Search

Quantum adiabatic optimization has long been expected to outperform classical methods in solving NP-type problems. While this has been proven in certain experiments, its main applications still reside in academic problems where the size of the system to be solved would not represent an obstacle to any modern desktop computer. Here we develop a systematic procedure to find the global optima of the modularity function to discover community structure in complex networks solely relying on pure annealers rather than hybrid solutions. We bypass the one-hot encoding constraints by hierarchically and recursively encoding binary instances of the problem that can be solved without the need to guess the exact penalties for the Lagrange multipliers. We study the variability, and robustness of the annealing process as a function of network size, directness of connections, topology, and the resolution of the communities. We show how our approach produces meaningful and at least equally optimal solutions to state-of-the-art community detection algorithms while maintaining tractable computing times. Lastly, due to its recursive nature, the annealing process returns intermediate subdivisions thus offering interpretable rather than black-box solutions. These \textit{dendrograms} can be used to unveil normal and pathological hidden hierarchies in brain networks hence opening the door to clinical workflows. Overall, this represents a first step towards an applicable practice-oriented usage of pure quantum annealing potentially bridging two segregated communities in modern science and engineering; that of network science and quantum computing.

Noise is a fundamental problem in learning theory with huge effects in the application of Machine Learning (ML) methods, due to real world data tendency to be noisy. Additionally, introduction of malicious noise can make ML methods fail critically, as is the case with adversarial attacks. Thus, finding and developing alternatives to improve robustness to noise is a fundamental problem in ML. In this paper, we propose a method to deal with noise: mitigating its effect through the use of data abstractions. The goal is to reduce the effect of noise over the model's performance through the loss of information produced by the abstraction. However, this information loss comes with a cost: it can result in an accuracy reduction due to the missing information. First, we explored multiple methodologies to create abstractions, using the training dataset, for the specific case of numerical data and binary classification tasks. We also tested how these abstractions can affect robustness to noise with several experiments that explore the robustness of an Artificial Neural Network to noise when trained using raw data \emph{vs} when trained using abstracted data. The results clearly show that using abstractions is a viable approach for developing noise robust ML methods.

In a world where Machine Learning (ML) is increasingly deployed to support decision-making in critical domains, providing decision-makers with explainable, stable, and relevant inputs becomes fundamental. Understanding how machine learning works under missing data and how this affects feature variability is paramount. This is even more relevant as machine learning approaches focus on standardising decision-making approaches that rely on an idealised set of features. However, decision-making in human activities often relies on incomplete data, even in critical domains. This paper addresses this gap by conducting a set of experiments using traditional machine learning methods that look for optimal decisions in comparison to a recently deployed machine learning method focused on a classification that is more descriptive and mimics human decision making, allowing for the natural integration of explainability. We found that the ML descriptive approach maintains higher classification accuracy while ensuring the stability of feature selection as data incompleteness increases. This suggests that descriptive classification methods can be helpful in uncertain decision-making scenarios., Comment: 30 pages, 6 figures, supplementary materials

Knotted proteins, when forced through the pores, can get stuck if the knots in their backbone tighten under force. Alternatively, the knot can slide off the chain, making translocation possible. We construct a simple energy landscape model of this process with a time-periodic potential that mimics the action of a molecular motor. We calculate the translocation time as a function of the period of the pulling force, discuss the asymptotic limits and biological relevance of the results., Comment: 6 pages

The availability of large data sets is providing an impetus for driving current artificial intelligent developments. There are, however, challenges for developing solutions with small data sets due to practical and cost-effective deployment and the opacity of deep learning models. The Comprehensive Abstraction and Classification Tool for Uncovering Structures called CACTUS is presented for improved secure analytics by effectively employing explainable artificial intelligence. It provides additional support for categorical attributes, preserving their original meaning, optimising memory usage, and speeding up the computation through parallelisation. It shows to the user the frequency of the attributes in each class and ranks them by their discriminative power. Its performance is assessed by application to the Wisconsin diagnostic breast cancer and Thyroid0387 data sets.

Stochastic restarting is a strategy of starting anew. Incorporation of the resetting to the random walks can result in the decrease of the mean first passage time, due to the ability to limit unfavorably meandering, sub-optimal trajectories. In the following manuscript we examine how stochastic resetting influences escape dynamics from the $(-\infty,1)$ interval in the presence of the single-well power-law $|x|^\kappa$ potentials with $\kappa>0$. Examination of the mean first passage time is complemented by the analysis of the coefficient of variation, which provides a robust and reliable indicator assessing efficiency of stochastic resetting. The restrictive nature of resetting is compared to placing a reflective boundary in the system at hand. In particular, for each potential, the position of the reflecting barrier giving the same mean first passage time as the optimal resetting rate is determined. Finally, in addition to reflecting, we compare effectiveness of other resetting strategies with respect to optimization of the mean first passage time., Comment: 7 pages

