Search

Whistle classification plays an essential role in studying the habitat and social behaviours of cetaceans. We obtained six categories of sweep whistles of two Tursiops aduncus individual signals using the passive acoustic mornitoring technique over a period of eight months in the Xiamen area. First, we propose a depthwise separable convolutional neural network for whistle classification. The proposed model adopts the depthwise convolution combined with the followed point-by-point convolution instead of the conventional convolution. As a result, it brings a better classification performance in sample sets with relatively independent features between different channels. Meanwhile, it leads to less computational complexity and fewer model parameters. Second, in order to solve the problem of an imbalance in the number of samples under each whistle category, we propose a random series method with five audio augmentation algorithms. The generalization ability of the trained model was improved by using an opening probability for each algorithm and the random selection of each augmentation factor within specific ranges. Finally, we explore the effect of the proposed augmentation method on the performance of our proposed architecture and find that it enhances the accuracy up to 98.53% for the classification of Tursiops aduncus whistles.

In this paper, we construct k-valued random analytic algebroid functions for the first time. By combining the properties of random series, we study the growth and Borel points of random analytic algebroid functions in the unit disc and obtain some interesting theorems.

We prove that we can identify three types of pointwise behaviour in the regularity of the (generalized) Rosenblatt process. This extends to a non Gaussian setting previous results known for the (fractional) Brownian motion. On this purpose, fine bounds on the increments of the Rosenblatt process are needed. Our analysis is essentially based on various wavelet methods., Comment: 53 pages

Light-adapted (LA) electroretinograms (ERGs) from 90 individuals with autism spectrum disorder (ASD), mean age (13.0 ± 4.2), were compared to 87 control subjects, mean age (13.8 ± 4.8). LA-ERGs were produced by a random series of nine different Troland based, full-field flash strengths and the ISCEV standard flash at 2/s on a 30 cd m−2 white background. A random effects mixed model analysis showed the ASD group had smaller b- and a-wave amplitudes at high flash strengths (p

The contribution of the stochastic thermodiffusion to the diffusion enhancement is studied. The thermodiffusion of particles suspended in a liquid may hold place when the spontaneous endo- or exothermal nanoscale events similar to elementary acts of enzymatic reactions occur as the random series in the space and time. In these events, the energy can be emitted or absorbed at nanoscale during few to hundreds of picoseconds. It may cause local spontaneous temperature spikes spreading quickly in the space and decaying with time. The random local temperature spikes create local transient temperature gradients, where thermodiffusion of the molecules and particles holds place as well as the change in the physical properties of the suspending medium due to heating. These thermodiffusion random walks may appear as the enhanced usual Stokes-Einstein diffusion when the energy absorption/generation is high enough. The evaluated relative contribution of the mentioned effect to the molecular mobility is shown to be in agreement with experimental data on enzymatic reactions from the literature.

