Most of the neural networks-based intrusion detection systems (IDS) examine all data features to detect intrusion or misuse patterns. Some of the features may be redundant or contribute little (if anything) to the detection process. That is why the purpose of this study is to identify important KDD features which will be used to train a neural network (NN), in order to best classify and detect attacks. Four NNs were studied: Modular, Recurrent, Principal Component Analysis (PCA), and Time-Lag recurrent (TLR) NNs. We investigated the performance of combining the Fisher's filter used as a feature selection technique, with one of the previously cited NNs. Our simulations show that using Fisher's filter improves largely the performance of the four considered NNs in terms of detection rate, attack classification, and computational time.
In the following paper, the use of fuzzy models in qualitative rating systems is analyzed in detail. The author works in an Austrian finance institution. There are at the moment two rating systems in use. The main purpose of such a rating system is to analyze company ratios to calculate a rating score, which is a measure for the financial situation and rigidity of a company. The first one is a solely hard fact rating system based on the Quicktest by Kralicek. The second one uses self-organizing maps and neural networks to calculate a rating classification and also offers the possibility to dispose personal appraisal in the calculation process.
The following work examines the application spectrum of fuzzy logic and fuzzy models in soft-fact rating systems.
We show that the use of fuzzy models in rating systems enables visualization of additional knowledge and offers the possibility to enhance the influence of a company's soft fact rating to the overall rating.
Multi-agent system is a system of autonomous, intelligent but resource-bounded agents. Particular agents have to be able to make decisions on their own, based on the activities of other agents and events within the system as well as in its environment. To this end agents make use of their own internal knowledge base which serves them as a memory. In this paper we focus on the design and management of such a knowledge base. After a brief description of some classical fundamental approaches to the knowledge base management, we propose an improvement based on the application of statistical methods. We focus in particular on the optimization of the process., Multi-agent systém je systém autonomních, inteligentních, ale zdrojově omezených agentů. Konkrétní agenti musí být schopni sami rozhodovat na základě činností jiných agentů a událostí v systému i v jeho prostředí. K tomuto účelu agenti využívají vlastní interní znalostní základnu, která jim slouží jako paměť. V tomto příspěvku se zaměřujeme na návrh a správu takové znalostní báze. Po stručném popisu některých klasických základních přístupů k řízení znalostní báze navrhujeme zlepšení založené na aplikaci statistických metod. Zaměřujeme se především na optimalizaci procesu., and Michal Košinár ; Ondřej Kohut
If the Poisson integral of the unit disc is replaced by its square root, it is known that normalized Poisson integrals of $L^{p}$ and weak $L^{p}$ boundary functions converge along approach regions wider than the ordinary nontangential cones, as proved by Rönning and the author, respectively. In this paper we characterize the approach regions for boundary functions in two general classes of Orlicz spaces. The first of these classes contains spaces $L^{\Phi }$ having the property $L^{\infty }\subset L^{\Phi }\subset L^{p}$, $1\le p<\infty $. The second contains spaces $L^{\Phi }$ that resemble $L^{p}$ spaces.
This article introduces an algorithm for implicit High Dimensional Model Representation (HDMR) of the Bellman equation. This approximation technique reduces memory demands of the algorithm considerably. Moreover, we show that HDMR enables fast approximate minimization which is essential for evaluation of the Bellman function. In each time step, the problem of parametrized HDMR minimization is relaxed into trust region problems, all sharing the same matrix. Finding its eigenvalue decomposition, we effectively achieve estimates of all minima. Their full-domain representation is avoided by HDMR and then the same approach is used recursively in the next time step. An illustrative example of N-armed bandit problem is included. We assume that the newly established connection between approximate HDMR minimization and the trust region problem can be beneficial also to many other applications.
This paper considers a distribution inventory system that consists of a single warehouse and several retailers. Customer demand arrives at the retailers according to a continuous-time renewal process. Material flow between echelons is driven by reorder point/order quantity inventory control policies. Our objective in this setting is to calculate the long-run inventory, backorder and customer service levels. The challenge in this system is to characterize the demand arrival process at the warehouse. We present a Markovian methodology to elucidate and approximate this process. We illustrate the use of this methodology in the distribution inventory system under stochastic transportation times with identical and non-identical retailers.
We discuss the convergence of approximate identities in Musielak-Orlicz spaces extending the results given by Cruz-Uribe and Fiorenza (2007) and the authors F.-Y. Maeda, Y. Mizuta and T. Ohno (2010). As in these papers, we treat the case where the approximate identity is of potential type and the case where the approximate identity is defined by a function of compact support. We also give a Young type inequality for convolution with respect to the norm in Musielak-Orlicz spaces.
In this paper, we consider Lipschitz conditions for tri-quadratic functional equations. We introduce a new notion similar to that of the left invariant mean and prove that a family of functions with this property can be approximated by tri-quadratic functions via a Lipschitz norm.
In this article we propose a method of parameters estimation for the class of discrete stable laws. Discrete stable distributions form a discrete analogy to classical stable distributions and share many interesting properties with them such as heavy tails and skewness. Similarly as stable laws discrete stable distributions are defined through characteristic function and do not posses a probability mass function in closed form. This inhibits the use of classical estimation methods such as maximum likelihood and other approach has to be applied. We depart from the H-method of maximum likelihood suggested by Kagan (1976) where the likelihood function is replaced by a function called informant which is an approximation of the likelihood function in some Hilbert space. For this method only some functionals of the distribution are required, such as probability generating function or characteristic function. We adopt this method for the case of discrete stable distributions and in a simulation study show the performance of this method.