associated with the low-frequency variability of the atmosphere, which are responsible for extended and temporally persistent heat waves or cold spells. Extreme value theory is used to benchmark our results.

We investigate leave-one-out cross validation (CV) as a determinator of the weight of the penalty term in the least absolute shrinkage and selection operator (LASSO). First, on the basis of the message passing algorithm and a perturbative discussion assuming that the number of observations is sufficiently large, we provide simple formulas for approximately assessing two types of CV errors, which enable us to significantly reduce the necessary cost of computation. These formulas also provide a simple connection of the CV errors to the residual sums of squares between the reconstructed and the given measurements. Second, on the basis of this finding, we analytically evaluate the CV errors when the design matrix is given as a simple random matrix in the large size limit by using the replica method. Finally, these results are compared with those of numerical simulations on finite-size systems and are confirmed to be correct. We also apply the simple formulas of the first type of CV error to an actual dataset of the supernovae.

We investigate the learning performance of the pseudolikelihood maximization method for inverse Ising problems. In the teacher–student scenario under the assumption that the teacher’s couplings are sparse and the student does not know the graphical structure, the learning curve and order parameters are assessed in the typical case using the replica and cavity methods from statistical mechanics. Our formulation is also applicable to a certain class of cost functions having locality; the standard likelihood does not belong to that class. The derived analytical formulas indicate that the perfect inference of the presence/absence of the teacher’s couplings is possible in the thermodynamic limit taking the number of spins N as infinity while keeping the dataset size M proportional to N, as long as α = M/ N > 2. Meanwhile, the formulas also show that the estimated coupling values corresponding to the truly existing ones in the teacher tend to be overestimated in the absolute value, manifesting the presence of estimation bias. These results are considered to be exact in the thermodynamic limit on locally tree-like networks, such as the regular random or Erdős–Rényi graphs. Numerical simulation results fully support the theoretical predictions. Additional biases in the estimators on loopy graphs are also discussed.

We carry out a systematic study of entanglement entropy in relativistic quantum field theory. This is defined as the von Neumann entropy S _A = −Tr ρ _Alogρ _A corresponding to the reduced density matrix ρ _A of a subsystem A. For the case of a 1+1-dimensional critical system, whose continuum limit is a conformal field theory with central charge c, we re-derive the result of Holzhey et al when A is a finite interval of length in an infinite system, and extend it to many other cases: finite systems, finite temperatures, and when A consists of an arbitrary number of disjoint intervals. For such a system away from its critical point, when the correlation length ξ is large but finite, we show that , where is the number of boundary points of A. These results are verified for a free massive field theory, which is also used to confirm a scaling ansatz for the case of finite size off-critical systems, and for integrable lattice models, such as the Ising and XXZ models, which are solvable by corner transfer matrix methods. Finally the free field results are extended to higher dimensions, and used to motivate a scaling form for the singular part of the entanglement entropy near a quantum phase transition.

We consider the use of deep learning methods for modeling complex phenomena like those occurring in natural physical processes. With the large amount of data gathered on these phenomena the data intensive paradigm could begin to challenge more traditional approaches elaborated over the years in fields like maths or physics. However, despite considerable successes in a variety of application domains, the machine learning field is not yet ready to handle the level of complexity required by such problems. Using an example application, namely sea surface temperature prediction, we show how general background knowledge gained from the physics could be used as a guideline for designing efficient deep learning models. In order to motivate the approach and to assess its generality we demonstrate a formal link between the solution of a class of differential equations underlying a large family of physical phenomena and the proposed model. Experiments and comparison with series of baselines including a state of the art numerical approach is then provided.

We study the probability distribution P( X _N   =   X, N) of the total displacement X _N of an N-step run and tumble particle on a line, in the presence of a constant nonzero drive E. While the central limit theorem predicts a standard Gaussian form for near its peak, we show that for large positive and negative X, the distribution exhibits anomalous large deviation forms. For large positive X, the associated rate function is nonanalytic at a critical value of the scaled distance from the peak where its first derivative is discontinuous. This signals a first-order dynamical phase transition from a homogeneous ‘fluid’ phase to a ‘condensed’ phase that is dominated by a single large run. A similar first-order transition occurs for negative large fluctuations as well. Numerical simulations are in excellent agreement with our analytical predictions.