Multifractal analysis is the mathematical study of the irregularity of objects or irregular functions. The two main classes of processes where multifractal analysis was performed are multiplicative processes (Mandelbrot cascades and their generalizations) and additive processes (Lévy processes, random wavelet series and their generalization). An important additive process is the random sum of pulses. These series are random sums of translations and dilatations of pulses supplied by a pulse of arbitrary form. The translation, expansion and amplitude parameters can be random (or deterministically related). Special cases of this model were introduced by Lovejoy and Mandelbrot to model the precipitations at a given point and further extensions were proposed by Ciosek-Georges, Taqqu, Mandelbrot,...Finally, Y. Demichel, studied in his thesis some fractal aspects of the sample paths of such processes. Some basic properties of these processes have been studied by these authors (existence, continuity, integrability, graph dimension, global regularity of trajectories in the Besov or Sobolev sense). However, despite some existing mathematical works, many questions are still open. The purpose of the thesis will be to introduce a more general model than those that have been studied in the past, and perform their multifractal analysis (i.e. to determine the pointwise regularity of its trajectories). This can be taken in the sense of the usual Hölderian regularity when the sample paths are locally bounded, or otherwise in the sense of DOLLARpDOLLAR-exponent.The thesis concludes with an applied part: simulation of the path and the study of the multifractal formalism for the sums of random pulses. Finally, we will be particularly interested in the physiological data of marathon runners provided by V. Billat and his teams., Les deux grandes classes de processus dont l'analyse multifractale a été réalisée sont les processus multiplicatifs (issus des cascades de Mandelbrot) et les processus additifs (processus de Lévy, séries aléatoires d'ondelettes et leurs généralisations). Une classe importante de processus se rattache à cette seconde catégorie : les "sommes de pulses aléatoires". Il s'agit de séries aléatoires où l'on somme des translatées-dilatées d'un "pulse" qui peut avoir une forme arbitraire. Les paramètres de translation, dilatation et d'amplitude peuvent être aléatoires (ou certains peuvent être reliés entre eux de façondéterministe). Des cas particuliers de ce modèle ont été introduits par Lovejoy et Mandelbrot pour modéliser la pluviométrie en un point donné, puis des extensions ont été proposées par Ciosek-Georges, Taqqu, Mandelbrot, ...Enfin, Y. Demichel, dans sa thèse, a étudié des aspects fractals des trajectoires de tels processus. Certaines propriétés de base de ces processus ont été étudiées (existence, continuité, intégrabilité, dimension de graphe, régularité globale des trajectoires au sens Besov ou Sobolev); cependant, malgré quelques travaux mathématiques déjà existants, de nombreuses questions sont encore ouvertes. Le but de la thèse sera d'introduire un modèle plus général que ceux qui ont été abordés dans le passé, et d'en effectuer également l'analyse multifractale, c'est à dire déterminer la régularité ponctuelle presque sûre des trajectoires. Celle-ci pourra être prise au sens de la régularité hölderienne usuelle quand les trajectoires sont localement bornées, ou au sens de DOLLARpDOLLAR-exposant dans le cas contraire.La thèse se conclut par une partie appliquée : simulation des trajectoires et l'étude du formalisme multifractal pour les sommes de pulses aléatoires. Enfin, on s'intéressera en particulier de données physiologiques de coureurs marathoniens fournies par V. Billat et ses équipes.

For any positive decreasing to zero sequence \begin{document} $a_n$ \end{document} such that \begin{document} $\sum a_n$ \end{document} diverges we consider the related series \begin{document} $\sum k_na_n$ \end{document} and \begin{document} $\sum j_na_n.$ \end{document} Here, \begin{document} $k_n$ \end{document} and \begin{document} $j_n$ \end{document} are real sequences such that \begin{document} $k_n∈\{0,1\}$ \end{document} and \begin{document} $j_n∈\{-1,1\}.$ \end{document} We study their convergence and characterize it in terms of the density of 1's in the sequences \begin{document} $k_n$ \end{document} and \begin{document} $j_n.$ \end{document} We extend our results to series \begin{document} $\sum m_na_n,$ \end{document} with \begin{document} $m_n∈\{-1,0,1\}$ \end{document} and apply them to study some associated random series.

We consider some random series parametrised by Martin-Lof random sequences. The simplest case is that of Rademacher series, independent of a time parameter. This is then extended to the case of Fourier series on the circle with Rademacher coefficients. Finally, a specific Fourier series which has coefficients determined by a computable function is shown to converge to an algorithmically random Brownian motion.

When key-press responses to targets have to be withheld until the presentation of response prompts, target-evoked P3 amplitudes are reduced and so is the P3 difference between rare and frequent targets (the oddball effect on P3). Recently we showed that this even applied when go-signals followed targets by 100ms. Here we aimed at replicating this result with more fine-grained temporal resolution in 100ms steps from 0ms to 500ms, and dissecting the P3 complex to stimulus- and response-related portions by applying residue iteration decomposition (RIDE). Frequent and rare target stimuli (in random series) were followed by go signals (and occasional no-go signals), with block-wise fixed stimulus-onset asynchronies (SOAs) from 0ms to 500ms. Target-evoked P3 amplitudes decreased monotonically across SOAs. Part of this decrease might have been due to an overlapping Contingent Negative Variation (CNV) prior to go signals, increasing across SOAs. When CNV was subtracted out by forming rare-frequent difference waveforms, oddball-P3 was largest at SOA 0, smallest at SOA 500, and equally large at SOAs 100-400. According to RIDE, it was P3's response-related part that was increased at SOA0. These results may be interpreted in terms of the stimulus-response-link reactivation hypothesis of P3.

