7 December 2017 in Astronomy & Astrophysics
We argue that comparison with observations of theoretical models for the velocity distribution of pulsars must be done directly with the observed quantities, that is parallax and the two components of proper motion. We have developed a formalism to do so, and applied it to pulsars with accurate VLBI measurements. For computational convenience, we model the data with Maxwellians. We find that a distribution with two Maxwellians improves significantly on a single Maxwellian. The “mixed” model takes into account that pulsars move away from their place of birth, a narrow region around the Galactic plane. The best model has 42% of the pulsars in a Maxwellian with average velocity km s-1, and 58% in a Maxwellian with average velocity 540 km s-1. About 5% of the pulsars has a velocity at birth less than 60 km s-1. For the youngest pulsars (τ c < 10 Myr), these numbers are 32% with 130 km s-1, 68% with 520 km s-1, and 3%, with appreciable uncertainties. Our analysis shows that the velocity distribution is wider than can be described with a single Maxwellian; it does not prove that two Maxwellians provide a better description than other wide models.
24 June 2013 in Physical Review E
The classical, continuous-time susceptible-infected-susceptible (SIS) Markov epidemic model on an arbitrary network is extended to incorporate infection and curing or recovery times each characterized by a general distribution (rather than an exponential distribution as in Markov processes). This extension, called the generalized SIS (GSIS) model, is believed to have a much larger applicability to real-world epidemics (such as information spread in online social networks, real diseases, malware spread in computer networks, etc.) that likely do not feature exponential times. While the exact governing equations for the GSIS model are difficult to deduce due to their non-Markovian nature, accurate mean-field equations are derived that resemble our previous -intertwined mean-field approximation (NIMFA) and so allow us to transfer the whole analytic machinery of the NIMFA to the GSIS model. In particular, we establish the criterion to compute the epidemic threshold in the GSIS model. Moreover, we show that the average number of infection attempts during a recovery time is the more natural key parameter, instead of the effective infection rate in the classical, continuous-time SIS Markov model. The relative simplicity of our mean-field results enables us to treat more general types of SIS epidemics, while offering an easier key parameter to measure the average activity of those general viral agents.
22 February 2024 in Neural Networks
As neural networks become more popular, the need for accompanying uncertainty estimates increases. There are currently two main approaches to test the quality of these estimates. Most methods output a density. They can be compared by evaluating their loglikelihood on a test set. Other methods output a prediction interval directly. These methods are often tested by examining the fraction of test points that fall inside the corresponding prediction intervals. Intuitively, both approaches seem logical. However, we demonstrate through both theoretical arguments and simulations that both ways of evaluating the quality of uncertainty estimates have serious flaws. Firstly, both approaches cannot disentangle the separate components that jointly create the predictive uncertainty, making it difficult to evaluate the quality of the estimates of these components. Specifically, the quality of a confidence interval cannot reliably be tested by estimating the performance of a prediction interval. Secondly, the loglikelihood does not allow a comparison between methods that output a prediction interval directly and methods that output a density. A better loglikelihood also does not necessarily guarantee better prediction intervals, which is what the methods are often used for in practice. Moreover, the current approach to test prediction intervals directly has additional flaws. We show why testing a prediction or confidence interval on a single test set is fundamentally flawed. At best, marginal coverage is measured, implicitly averaging out overconfident and underconfident predictions. A much more desirable property is pointwise coverage, requiring the correct coverage for each prediction. We demonstrate through practical examples that these effects can result in favouring a method, based on the predictive uncertainty, that has undesirable behaviour of the confidence or prediction intervals. Finally, we propose a simulation-based testing approach that addresses these problems while still allowing easy comparison between different methods. This approach can be used for the development of new uncertainty quantification methods.
1 May 2014 in Physical Review E
By invoking the famous Fortuin, Kasteleyn, and Ginibre (FKG) inequality, we prove the conjecture that the correlation of infection at the same time between any pair of nodes in a network cannot be negative for (exact) Markovian susceptible-infected-susceptible (SIS) and susceptible-infected-removed (SIR) epidemics on networks. The truth of the conjecture establishes that the $N$-intertwined mean-field approximation (NIMFA) upper bounds the infection probability in any graph so that network design based on NIMFA always leads to safe protections against malware spread. However, when the infection or/and curing are not Poisson processes, the infection correlation between two nodes can be negative.
8 March 2022 in IEEE Transactions on Intelligent Transportation Systems
The development of assessment methods for the performance of Automated Vehicles (AVs) is essential to enable the deployment of automated driving technologies, due to the complex operational domain of AVs. One candidate is scenario-based assessment, in which test cases are derived from real-world road traffic scenarios obtained from driving data. Because of the high variety of the possible scenarios, using only observed scenarios for the assessment is not sufficient. Therefore, methods for generating additional scenarios are necessary. Our contribution is twofold. First, we propose a method to determine the parameters that describe the scenarios to a sufficient degree while relying less on strong assumptions on the parameters that characterize the scenarios. By estimating the probability density function (pdf) of these parameters, realistic parameter values can be generated. Second, we present the Scenario Representativeness (SR) metric based on the Wasserstein distance, which quantifies to what extent the scenarios with the generated parameter values are representative of real-world scenarios while covering the actual variety found in the real-world scenarios. A comparison of our proposed method with methods relying on assumptions of the scenario parameterization and pdf estimation shows that the proposed method can automatically determine the optimal scenario parameterization and pdf estimation. Furthermore, it is demonstrated that our sr metric can be used to choose the (number of) parameters that best describe a scenario. The presented method is promising, because the parameterization and pdf estimation can directly be applied to already available importance sampling strategies for accelerating the evaluation of AVs.
7 June 2024 in Neurocomputing