Properties of the noise-driven escape kinetics are mainly determined by the stochastic component of the system dynamics. Nevertheless, the escape dynamics is also sensitive to deterministic forces. Here, we are exploring properties of the overdamped drifted escape from finite intervals under the action of symmetric $\alpha$-stable noises. We show that the properly rescaled mean first passage time follows the universal pattern as a function of the generalized P\'ecklet number, which can be used to efficiently discriminate between domains where drift or random force dominate. Stochastic driving of the $\alpha$-stable type is capable of diminishing the significance of the drift in the regime when the drift prevails., Comment: 8 pages

Stochastic resetting and noise-enhanced stability are two phenomena which can affect the lifetime and relaxation of nonequilibrium states. They can be considered as measures of controlling the efficiency of the completion process when a stochastic system has to reach a desired state. Here, we study interaction of random (Poissonian) resetting and stochastic dynamics in unstable potentials. Unlike noise-induced stability which increases the relaxation time, the stochastic resetting may eliminate winding trajectories contributing to the lifetime and accelerate the escape kinetics from unstable states. In the paper we present a framework to analyze compromises between the two contrasting phenomena in a noise-driven kinetics subject to random restarts., Comment: 7 pages

The escape of the randomly accelerated undamped particle from the finite interval under action of stochastic resetting is studied. The motion of such a particle is described by the full Langevin equation and the particle is characterized by the position and velocity. We compare three resetting protocols, which restarts velocity or position (partial resetting) and the whole motion (position and velocity -- full resetting). Using the mean first passage time we assess efficiency of restarting protocols in facilitating or suppressing the escape kinetics. There are fundamental differences between partial resetting scenarios which restart velocity or position, as in the limit of very frequent resets only the position resetting (regardless of initial velocity) can trap the particle in the finite domain of motion. The velocity resetting or the simultaneous position and velocity restarting provide a possibility of facilitating the undamped escape kinetics., Comment: 23 pages

Theory of stochastic processes provides theoretical tools which can be efficiently used to explore properties of noise induced escape kinetics. Since noise facilitated escape over the potential barrier resembles free climbing, one can use the first passage time theory in analysis of rock climbing. We perform the analysis of the mean first passage time in order to answer the question regarding the optimal, i.e., resulting in the fastest climbing, rope length. It is demonstrated that there is a discrete set of favorable rope lengths assuring shortest climbing times, as they correspond to local minima of mean first passage time. Within the set of favorable rope lengths there is the optimal rope giving rise to the shortest climbing time. In particular, more experienced climbers can decrease their climbing time by using longer ropes., Comment: 7 pages, 6 figures

The escape from a given domain is one of the fundamental problems in statistical physics and the theory of stochastic processes. Here, we explore properties of the escape of an inertial particle driven by L\'evy noise from a bounded domain, restricted by two absorbing boundaries.Presence of two absorbing boundaries assures that the escape process can be characterized by the finite mean first passage time. The detailed analysis of escape kinetics shows that properties of the mean first passage time for the integrated Ornstein--Uhlenbeck process driven by L\'evy noise are closely related to properties of the integrated L\'evy motions which, in turn, are close to properties of the integrated Wiener process. The extensive studies of the mean first passage time were complemented by examination of the escape velocity and energy along with their sensitivity to initial conditions., Comment: 10 pages

We consider properties of one-dimensional diffusive dichotomous flow and discuss effects of resonant activation in the presence of statistically independent random resetting mechanism. Resonant activation and stochastic resetting are two similar effects, as both of them can optimize the noise induced escape. Our studies show completely different origins of optimization in adapted setups. Efficiency of stochastic resetting relies on elimination of suboptimal trajectories while resonant activation is associated with matching of time scales in the dynamic environment. Consequently, both effects can be easily tracked by studying their asymptotic properties. Finally, stochastic resetting cannot be easily used to further optimization of the resonant activation in symmetric setups., Comment: 8 pages

Combined action of noise and deterministic force in dynamical systems can induce resonant effects. Here, we demonstrate a minimal, deterministic-force-free, setup allowing for occurrence of resonant, noise induced effects. We show that in the archetypal problem of escape from finite intervals driven by $\alpha$-stale noise with the periodically modulated stability index, depending on the initial direction of the modulation, resonant-activation-like or noise-enhanced-stability-like phenomena can be observed., Comment: 8 pages

The escape from a potential well is an archetypal problem in the study of stochastic dynamical systems, representing real-world situations from chemical reactions to leaving an established home range in movement ecology. Concurrently, L{\'e}vy noise is a well-established approach to model systems characterized by statistical outliers and diverging higher-order moments, ranging from gene expression control to the movement patterns of animals and humans. Here, we study the problem of L\'evy noise-driven escape from an almost rectangular, arctan potential well restricted by two absorbing boundaries. We unveil analogies of the observed transient dynamics to the general properties of stationary states of L{\'e}vy processes in single-well potentials. The first escape dynamics is shown to exhibit exponential tails. We examine the dependence of the escape on the shape parameters, steepness and height, of the arctan potential. Finally, we explore in detail the behavior of the probability densities of the first-escape time and the last-hitting point., Comment: 12 pages