We consider a single Brownian particle in one dimension in a medium at a constant temperature in the underdamped regime. We stochastically reset the position of the Brownian particle to a fixed point in the space with a constant rate r whereas its velocity evolves irrespective of the position of the particle and the stochastic resetting mechanism. The nonequilibrium steady state of the position distribution is studied for this model system. Further, we study the distribution of the position of the particle in the finite time and the approach to the nonequilibrium steady state distribution with time. Numerical simulations are done to verify the analytical results.

We uncover a duality between relaxation and first passage processes in ergodic reversible Markovian dynamics in both discrete and continuous state-space. The duality exists in the form of a spectral interlacing—the respective time scales of relaxation and first passage are shown to interlace. Our canonical theory allows for the first time to determine the full first passage time distribution analytically from the simpler relaxation eigenspectrum. The duality is derived and proven rigorously for both discrete state Markov processes in arbitrary dimension and effectively one-dimensional diffusion processes, whereas we also discuss extensions to more complex scenarios. We apply our theory to a simple discrete-state protein folding model and to the Ornstein–Uhlenbeck process, for which we obtain the exact first passage time distribution analytically in terms of a Newton series of determinants of almost triangular matrices.

Exact solutions of interacting random walk models, such as 1D lattice gases, offer precise insight into the origin of nonequilibrium phenomena. Here, we study a model of run-and-tumble particles on a ring lattice interacting via hardcore exclusion. We present the exact solution for one and two particles using a generating function technique. For two particles, the eigenvectors and eigenvalues are explicitly expressed using two parameters reminiscent of Bethe roots, whose numerical values are determined by polynomial equations which we derive. The spectrum depends in a complicated way on the ratio of direction reversal rate to lattice jump rate, . For both one and two particles, the spectrum consists of separate real bands for large , which mix and become complex-valued for small . At exceptional values of , two or more eigenvalues coalesce such that the Markov matrix is non-diagonalizable. A consequence of this intricate parameter dependence is the appearance of dynamical transitions: non-analytic minima in the longest relaxation times as functions of (for a given lattice size). Exceptional points are theoretically and experimentally relevant in, e.g. open quantum systems and multichannel scattering. We propose that the phenomenon should be a ubiquitous feature of classical nonequilibrium models as well, and of relevance to physical observables in this context.

In this article we derive a measure of quantumness in quantum multi-parameter estimation problems. We can show that the ratio between the mean Uhlmann curvature and the Fisher information provides a figure of merit which estimates the amount of incompatibility arising from the quantum nature of the underlying physical system. This ratio accounts for the discrepancy between the attainable precision in the simultaneous estimation of multiple parameters and the precision predicted by the Cramér–Rao bound. As a testbed for this concept, we consider a quantum many-body system in thermal equilibrium, and explore the quantum compatibility of the model across its phase diagram.

Random deposition with a relaxation model in (u, v) flower networks is studied. In a (2, 2) flower network, the surface width W(t, N) was found to grow as b ln t in the early period and follows a ln N in the saturated regime, where t and N are the evolution time and the number of nodes in the network, respectively. The dynamic exponent z, obtained by the relation z = a/b, was z ≈ 2.11(10), which is consistent with the random walk exponent d_w = 2 in the network. For u + v ≥ 5(u ≥ 2, v ≥ u), i.e. the (2, 3) and (3, 3) flower networks, the surface width grows following power-law behavior with some corrections, where the growth exponent β and roughness exponent α are controlled by spectral dimension d_s and fractal dimension d_f of the substrate network.

