Entropy in Subordination and Filtering

We propose a stochastic model of transmitting random information at random time. In this model, the signal is observed as a random sampling according to an increasing stable stochastic process. Thus we are given a subordinate stochastic process which is a typical irreversible process. As the characteristic of this phenomea we observe the loss of entropy.     

Neural network calibrated stochastic processes: forecasting financial assets

If a given dynamical process contains an inherently unpredictable component, it may be modeled as a stochastic process. Typical examples from financial markets are the dynamics of prices (e.g. prices of stocks or commodities) or fundamental rates (exchange rates etc.). The unknown future value of the corresponding stochastic process is usually estimated as the expected value under a suitable measure, which may be determined from distribution of past (historical) values. The predictive power of this estimation is limited by the simplifying assumptions of common calibration methods. Here we propose a novel method of “intelligent” calibration, using learning (2-layer) neural networks in order to dynamically adapt the parameters of a stochastic model to the most recent time series of fixed length (memory depth) to the past. The process parameters are determined by the weights of the intermediate layer of the neural network. The final layer combines these parameters in a meaningful manner yielding the forecasting value for the stochastic process. On each actual finite memory, the neural network is trained by back-propagation, obtaining a much more flexible and realistic parameter calibration than an analogous fit to an autoregressive models could do. In the context of processes related to financial assets, the final combination of the output layer relates to their market-price-of-risk. The back propagation is limited to the typical memory length of the financial market (for example 10 previous business days). We demonstrate the learning efficiency of the new algorithm by tracking the next-day forecasts with one typical examples each, for the asset classes of currencies and stocks.     

具有跳跃的随机神经网络的均方稳定性

研究了一类新的具有脉冲跳跃的Hopfield神经网络系统模型,其中脉冲时刻的跳跃是由一般的随机序列所引起,通过运用Lyapunov函数方法,获取了一些新的均方稳定性结果.由于脉冲的跳跃使得不稳定的神经网络变成稳定,因而所得的结果也可以运用到其他相关领域.     

Asymptotic Multivariate Dominance: A Financial Application

We propose a multivariate stochastic dominance relation aimed at ranking different financial markets/sectors from the point of view of a non-satiable risk averse investor. In particular, we assume that the vector of returns of a given market is in the domain of attraction of a symmetric stable Paretian law in order to take into account the asymptotic behaviour of the financial returns. We determine the stochastic dominance rule for stable symmetric distributions, where the stability parameter plays a crucial role. Consequently, the multivariate rule for ordering markets is based on a comparison between i) location parameters, ii) dispersion parameters, and iii) stability indices. Finally, we apply the method to the equity markets of the four countries with the highest gross domestic product in 2013, namely, the US, China, Japan and Germany. In this empirical comparison we examine the ex ante and ex post dominance between stock markets, either assuming that the returns are jointly (or conditionally, for a robust approach) Gaussian distributed, or in the domain of attraction of a stable sub-Gaussian law.     

Robust stability for stochastic Hopfield neural networks with time delays

In this paper, the asymptotic stability analysis problem is considered for a class of uncertain stochastic neural networks with time delays and parameter uncertainties. The delays are time-invariant, and the uncertainties are norm-bounded that enter into all the network parameters. The aim of this paper is to establish easily verifiable conditions under which the delayed neural network is robustly asymptotically stable in the mean square for all admissible parameter uncertainties. By employing a Lyapunov–Krasovskii functional and conducting the stochastic analysis, a linear matrix inequality (LMI) approach is developed to derive the stability criteria. The proposed criteria can be checked readily by using some standard numerical packages, and no tuning of parameters is required. Examples are provided to demonstrate the effectiveness and applicability of the proposed criteria.     

The split-step θ-methods for stochastic delay Hopfield neural networks

In this paper, we construct a new split-step numerical method for stochastic delay Hopfield neural networks. The main aim of this paper is to investigate the mean-square stability of this split-step θ-methods for stochastic delay Hopfield neural networks. It is proved that the split-step θ-methods are mean-square stable under suitable conditions. Numerical experiments verify the numerical stability results obtained from theory. A comparison between this work and Ronghua et al. [8] is also discussed in the example.     

Stochastic finite-time boundedness of Markovian jumping neural network with uncertain transition probabilities

The stochastic finite-time boundedness problem is considered for a class of uncertain Markovian jumping neural networks (MJNNs) that possess partially known transition jumping parameters. The transition of the jumping parameters is governed by a finite-state Markov process. By selecting the appropriate stochastic Lyapunov–Krasovskii functional, sufficient conditions of stochastic finite time boundedness of MJNNs are presented and proved. The boundedness criteria are formulated in the form of linear matrix inequalities and the designed algorithms are described as optimization ones. Simulation results illustrate the effectiveness of the developed approaches.     