International audience; Pipiras introduced in the early 2000s an almost surely and uniformly convergent (on compact intervals) wavelet-type expansion of classical Rosenblatt process. Yet, the issue of estimating, almost surely, its uniform rate of convergence remained an open question. The main goal of our present article is to provide an answer to it in the more general framework of generalized Rosenblatt process, under the assumption that the underlying wavelet basis belongs to the class due to Meyer. The main ingredient of our strategy consists in expressing in a non-classical (new) way the approximation errors related with the approximation spaces of a multiresolution analysis of L 2 (R 2). Such a non-classical expression may also be of interest in its own right.

Let $(X_n)_{n\in \mathbb{N}}$ be a sequence of i.i.d. random variables with distribution $\mathbb P(X_1=1)=\mathbb P(X_1=-1)=1/2$. Let $F(\sigma)=\sum_{n=1}^\infty X_nn^{-\sigma}$. We prove that the following holds almost surely \begin{equation*} \limsup_{\sigma\to 1/2^+}\frac{F(\sigma)}{\sqrt{2\mathbb E F(\sigma)^2\log\log \mathbb E F(\sigma)^2}}=1. \end{equation*}, Comment: 19 pages, accepted version. To appear in ECP

Cycle times found in many oceanic time series have been explained with references to external mechanisms that act on the systems. Here we show that when we extract cycle times from 100 sets of paired random series, we find six distinct clusters of common cycle times ranging from about 3 years to about 32 years. Cycle times, CT, get shorter when one series in a pair is an increasingly stronger leading series, measured as LL-strength, to the other, CT ≈ - (minus) LL-strength. This may explain the frequent finding that many global warming time series, e.g., the Southern oscillation index and the Pacific decadal oscillation, show distinct cycle times (Power spectral analysis: 3-5, 7-8, 13-15, 22-24, and 29-30 years). An important implication of these findings is that processes that strengthen the impact of one ocean variable on another may cause more frequent adverse climate conditions.

There are two research traditions studying people's reactions to random binary events: one concerns serial choice reaction times, the other concerns predictions of events in a series. The present studies focused on comparing expectations between these two approaches. We formed and tested a general hypothesis that, regardless of the type of task, when an individual faces a sequence of events they initially expect trend continuation. Only when people assume that a sequence is random might they override the default and expect trend reversal instead. In a series of experiments we found that limitation of access to cognitive resources enhances expectations of trend continuation. Our interpretation of this finding is that an expectation of trend continuation is the default for the human cognitive system and that a belief in trend reversal requires access to cognitive resources to overcome the tendency to expect trend continuation.

Reading performance that can be measured by reading tests depends on whether reading material with or without contextual continuity is used.The goal of this study was to create a German version of the SKread test and to evaluate it in a clinical setting.The evaluation of the SKread test was first performed on two groups of visually healthy subjects of different ages: a junior group of 25 persons with ages between 20 and 30 years (mean = 25.84 years, SD ± 2.41 years) and a senior group of 25 persons with ages between 51 and 84 years (mean = 62.40 ± 8.46 years). The same measurements were also performed on a group of 18 patients with age-related macular degeneration (AMD) with ages between 75 and 95 years (mean = 81.89 ± 5.48 years). The reading performance was also measured using Radner charts.Using reading material without syntactic continuity considerably slowed down the reading speed and increased the error rate. Median reading rates of 11.53 characters/s (CPS) for the junior group and 8.96 CPS for the senior group were clearly lower than those for the Radner charts (22.02 CPS and 18.48 CPS, respectively). In the AMD patients, a statistical analysis of the error rates showed a highly significant difference between the Radner charts and the SKread test (p = 0.00014). Furthermore, by analyzing the errors made in the SKread test information could be obtained about the position of central scotomas. The test-retest reliability of the SKread was very good.Information about the position of a central scotoma can be acquired by using the SKread test and an analysis of reading errors, which can augment effective clinical monitoring in AMD and subsequent visual rehabilitation.