Generating feature spaces for linear algorithms with regularized sparse kernel slow feature analysis

Without non-linear basis functions many problems can not be solved by linear algorithms. This article proposes a method to automatically construct such basis functions with slow feature analysis (SFA). Non-linear optimization of this unsupervised learning method generates an orthogonal basis on the unknown latent space for a given time series. In contrast to methods like PCA, SFA is thus well suited for techniques that make direct use of the latent space. Real-world time series can be complex, and current SFA algorithms are either not powerful enough or tend to over-fit. We make use of the kernel trick in combination with sparsification to develop a kernelized SFA algorithm which provides a powerful function class for large data sets. Sparsity is achieved by a novel matching pursuit approach that can be applied to other tasks as well. For small data sets, however, the kernel SFA approach leads to over-fitting and numerical instabilities. To enforce a stable solution, we introduce regularization to the SFA objective. We hypothesize that our algorithm generates a feature space that resembles a Fourier basis in the unknown space of latent variables underlying a given real-world time series. We evaluate this hypothesis at the example of a vowel classification task in comparison to sparse kernel PCA. Our results show excellent classification accuracy and demonstrate the superiority of kernel SFA over kernel PCA in encoding latent variables.     

Path Planning for UAVs Under Communication Constraints Using SPLAT! and MILP

We will in this paper address the problem of offline path planning for Unmanned Aerial Vehicles (UAVs). Our goal is to find paths that meet mission objectives, are safe with respect to collision and grounding, fuel efficient and satisfy criteria for communication. Due to the many nonconvex constraints of the problem, Mixed Integer Linear Programming (MILP) will be used in finding the path. Approximate communication constraints and terrain avoidance constraints are used in the MILP formulation. To achieve more accurate prediction of the ability to communicate, the path is then analyzed in the radio propagation toolbox SPLAT!, and if the UAVs are not able to communicate according to design criteria for bandwidth, constraints are modified in the optimization problem in an iterative manner. The approach is exemplified with the following setup: The path of two UAVs are planned so they can serve as relay nodes between a target without line of sight to the base station.     

MICAST — The effect of magnetically controlled fluid flow on microstructure evolution in cast technical Al-alloys

The MICAST research program focuses on a systematic analysis of the effect of convection on the microstructure evolution in cast Al-alloys. The experiments within MICAST are carried out under well defined thermally and magnetically controlled, convective boundary conditions and analyzed using advanced diagnostics and theoretical modeling, involving phase field simulation, micro-modeling and global simulation of heat and mass transport. The MICAST team uses as a model material the Al-Si base alloys. This paper gives a brief overview on recent experimental results of the MICAST team on the effect of rotating magnetic fields on microstructure in AlSi alloys.     

A 2 MV Van de Graaff accelerator as a tool for planetary and impact physics research

Investigating the dynamical and physical properties of cosmic dust can reveal a great deal of information about both the dust and its many sources. Over recent years, several spacecraft (e.g., Cassini, Stardust, Galileo, and Ulysses) have successfully characterised interstellar, interplanetary, and circumplanetary dust using a variety of techniques, including in situ analyses and sample return. Charge, mass, and velocity measurements of the dust are performed either directly (induced charge signals) or indirectly (mass and velocity from impact ionisation signals or crater morphology) and constrain the dynamical parameters of the dust grains. Dust compositional information may be obtained via either time-of-flight mass spectrometry of the impact plasma or direct sample return. The accurate and reliable interpretation of collected spacecraft data requires a comprehensive programme of terrestrial instrument calibration. This process involves accelerating suitable solar system analogue dust particles to hypervelocity speeds in the laboratory, an activity performed at the Max Planck Institut fu?r Kernphysik in Heidelberg, Germany. Here, a 2 MV Van de Graaff accelerator electrostatically accelerates charged micron and submicron-sized dust particles to speeds up to 80 km s(-1). Recent advances in dust production and processing have allowed solar system analogue dust particles (silicates and other minerals) to be coated with a thin conductive shell, enabling them to be charged and accelerated. Refinements and upgrades to the beam line instrumentation and electronics now allow for the reliable selection of particles at velocities of 1-80 km s(-1) and with diameters of between 0.05 μm and 5 μm. This ability to select particles for subsequent impact studies based on their charges, masses, or velocities is provided by a particle selection unit (PSU). The PSU contains a field programmable gate array, capable of monitoring in real time the particles' speeds and charges, and is controlled remotely by a custom, platform independent, software package. The new control instrumentation and electronics, together with the wide range of accelerable particle types, allow the controlled investigation of hypervelocity impact phenomena across a hitherto unobtainable range of impact parameters.     

