Multilevel Training of Binary Morphological Operators

The design of binary morphological operators that are translation-invariant and locally defined by a finite neighborhood window corresponds to the problem of designing Boolean functions. As in any supervised classification problem, morphological operators designed from a training sample also suffer from overfitting. Large neighborhood tends to lead to performance degradation of the designed operator. This work proposes a multilevel design approach to deal with the issue of designing large neighborhood-based operators. The main idea is inspired by stacked generalization (a multilevel classifier design approach) and consists of, at each training level, combining the outcomes of the previous level operators. The final operator is a multilevel operator that ultimately depends on a larger neighborhood than of the individual operators that have been combined. Experimental results show that two-level operators obtained by combining operators designed on subwindows of a large window consistently outperform the single-level operators designed on the full window. They also show that iterating two-level operators is an effective multilevel approach to obtain better results.

Local star formation triggered by supernova shocks in magnetized diffuse neutral clouds

In this work, considering the impact of a supernova remnant (SNR) with a neutral magnetized cloud we derived analytically a set of conditions that are favourable for driving gravitational instability in the cloud and thus star formation. Using these conditions, we have built diagrams of the SNR radius, R(SNR), versus the initial cloud density, n(c), that constrain a domain in the parameter space where star formation is allowed. This work is an extension to previous study performed without considering magnetic fields (Melioli et al. 2006, hereafter Paper I). The diagrams are also tested with fully three-dimensional MHD radiative cooling simulations involving a SNR and a self-gravitating cloud and we find that the numerical analysis is consistent with the results predicted by the diagrams. While the inclusion of a homogeneous magnetic field approximately perpendicular to the impact velocity of the SNR with an intensity similar to 1 mu G within the cloud results only a small shrinking of the star formation zone in the diagram relative to that without magnetic field, a larger magnetic field (similar to 10 mu G) causes a significant shrinking, as expected. Though derived from simple analytical considerations these diagrams provide a useful tool for identifying sites where star formation could be triggered by the impact of a supernova blast wave. Applications of them to a few regions of our own Galaxy (e.g. the large CO shell in the direction of Cassiopeia, and the Edge Cloud 2 in the direction of the Scorpious constellation) have revealed that star formation in those sites could have been triggered by shock waves from SNRs for specific values of the initial neutral cloud density and the SNR radius. Finally, we have evaluated the effective star formation efficiency for this sort of interaction and found that it is generally smaller than the observed values in our own Galaxy (SFE similar to 0.01-0.3). This result is consistent with previous work in the literature and also suggests that the mechanism presently investigated, though very powerful to drive structure formation, supersonic turbulence and eventually, local star formation, does not seem to be sufficient to drive global star formation in normal star-forming galaxies, not even when the magnetic field in the neutral clouds is neglected.

Delayed local inflammatory response induced by Thalassophryne nattereri venom is related to extracellular matrix degradation

Symptoms evoked by Thalassophryne nattereri fish envenomation include local oedema, severe pain and intense necrosis with strikingly inefficient healing, continuing for several weeks or months. Investigations carried out in our laboratory showed that, in the venom-induced acute inflammation, thrombosis in venules and constrictions in arterioles were highly visible, in contrast to a notable lack of inflammatory cell. Nevertheless, the reason that the venom toxins favour delayed local inflammatory response is poorly defined. In this study, we analysed the movement of leucocytes after T. nattereri venom injection in the intraplantar region of Swiss mice, the production of pro-inflammatory mediators and the venom potential to elicit matrix metalloproteinase production and extracellular matrix degradation. Total absence of mononuclear and neutrophil influx was observed until 14 days, but the venom stimulates pro-inflammatory mediator secretion. Matrix metalloproteinases (MMP)-2 and MMP-9 were detected in greater quantities, accompanied by tissue degradation of collagenous fibre. An influx of mononuclear cells was noted very late and at this time the levels of IL-6, IL-1 beta and MMP-2 remained high. Additionally, the action of venom on the cytoskeletal organization was assessed in vitro. Swift F-actin disruption and subsequent loss of focal adhesion was noted. Collectively these findings show that the altered specific interaction cell-matrix during the inflammatory process creates an inadequate environment for infiltration of inflammatory cells.

Protecting clean critical points by local disorder correlations

We show that a broad class of quantum critical points can be stable against locally correlated disorder even if they are unstable against uncorrelated disorder. Although this result seemingly contradicts the Harris criterion, it follows naturally from the absence of a random-mass term in the associated order parameter field theory. We illustrate the general concept with explicit calculations for quantum spin-chain models. Instead of the infinite-randomness physics induced by uncorrelated disorder, we find that weak locally correlated disorder is irrelevant. For larger disorder, we find a line of critical points with unusual properties such as an increase of the entanglement entropy with the disorder strength. We also propose experimental realizations in the context of quantum magnetism and cold-atom physics. Copyright (C) EPLA, 2011

Inference and local influence assessment in skew-normal null intercept measurement error model

In this article, we discuss inferential aspects of the measurement error regression models with null intercepts when the unknown quantity x (latent variable) follows a skew normal distribution. We examine first the maximum-likelihood approach to estimation via the EM algorithm by exploring statistical properties of the model considered. Then, the marginal likelihood, the score function and the observed information matrix of the observed quantities are presented allowing direct inference implementation. In order to discuss some diagnostics techniques in this type of models, we derive the appropriate matrices to assessing the local influence on the parameter estimates under different perturbation schemes. The results and methods developed in this paper are illustrated considering part of a real data set used by Hadgu and Koch [1999, Application of generalized estimating equations to a dental randomized clinical trial. Journal of Biopharmaceutical Statistics, 9, 161-178].