Generation of optimal artificial neural networks using a pattern search algorithm: application to approximation of chemical systems

A pattern search optimization method is applied to the generation of optimal artificial neural networks (ANNs). Optimization is performed using a mixed variable extension to the generalized pattern search method. This method offers the advantage that categorical variables, such as neural transfer functions and nodal connectivities, can be used as parameters in optimization. When used together with a surrogate, the resulting algorithm is highly efficient for expensive objective functions. Results demonstrate the effectiveness of this method in optimizing an ANN for the number of neurons, the type of transfer function, and the connectivity among neurons. The optimization method is applied to a chemistry approximation of practical relevance. In this application, temperature and a chemical source term are approximated as functions of two independent parameters using optimal ANNs. Comparison of the performance of optimal ANNs with conventional tabulation methods demonstrates equivalent accuracy by considerable savings in memory storage. The architecture of the optimal ANN for the approximation of the chemical source term consists of a fully connected feedforward network having four nonlinear hidden layers and 117 synaptic weights. An equivalent representation of the chemical source term using tabulation techniques would require a 500 x 500 grid point discretization of the parameter space.     

SEMPA – Ein Ansatz des Semantischen Prozessmanagements zur Planung von Prozessmodellen

Currently process modeling is mostly done manually. Therefore, the initial design of process models as well as changes to process models which are frequently necessary to react to new market developments or new regulations are time-consuming tasks. In this paper we introduce SEMPA, an approach for the partly automatic planning of process models. Using ontologies to semantically describe actions – as envisioned in Semantic Business Process Management –, a process model for a specified problem setting can be created automatically. In comparison to existing planning algorithms our approach creates process models including control structures and is able to cope with complex and numerical input and output parameters of actions. The prototypical implementation as well as an example taken from the financial services domain illustrate the practical benefit of our approach.     

Bringing law and order to IEEE 802.11 networks—A case for DiscoSec

To improve the tarnished reputation of WLAN security, the new IEEE 802.11i standard provides means for mutual user authentication and assures confidentiality of user data. However, the IEEE 802.11 link-layer is still highly vulnerable to a plethora of simple, yet effective attacks which further jeopardize the already fragile security of wireless communications.Some of these vulnerabilities are related to limited hardware capabilities of access points and their abuse may result in serious degradation of control over the wireless connection, which, especially in the case of broadcast communication, allows for client hijacking attacks. Although these issues are known and their impact is expected to be less prevalent on modern equipment, this work demonstrates the opposite. In our experimental analysis, we tested frequently used access points, and by forcing them to operate on their performance limits, we identified significant operational anomalies and demonstrated their impact on security by implementing a novel version of the Man-In-The-Middle attack, to which we refer as the Muzzle attack.Secondly, this work describes DiscoSec, a solution for “patching” WLANs against a variety of such link-layer attacks. DiscoSec provides DoS-resilient key exchange, an efficient frame authentication, and a performance-oriented implementation. By means of extensive real-world measurements DiscoSec is evaluated, showing that even on very resource-limited devices the network throughput is decreased by only 22% compared to the throughput without any authentication, and by 6% on more performance-capable hardware. To demonstrate its effectiveness, DiscoSec is available as an open-source IEEE 802.11 device driver utilizing well-established cryptographic primitives provided by the Linux Crypto API and OpenSSL library.     

The process map as an instrument to standardize processes: design and application at a financial service provider

The standardization of processes and the identification of shared business services in a service-oriented architecture (SOA) are currently widely discussed. Above all in practice, however, there still is a lack of appropriate instruments to support these tasks. In this paper an approach for a process map is introduced which allows for a systematic presentation—as complete as possible—of the processes in an enterprise (division). After a consistent refinement of the process has taken place by means of aggregation/disaggregation respectively, generalization/specialization relations, it is possible to identify primarily functional similarities of the detailed sub-processes. The application of the process map at a financial service provider (FSP) highlights how these similarities can be taken as a basis to standardize processes and to identify shared services.     

Charting online OMICS resources: A navigational chart for clinical researchers

The life sciences have sprouted several popular and successful OMICS technologies that span all levels of biological information transfer. Ever since the start of the Human Genome Project, the then revolutionary idea to make all resulting data publicly available has been central to all of the efforts across OMICS technologies. As a result, a great variety of publicly available data repositories and resources is currently available to the research community. This widespread availability of data does come at the price of increased confusion on the part of the users, especially for those that see the OMICS technologies as tools to help unravel a larger biological or clinical question. We therefore provide a comprehensive overview of the available resources across OMICS fields, with a special emphasis on those databases that are relevant to the study of proteins. Additionally, we also describe various integrative systems that have been established, and highlight new developments in the field that can revolutionize the way in which live data integration is achieved over the internet.     

Perfect sequence excitation of the NLMS algorithm and its application to acoustic echo control

Adaptive echo cancellers are currently being studied for applications such as audio teleconference systems or hands-free telephone sets with high speech quality. The purpose of the echo control is to eliminate the acoustic feedback from the loudspeaker to the microphone. One problem of echo cancellers using e.g. the normalized least mean square algorithm(nlms) for the adaptation of the coefficients is that the convergence properties degrade with colored signal input such as speech signals [8, 9, 17]. One approach to accelerate the convergence speed is to introduce linear prediction filters in order to decorrelate the speech signal [1, 2, 12, 15, 20]. This paper presents a new approach, named the excited lms algorithm orelms algorithm, which prewhitens the input signal applying perfect sequences. Coincidently, the proposed algorithm can be interpreted as a combination of the conventionalnlms algorithm and a system identification approach using m-sequences or related sequences.     

Resonant metal-mesh bandpass filters for the far infrared

The spectral performance of freestanding resonant metal-mesh bandpass filters operating with center frequencies ranging from 585 GHz to 2.1 THz is presented. These filters are made up of a 12-μm-thick copper film with an array of cross-shaped apertures that fill a circular area with a 50-mm diameter. The filters exhibit power transmission in the range 97-100% at their respective center frequencies and stop-band rejection in excess of 18 dB. The theoretically predicted nondiffracting properties of the meshes are experimentally verified through high-resolution beam mapping. Scalability of the filter spectra with mesh dimensions is demonstrated over a wide spectral range. Several modeling methods are considered, and results from the models are shown.