A “four-ferrocene” modified stem-loop structure as a probe for sensitive detection and single-base mismatch discrimination of DNA

We report the use of a four-ferrocene modified oligonucleotide as a probe for DNA detection with a gold electrode microsystem. This oligonucleotide is synthesized by automated solid-phase synthesis with four successive ferrocene moieties at the 5′-end and a C6-thiol modifier group at the 3′-end. The grafting of this 4Fc-DNA probe on a gold electrode microsystem results in the appearance of the ferrocene redox couple in cyclic voltammetry. The probe sequence is a stem-loop structure that folds efficiently on the electrode, thus optimizing electron transfer. Such architecture serves as sensor for DNA detection which is based on hybridization. The resulting disposable voltammetric sensor allowed direct, reagentless DNA detection in a small volume (20 μL). Electrochemical response upon hybridization with complementary short sequence (30-base length) and long sequence (50-base length) strands was observed by differential pulse voltammetry. Current variations were compared. The longer the sequence, the greater the decrease in current. The system's detection limit was estimated at 3.5 pM (0.07 fmol in 20 μL) with the 50-base length target and provided a dynamic detection range between 3.5 pM and 5 nM. Single mismatch detection showed a good level of sensitivity. The system was regenerated twice with no significant loss of Fc signal. Finally, 1 pM sensitivity was reached with a long chain analog of DNA PCR products of Influenza virus.     

Compressive and Tensile Behaviours of PLLA Matrix Composites Reinforced with Randomly Dispersed Flax Fibres

Nowadays, the ecological footprint of a material is becoming tremendously important. The Poly l-Lactide Acid (PLLA) matrix composites reinforced by randomly scattered flax fibres have mechanical properties similar to polyester/glass composites [1], lower environmental impacts and can be compost at the end of their lives. In this study, the mechanical characterization of biocomposites has been pushed further with the determination of the compressive and tensile properties. Furthermore, the mechanical properties of single flax fibres have been measured and implemented in a micro-mechanical estimation of the composite elastic modulus. Tensile and compressive stiffness determined by the mechanical analyses show very good correlations with the mathematical estimation.     

Comparison of Cell Sizes for Cost Efficient Deployment of a Sensor Network Aided Cognitive Radio System

To exploit spectrum resources on a secondary basis, a Sensor Network Aided Cognitive Radio Network uses a wireless sensor network that assists a secondary cognitive radio network by providing information about the current primary spectrum occupancy. In this paper we determine the performance for different cell sizes when considering the costs for deployment of base stations in a secondary network that exploits spectrum holes identified by the wireless sensor network. The secondary base station is deployed co-located with a mobile primary network that uses a cellular reuse pattern with seven frequencies. Performance of the secondary system and impact on the primary system are mainly studied in terms of throughput, packet loss and coverage when using spectrum holes in the space, time and frequency domains. Especially, we find that the cell size and configured transmit powers for the secondary system are important for optimal system performance, and that smaller cell sizes and less expensive base stations for the secondary system are beneficial. The impact on primary system performance was found to be low, but that optimal tuning of the sensor network is important.     

Auslegung von Blasensäulen mithilfe von Compartmentmodellen

In bubble columns, the phenomena of mass and heat transfer as well as the reaction are closely linked to the complex fluid dynamics. Compartment modeling offers the opportunity to integrate these phenomena while enabling an axial and radial distribution with acceptable computing effort. This article includes methods for generating the compartment geometry and fluid dynamic parameters of this modeling approach, facilitating the opportunity to optimize an industrial bubble column.     

Angiogenic potential of boron-containing bioactive glasses: in vitro study

Boron-containing bioactive glasses (BGs) are being extensively researched for the treatment and regeneration of bone defects because of their osteostimulatory and neovascularization potential. In this study, we report the effects of the ionic dissolution products (IDPs) of different boron-doped, borosilicate, and borate BG scaffolds on mouse bone marrow stromal cells in vitro, using an angiogenesis assay. Five different BG scaffolds of the system SiO₂–Na₂O–K₂O–MgO–CaO–P₂O₅–B₂O₃ (with varying amounts of SiO₂ and B₂O₃) were fabricated by the foam replication technique. Bone marrow stromal cells were cultivated in contact with the IDPs of the boron-containing BG scaffolds at different concentrations for 48 h. The expression and secretion of vascular endothelial growth factor (VEGF) from the cultured cells was measured quantitatively using the VEGF ELISA Kit. Cell viability and cell morphology were determined using WST-8 assay and H&E staining, respectively. The cellular response was found to be dependent on boron content and the B release profile from the glasses corresponded to the positive or negative biological activity of the BGs.