We explore a variant of the Katz–Lebowitz–Spohn (KLS) driven lattice gas in two dimensions, where the lattice is split into two regions that are coupled to heat baths with distinct temperatures. The geometry was arranged such that the temperature boundaries are oriented parallel to the external particle drive and resulting net current. We have explored the changes in the dynamical behavior that are induced by our choice of the hopping rates across the temperature boundaries. If these hopping rates at the interfaces satisfy particle-hole symmetry, the current difference across them generates a vector flow diagram akin to a vortex sheet. We have studied the finite-size scaling of the particle density fluctuations in both temperature regions, and observed that it is controlled by the respective temperature values. Specifically, if the colder subsystem is maintained at the KLS critical temperature, while the hotter subsystem's temperature is set much higher, the interface current greatly suppresses particle exchange between the two regions. As a result of the ensuing effective subsystem decoupling, strong fluctuations persist in the critical region, whence the particle density fluctuations scale with the KLS critical exponents. However, if both temperatures are set well above the critical temperature, the particle density fluctuations scale according to the totally asymmetric exclusion process. In addition, we have measured the entropy production rate in both subsystems; it displays intriguing algebraic decay in the critical region, while it saturates quickly at a small but non-zero level in the hotter region. We have also considered another possible choice of the hopping rates across the temperature interfaces that explicitly breaks particle-hole symmetry. In that case the boundary rates induce a net particle flux across the interfaces that displays power-law behavior, until ultimately the particle exclusion constraints generate a clogging transition to an inert state.

Two distinct limits for deep learning have been derived as the network width h → ∞, depending on how the weights of the last layer scale with h. In the neural tangent Kernel (NTK) limit, the dynamics becomes linear in the weights and is described by a frozen kernel Θ (the NTK). By contrast, in the mean-field limit, the dynamics can be expressed in terms of the distribution of the parameters associated with a neuron, that follows a partial differential equation. In this work we consider deep networks where the weights in the last layer scale as αh^−1/2 at initialization. By varying α and h, we probe the crossover between the two limits. We observe two the previously identified regimes of 'lazy training' and 'feature training'. In the lazy-training regime, the dynamics is almost linear and the NTK barely changes after initialization. The feature-training regime includes the mean-field formulation as a limiting case and is characterized by a kernel that evolves in time, and thus learns some features. We perform numerical experiments on MNIST, Fashion-MNIST, EMNIST and CIFAR10 and consider various architectures. We find that: (i) the two regimes are separated by an α* that scales as . (ii) Network architecture and data structure play an important role in determining which regime is better: in our tests, fully-connected networks perform generally better in the lazy-training regime, unlike convolutional networks. (iii) In both regimes, the fluctuations δF induced on the learned function by initial conditions decay as , leading to a performance that increases with h. The same improvement can also be obtained at an intermediate width by ensemble-averaging several networks that are trained independently. (iv) In the feature-training regime we identify a time scale , such that for t t₁ the dynamics is linear. At t ~ t₁, the output has grown by a magnitude and the changes of the tangent kernel | |ΔΘ| | become significant. Ultimately, it follows for ReLU and Softplus activation functions, with a < 2 and a → 2 as depth grows. We provide scaling arguments supporting these findings.

General diffusion processes involve one or more diffusing species and are usually modelled by Fick's law, which assumes infinite propagation velocity. In this article, searching for the effect of finite propagation speeds in a system with two reacting species, we investigate diffusing and reacting particles governed by a hyperbolic diffusion equation, that is, the Cattaneo equation, which describes a diffusion process with finite propagation velocity, in the presence of a constant external field and reaction terms. These reaction terms are linear and may be related to irreversible and reversible processes, including memory effects, depending on the choices of the reaction rates. We obtain exact solutions for the equilibrium concentrations and explore the rich variety of behaviours exhibited by the species involved in reaction processes. Our results may shine new light into systems with more than one kind of diffusing and reacting particles, as is the case in several industrial and biological process, when finite speeds and memory effects are involved.

The spherical p-spin is a fundamental model for glassy physics, thanks to its analytical solution achievable via the replica method. Unfortunately, the replica method has some drawbacks: it is very hard to apply to diluted models and the assumptions beyond it are not immediately clear. Both drawbacks can be overcome by the use of the cavity method; however, this needs to be applied with care to spherical models. Here, we show how to write the cavity equations for spherical p-spin models, both in the replica symmetric (RS) ansatz (corresponding to belief propagation) and in the one-step replica-symmetry-breaking (1RSB) ansatz (corresponding to survey propagation). The cavity equations can be solved by a Gaussian RS and multivariate Gaussian 1RSB ansatz for the distribution of the cavity fields. We compute the free energy in both ansatzes and check that the results are identical to the replica computation, predicting a phase transition to a 1RSB phase at low temperatures. The advantages of solving the model with the cavity method are many. The physical meaning of the ansatz for the cavity marginals is very clear. The cavity method works directly with the distribution of local quantities, which allows us to generalize the method to diluted graphs. What we are presenting here is the first step towards the solution of the diluted version of the spherical p-spin model, which is a fundamental model in the theory of random lasers and interesting per se as an easier-to-simulate version of the classical fully connected p-spin model.