The noise driven motion in a bistable potential acts as the archetypal model of various physical phenomena. Here, we contrast the overdamped dynamics with the full (underdamped) dynamics. For the overdamped particle driven by a non-equilibrium, $\alpha$-stable noise the ratio of forward and backward transition rates depends only on the width of a potential barrier separating both minima. Using analytical and numerical methods, we show that in the regime of full dynamics, contrary to the overdamped case, the ratio of transition rates depends both on widths and heights of the potential barrier. The analytical formula for the ratio of transition rates is corroborated by extensive numerical simulations., Comment: 9 pages

Stationary states of overdamped anharmonic stochastic oscillators driven by L\'evy noise are typically multimodal. The very same situation is recorded for an underdamped L\'evy noise driven motion in single-well potentials with linear friction. Within current manuscript we relax the assumption that the friction experienced by a particle is linear. Using computer simulations, we study underdamped motion in single-well potentials in the regime of nonlinear friction. We demonstrate that it is relatively easy to observe multimodality in the velocity distribution as it is determined by the friction itself and it is the same as the multimodality in the overdamped case with the analogous deterministic force. Contrary to the velocity marginal density, it is more difficult to produce multimodality in the position. Nevertheless, for fine-tuned nonlinear friction, the spatial multimodality can be recorded., Comment: 10 pages

Stochastic evolution of various dynamic systems and reaction networks is commonly described in terms of noise assisted escape of an overdamped particle from a potential well, as devised by the paradigmatic Langevin equation in which additive Gaussian stochastic force reproduces effects of thermal fluctuations from the reservoir. When implemented for systems close to equilibrium, the approach correctly explains emergence of Boltzmann distribution for the ensemble of trajectories generated by Langevin equation and relates intensity of the noise strength to the mobility. This scenario can be further generalized to include effects of non-Gaussian, burst-like forcing modeled by L\'evy noise. In this case however, the pulsatile additive noise cannot be treated as the internal (thermal), since the relation between the strength of the friction and variance of the noise is violated. Heavy tails of L\'evy noise distributions not only facilitate escape kinetics, but more importantly, change the escape protocol by altering final stationary state to a non-Boltzmann, non-equilibrium form. As a result, contrary to the kinetics induced by a Gaussian white noise, escape rates in environments with L\'evy noise are determined not by the barrier height, but instead, by the barrier width. We further discuss consequences of simultaneous action of thermal and L\'evy noises on statistics of passage times and population of reactants in double-well potentials., Comment: 12 pages

Using methods of stochastic dynamics, we have studied stationary states in the underdamped anharmonic stochastic oscillators driven by Cauchy noise. Shape of stationary states depend both on the potential type and the damping. If the damping is strong enough, for potential wells which in the overdamped regime produce multimodal stationary states, stationary states in the underdamped regime can be multimodal with the same number of modes like in the overdamped regime. For the parabolic potential, the stationary density is always unimodal and it is given by the two dimensional $\alpha$-stable density. For the mixture of quartic and parabolic single-well potentials the stationary density can be bimodal. Nevertheless, the parabolic addition, which is strong enough, can destroy bimodlity of the stationary state., Comment: 9 pages

A L\'evy noise is an efficient description of out-of-equilibrium systems. The presence of L\'evy flights results in a plenitude of noise-induced phenomena. Among others, L\'evy flights can produce stationary states with more than one modal value in single-well potentials. Here, we explore stationary states in special double-well potentials demonstrating that a sufficiently high potential barrier separating potential wells can produce bimodal stationary states in each potential well. Furthermore, we explore how the decrease in the barrier height affects the multimodality of stationary states. Finally, we explore a role of the multimodality of stationary states on the noise induced escape over the static potential barrier., Comment: 10 pages, 11 figures

Stationary states for a particle moving in a single-well, steeper than parabolic, potential driven by L\'evy noise can be bi-modal. Here, we explore in details conditions that are required in order to induce multimodal stationary states having more than two modal values. Phenomenological arguments determining necessary conditions for emergence of stationary states of higher multimodality are provided. Basing on these arguments, appropriate symmetric single-well potentials are constructed. Finally, using numerical methods it is verified that stationary states have anticipated multimodality., Comment: 15 pages 10 figures

L\'evy walks are continuous time random walks with spatio-temporal coupling of jump lengths and waiting times, often used to model superdiffusive spreading processes such as animals searching for food, tracer motion in weakly chaotic systems, or even the dynamics in quantum systems such as cold atoms. In the simplest version L\'evy walks move with a finite speed. Here, we present an extension of the L\'evy walk scenario for the case when external force fields influence the motion. The resulting motion is a combination of the response to the deterministic force acting on the particle, changing its velocity according to the principle of total energy conservation, and random velocity reversals governed by the distribution of waiting times. For the fact that the motion stays conservative, that is, on a constant energy surface, our scenario is fundamentally different from thermal motion in the same external potentials. In particular, we present results for the velocity and position distributions for single well potentials of different steepness. The observed dynamics with its continuous velocity changes enriches the theory of L\'evy walk processes and will be of use in a variety of systems, for which the particles are externally confined., Comment: 27 pages, 17 figures