Global asymptotic stability of stochastic BAM neural networks with distributed delays and reaction-diffusion terms

This paper is concerned with global asymptotic stability of a class of reaction-diffusion stochastic Bi-directional Associative Memory (BAM) neural networks with discrete and distributed delays. Based on suitable assumptions, we apply the linear matrix inequality (LMI) method to propose some new sufficient stability conditions for reaction-diffusion stochastic BAM neural networks with discrete and distributed delays. The obtained results are easy to check and improve upon the existing stability results. An example is also given to demonstrate the effectiveness of the obtained results.     

Periodic solution of a stochastic SIQR epidemic model incorporating media coverage

In this paper, we propose a stochastic SIQR epidemic model with periodic parameters and media coverage. Firstly, we study that the stochastic non-autonomous periodic system has a unique global positive solution. Secondly, by using the Khasminskii''s theory, we prove that this stochastic periodic system has a nontrivial positive periodic solution. Then, we obtain the sufficient condition for extinction of the disease. Finally, numerical simulations are employed to illustrate our theoretical analysis.     

Statistical Inference for Partially Observed Markov-Modulated Diffusion Risk Model

We propose a method for obtaining the maximum likelihood estimators of the parameters of the Markov-Modulated Diffusion Risk Model in which the inter-claim times, the claim sizes, and the volatility diffusion process are influenced by an underlying Markov jump process. We consider cases when this process has been observed in two scenarios: first, only observing the inter-claim times and the claim sizes in an interval time, and second, considering the number of claims and the underlying Markov jump process at discrete times. In both cases, the data can be viewed as incomplete observations of a model with a tractable likelihood function, so we propose to use algorithms based on stochastic Expectation-Maximization algorithms to do the statistical inference. For the second scenario, we present a simulation study to estimate the ruin probability. Moreover, we apply the Markov-Modulated Diffusion Risk Model to fit a real dataset of motor insurance.

     

The optimal harvesting problem under price uncertainty

In this paper we study the exploitation of a one species forest plantation when timber price is governed by a stochastic process. The work focuses on providing closed expressions for the optimal harvesting policy in terms of the parameters of the price process and the discount factor, with finite and infinite time horizon. We assume that harvest is restricted to mature trees older than a certain age and that growth and natural mortality after maturity are neglected. We use stochastic dynamic programming techniques to characterize the optimal policy and we model price using a geometric Brownian motion and an Ornstein–Uhlenbeck process. In the first case we completely characterize the optimal policy for all possible choices of the parameters. In the second case we provide sufficient conditions, based on explicit expressions for reservation prices, assuring that harvesting everything available is optimal. In addition, for the Ornstein–Uhlenbeck case we propose a policy based on a reservation price that performs well in numerical simulations. In both cases we solve the problem for every initial condition and the best policy is obtained endogenously, that is, without imposing any ad hoc restrictions such as maximum sustained yield or convergence to a predefined final state.     

Delay-dependent stochastic stability of delayed Hopfield neural networks with Markovian jump parameters

In this paper, the problem of stochastic stability for a class of time-delay Hopfield neural networks with Markovian jump parameters is investigated. The jumping parameters are modeled as a continuous-time, discrete-state Markov process. Without assuming the boundedness, monotonicity and differentiability of the activation functions, some results for delay-dependent stochastic stability criteria for the Markovian jumping Hopfield neural networks (MJDHNNs) with time-delay are developed. We establish that the sufficient conditions can be essentially solved in terms of linear matrix inequalities.     

Ill-posedness and multi-parameter regularization in an identification problem for jump diffusion processes

In many financial applications jump diffusion processes are used to describe price dynamics of assets. Therefore, methods to estimate the parameters of this stochastic process from some market data are required. Here we briefly discuss the unique and stable solvability of this specific inverse problem. (© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)     