The phenomenon of spontaneous synchronization, particularly within the framework of the Kuramoto model, has been a subject of intense research over the years. The model comprises oscillators with distributed natural frequencies interacting through a mean-field coupling, and serves as a paradigm to study synchronization. In this review, we put forward a general framework in which we discuss in a unified way known results with more recent developments obtained for a generalized Kuramoto model that includes inertial effects and noise. We describe the model from a different perspective, highlighting the long-range nature of the interaction between the oscillators, and emphasizing the equilibrium and out-of-equilibrium aspects of its dynamics from a statistical physics point of view. In this review, we first introduce the model and discuss both for the noiseless and noisy dynamics and for unimodal frequency distributions the synchronization transition that occurs in the stationary state. We then introduce the generalized model, and analyze its dynamics using tools from statistical mechanics. In particular, we discuss its synchronization phase diagram for unimodal frequency distributions. Next, we describe deviations from the mean-field setting of the Kuramoto model. To this end, we consider the generalized Kuramoto dynamics on a one-dimensional periodic lattice on the sites of which the oscillators reside and interact with one another with a coupling that decays as an inverse power-law of their separation along the lattice. For two specific cases, namely, in the absence of noise and inertia, and in the case when the natural frequencies are the same for all the oscillators, we discuss how the long-time transition to synchrony is governed by the dynamics of the mean-field mode (zero Fourier mode) of the spatial distribution of the oscillator phases.

A unified set of hydrodynamic equations describing condensed phases of matter with broken continuous symmetries is derived using a generalization of the statistical-mechanical approach based on the local equilibrium distribution. The dissipativeless and dissipative parts of the current densities and the entropy production are systematically deduced in this approach by expanding in powers of the gradients of the macrofields. Green–Kubo formulas are obtained for all the linear transport coefficients. The consequences of microreversibility and spatial symmetries are investigated, leading to the prediction of cross effects resulting from Onsager–Casimir reciprocal relations. Crystalline solids and liquid crystals are potential examples of application. The approach is clarifying the links between the microscopic Hamiltonian dynamics and the thermodynamic and transport properties at the macroscale.

Variants of the Voronoi construction, commonly applied to divide space, are analysed for quasi-two-dimensional hard sphere systems. Configurations are constructed from a polydisperse lognormal distribution of sphere radii, mimicking recent experimental investigations. In addition, experimental conditions are replicated where spheres lie on a surface such that their respective centres do not occupy a single plane. Significantly, we demonstrate that using an unweighted (no dependence on sphere size) two-dimensional Voronoi construction (in which the sphere centres are simply projected onto a single plane) is topologically equivalent to taking the lowest horizontal section through a three-dimensional construction in which the division of space is weighted in terms of sphere size. The problem is then generalised by considering the tessellations formed from horizontal sections through the three-dimensional construction at arbitrary cut height above the basal plane. This further suggests a definition of the commonly-applied packing fraction which avoids the counter-intuitive possibility of it becoming greater than unity. Key network and Voronoi cell properties (the fraction of six-membered rings, assortativity and cell height) and are analysed as a function of separation from the basal plane and the limits discussed. Finally, practical conclusions are drawn of direct relevance to on-going experimental investigations.

We consider the variable selection problem of generalized linear models (GLMs). Stability selection (SS) is a promising method proposed for solving this problem. Although SS provides practical variable selection criteria, it is computationally demanding because it needs to fit GLMs to many re-sampled datasets. We propose a novel approximate inference algorithm that can conduct SS without the repeated fitting. The algorithm is based on the replica method of statistical mechanics and vector approximate message passing of information theory. For datasets characterized by rotation-invariant matrix ensembles, we derive state evolution equations that macroscopically describe the dynamics of the proposed algorithm. We also show that their fixed points are consistent with the replica symmetric solution obtained by the replica method. Numerical experiments indicate that the algorithm exhibits fast convergence and high approximation accuracy for both synthetic and real-world data.

In this paper, we propose a novel probability distribution that asymptotically represents a power-law, ψ( t) ∼ t ^{− α−1}, with 0 < α < 2. The main feature of the distribution is that it has a simple expression in the Laplace transform representation, making it suitable for performing calculations in stochastic processes, particularly non-Poissonian processes.

We investigate the learning performance of the pseudolikelihood maximization method for inverse Ising problems. In the teacher–student scenario under the assumption that the teacher’s couplings are sparse and the student does not know the graphical structure, the learning curve and order parameters are assessed in the typical case using the replica and cavity methods from statistical mechanics. Our formulation is also applicable to a certain class of cost functions having locality; the standard likelihood does not belong to that class. The derived analytical formulas indicate that the perfect inference of the presence/absence of the teacher’s couplings is possible in the thermodynamic limit taking the number of spins N as infinity while keeping the dataset size M proportional to N, as long as α = M/ N > 2. Meanwhile, the formulas also show that the estimated coupling values corresponding to the truly existing ones in the teacher tend to be overestimated in the absolute value, manifesting the presence of estimation bias. These results are considered to be exact in the thermodynamic limit on locally tree-like networks, such as the regular random or Erdős–Rényi graphs. Numerical simulation results fully support the theoretical predictions. Additional biases in the estimators on loopy graphs are also discussed.

A particle entering a scattering and absorbing medium executes a random walk through a sequence of scattering events. The particle ultimately either achieves first-passage, leaving the medium, or it is absorbed. The Kubelka–Munk model describes a flux of such particles moving perpendicular to the surface of a plane-parallel medium with a scattering rate and an absorption rate. The particle path alternates between the positive direction into the medium and the negative direction back towards the surface. Backscattering events from the positive to the negative direction occur at local maxima or peaks, while backscattering from the negative to the positive direction occur at local minima or valleys. The probability of a particle avoiding absorption as it follows its path decreases exponentially with the path-length λ. The reflectance of a semi-infinite slab is therefore the Laplace transform of the distribution of path-length that ends with a first-passage out of the medium. In the case of a constant scattering rate the random walk is a Poisson process. We verify our results with two iterative calculations, one using the properties of iterated convolution with a symmetric kernel and the other via direct calculation with an exponential step-length distribution. We present a novel demonstration, based on fluctuation theory of sums of random variables, that the first-passage probability as a function of the number of peaks n in the alternating path is a step-length distribution-free combinatoric expression involving Catalan numbers. Counting paths with backscattering on the real half-line results in the same Catalan number coefficients as Dyck paths on the whole numbers. Including a separate forward-scattering Poisson process results in a combinatoric expression related to counting Motzkin paths. We therefore connect walks on the real line to discrete path combinatorics.

High future discounting rates favor inaction on present expending while lower rates advise for a more immediate political action. A possible approach to this key issue in global economy is to take historical time series for nominal interest rates and inflation, and to construct then real interest rates and finally obtaining the resulting discount rate according to a specific stochastic model. Extended periods of negative real interest rates, in which inflation dominates over nominal rates, are commonly observed, occurring in many epochs and in all countries. This feature leads us to choose a well-known model in statistical physics, the Ornstein–Uhlenbeck model, as a basic dynamical tool in which real interest rates randomly fluctuate and can become negative, even if they tend to revert to a positive mean value. By covering 14 countries over hundreds of years we suggest different scenarios and include an error analysis in order to consider the impact of statistical uncertainty in our results. We find that only 4 of the countries have positive long-run discount rates while the other ten countries have negative rates. Even if one rejects the countries where hyperinflation has occurred, our results support the need to consider low discounting rates. The results provided by these fourteen countries significantly increase the priority of confronting global actions such as climate change mitigation. We finally extend the analysis by first allowing for fluctuations of the mean level in the Ornstein–Uhlenbeck model and secondly by considering modified versions of the Feller and lognormal models. In both cases, results remain basically unchanged thus demonstrating the robustness of the results presented.