A mathematical framework for stochastic climate models

There has been a recent burst of activity in the atmosphere‐ocean sciences community in utilizing stable linear Langevin stochastic models for the unresolved degrees of freedom in stochastic climate prediction. Here a systematic mathematical strategy for stochastic climate modeling is developed, and some of the new phenomena in the resulting equations for the climate variables alone are explored. The new phenomena include the emergence of both unstable linear Langevin stochastic models for the climate mean variables and the need to incorporate both suitable nonlinear effects and multiplicative noise in stochastic models under appropriate circumstances. All of these phenomena are derived from a systematic self‐consistent mathematical framework for eliminating the unresolved stochastic modes that is mathematically rigorous in a suitable asymptotic limit. The theory is illustrated for general quadratically nonlinear equations where the explicit nature of the stochastic climate modeling procedure can be elucidated. The feasibility of the approach is demonstrated for the truncated equations for barotropic flow with topography. Explicit concrete examples with the new phenomena are presented for the stochastically forced three‐mode interaction equations. The conjecture of Smith and Waleffe [Phys. Fluids 11 (1999), 1608–1622] for stochastically forced three‐wave resonant equations in a suitable regime of damping and forcing is solved as a byproduct of the approach. Examples of idealized climate models arising from the highly inhomogeneous equilibrium statistical mechanics for geophysical flows are also utilized to demonstrate self‐consistency of the mathematical approach with the predictions of equilibrium statistical mechanics. In particular, for these examples, the reduced stochastic modeling procedure for the climate variables alone is designed to reproduce both the climate mean and the energy spectrum of the climate variables. © 2001 John Wiley & Sons, Inc.     

Stochastic MOLP with Incomplete Information: An Interactive Approach with Recourse

Numerous multiobjective linear programming (MOLP) methods have been proposed in the last two decades, but almost all for contexts where the parameters of problems are deterministic. However, in many real situations, parameters of a stochastic nature arise. In this paper, we suppose that the decision-maker is confronted with a situation of partial uncertainty where he possesses incomplete information about the stochastic parameters of the problem, this information allowing him to specify only the limits of variation of these parameters and eventually their central values. For such situations, we propose a multiobjective stochastic linear programming methodology; it implies the transformation of the stochastic objective functions and constraints in order to obtain an equivalent deterministic MOLP problem and the solving of this last problem by an interactive approach derived from the STEM method. Our methodology is illustrated by a didactical example.     

Stochastic stability analysis for delayed neural networks of neutral type with Markovian jump parameters

In this paper, the problem of stochastic stability for a class of delayed neural networks of neutral type with Markovian jump parameters is investigated. The jumping parameters are modelled as a continuous-time, discrete-state Markov process. A sufficient condition guaranteeing the stochastic stability of the equilibrium point is derived for the Markovian jumping delayed neural networks (MJDNNs) with neutral type. The stability criterion not only eliminates the differences between excitatory and inhibitory effects on the neural networks, but also can be conveniently checked. The sufficient condition obtained can be essentially solved in terms of linear matrix inequality. A numerical example is given to show the effectiveness of the obtained results.     

Non-convergence of stochastic gradient descent in the training of deep neural networks

Deep neural networks have successfully been trained in various application areas with stochastic gradient descent. However, there exists no rigorous mathematical explanation why this works so well. The training of neural networks with stochastic gradient descent has four different discretization parameters: (i) the network architecture; (ii) the amount of training data; (iii) the number of gradient steps; and (iv) the number of randomly initialized gradient trajectories. While it can be shown that the approximation error converges to zero if all four parameters are sent to infinity in the right order, we demonstrate in this paper that stochastic gradient descent fails to converge for ReLU networks if their depth is much larger than their width and the number of random initializations does not increase to infinity fast enough.     

Stable limits for sums of dependent infinite variance random variables

The aim of this paper is to provide conditions which ensure that the affinely transformed partial sums of a strictly stationary process converge in distribution to an infinite variance stable distribution. Conditions for this convergence to hold are known in the literature. However, most of these results are qualitative in the sense that the parameters of the limit distribution are expressed in terms of some limiting point process. In this paper we will be able to determine the parameters of the limiting stable distribution in terms of some tail characteristics of the underlying stationary sequence. We will apply our results to some standard time series models, including the GARCH(1, 1) process and its squares, the stochastic volatility models and solutions to stochastic recurrence equations.     

On Maximal Inequalities for Stable Stochastic Integrals

Sharp maximal inequalities in large and small range are derived for stable stochastic integrals. In order to control the tail of a stable process, we introduce a truncation level in the support of its Lévy measure: we show that the contribution of the compound Poisson stochastic integral is negligible as the truncation level is large, so that the study is reduced to establish maximal inequalities for the martingale part with a suitable choice of truncation level. The main problem addressed in this paper is to give upper bounds which remain bounded as the parameter of stability of the underlying stable process goes to 2. Applications to estimates of first passage times of symmetric stable processes above positive continuous curves complete this work.