Learning with "relevance": using a third factor to stabilize Hebbian learning

It is a well-known fact that Hebbian learning is inherently unstable because of its self-amplifying terms: the more a synapse grows, the stronger the postsynaptic activity, and therefore the faster the synaptic growth. This unwanted weight growth is driven by the autocorrelation term of Hebbian learning where the same synapse drives its own growth. On the other hand, the cross-correlation term performs actual learning where different inputs are correlated with each other. Consequently, we would like to minimize the autocorrelation and maximize the cross-correlation. Here we show that we can achieve this with a third factor that switches on learning when the autocorrelation is minimal or zero and the cross-correlation is maximal. The biological counterpart of such a third factor is a neuromodulator that switches on learning at a certain moment in time. We show in a behavioral experiment that our three-factor learning clearly outperforms classical Hebbian learning.     

A flexible multi-volume shader framework for arbitrarily intersecting multi-resolution datasets

We present a powerful framework for 3D-texture-based rendering of multiple arbitrarily intersecting volumetric datasets. Each volume is represented by a multi-resolution octree-based structure and we use out-of-core techniques to support extremely large volumes. Users define a set of convex polyhedral volume lenses, which may be associated with one or more volumetric datasets. The volumes or the lenses can be interactively moved around while the region inside each lens is rendered using interactively defined multi-volume shaders. Our rendering pipeline splits each lens into multiple convex regions such that each region is homogenous and contains a fixed number of volumes. Each such region is further split by the brick boundaries of the associated octree representations. The resulting puzzle of lens fragments is sorted in front-to-back or back-to-front order using a combination of a view-dependent octree traversal and a GPU-based depth peeling technique. Our current implementation uses slice-based volume rendering and allows interactive roaming through multiple intersecting multi-gigabyte volumes.     

Topology-controlled volume rendering

Topology provides a foundation for the development of mathematically sound tools for processing and exploration of scalar fields. Existing topology-based methods can be used to identify interesting features in volumetric data sets, to find seed sets for accelerated isosurface extraction, or to treat individual connected components as distinct entities for isosurfacing or interval volume rendering. We describe a framework for direct volume rendering based on segmenting a volume into regions of equivalent contour topology and applying separate transfer functions to each region. Each region corresponds to a branch of a hierarchical contour tree decomposition, and a separate transfer function can be defined for it. The novel contributions of our work are: 1) a volume rendering framework and interface where a unique transfer function can be assigned to each subvolume corresponding to a branch of the contour tree, 2) a runtime method for adjusting data values to reflect contour tree simplifications, 3) an efficient way of mapping a spatial location into the contour tree to determine the applicable transfer function, and 4) an algorithm for hardware-accelerated direct volume rendering that visualizes the contour tree-based segmentation at interactive frame rates using graphics processing units (GPUs) that support loops and conditional branches in fragment programs     

Efficient computation of Morse-Smale complexes for three-dimensional scalar functions

The Morse-Smale complex is an efficient representation of the gradient behavior of a scalar function, and critical points paired by the complex identify topological features and their importance. We present an algorithm that constructs the Morse-Smale complex in a series of sweeps through the data, identifying various components of the complex in a consistent manner. All components of the complex, both geometric and topological, are computed, providing a complete decomposition of the domain. Efficiency is maintained by representing the geometry of the complex in terms of point sets.     

Autonomic information flow improves prognostic impact of task force HRV monitoring

Heart rate variability (HRV) represents the cardiovascular control mediated by the autonomic nervous system and other mechanisms. In the established task force HRV monitoring different cardiovascular control mechanisms can approximately be identified at typical frequencies of heart rate oscillations by power spectral analysis. HRV measures assessing complex and fractal behavior partly improved clinical risk stratification. However, their relationship to (patho-)physiology is not sufficiently explored. Objective of the present work is the introduction of complexity measures of different physiologically relevant time scales. This is achieved by a new concept of the autonomic information flow (AIF) analysis which was designed according to task force HRV. First applications show that different time scales of AIF improve the risk stratification of patients with multiple organ dysfunction syndrome and cardiac arrest patients in comparison to standard HRV. Each group's significant time scales correspond to their respective pathomechanisms.     

A learning automata-based fault-tolerant routing algorithm for mobile ad hoc networks

Reliable routing of packets in a Mobile Ad Hoc Network (MANET) has always been a major concern. The open medium and the susceptibility of the nodes of being fault-prone make the design of protocols for these networks a challenging task. The faults in these networks, which occur either due to the failure of nodes or due to reorganization, can eventuate to packet loss. Such losses degrade the performance of the routing protocols running on them. In this paper, we propose a routing algorithm, named as learning automata based fault-tolerant routing algorithm (LAFTRA), which is capable of routing in the presence of faulty nodes in MANETs using multipath routing. We have used the theory of Learning Automata (LA) for optimizing the selection of paths, reducing the overhead in the network, and for learning about the faulty nodes present in the network. The proposed algorithm can be juxtaposed to any existing routing protocol in a MANET. The results of simulation of our protocol using network simulator 2 (ns-2) shows the increase in packet delivery ratio and decrease in overhead compared to the existing protocols. The proposed protocol gains an edge over FTAR, E2FT by nearly 2% and by more than 10% when compared with AODV in terms of packet delivery ratio with nearly 30% faulty nodes in the network. The overhead generated by our protocol is lesser by 1% as compared to FTAR and by nearly 17% as compared to E2FT when there are nearly 30% faulty nodes.     

Pulsed laser deposition of La0.67Ca0.33MnO3 thin films on LiNbO3 and surface acoustic wave studies

Surface Acoustic Waves on piezoelectric substrates can be used to investigate the dynamic conductivity of thin films in a non-contact and very sensitive way, especially at low conductivities. Here, we report on such surface acoustic wave studies to characterize thin manganite film like La_0.67Ca_0.33MnO₃, exhibiting a Jan Teller effect with a strong electron phonon interaction and a metal insulator transition at high temperatures.

We report on the deposition of La_0.67Ca_0.33MnO₃ on piezoelectric substrates (LiNbO₃ in different crystal cuts, employing a pulsed laser deposition technique. The structural quality of the thin films are examined by X-Ray Diffraction, Scanning Electron Microscope and Energy Dispersive X-ray spectroscopy. For the electrical characterization, we employ the surface acoustic wave technique, accompanied by conventional direct current resistance measurements for comparison.     

Methods of Image Compression by PHL Transform

In this paper an image data compression scheme based on Periodic Haar Piecewise-Linear (PHL) transform and quantization tables is proposed. Effectiveness of the compression for different classes of images is evaluated. The comparison of the compression quality using PHL and DCT transforms is given.     

FSAP Protects against Histone-Mediated Increase in Endothelial Permeability In Vitro

Factor-VII-activating protease (FSAP) is involved in the regulation of hemostasis and inflammation. Extracellular histones play a role in inflammation and the conversion of latent pro-FSAP into active FSAP. FSAP has been shown to regulate endothelial permeability, but the mechanisms are not clear. Here, we have investigated the effects of FSAP on endothelial permeability in vitro. A mixture of histones from calf thymus stimulated permeability, and the wild-type (WT) serine protease domain (SPD) of FSAP blocked this effect. WT–SPD–FSAP did not influence permeability on its own, nor that stimulated by thrombin or vascular endothelial growth factor (VEGF)-A₁₆₅. Histones induced a large-scale rearrangement of the junction proteins VE-cadherin and zona occludens-1 from a clear junctional distribution to a diffuse pattern. The presence of WT–SPD–FSAP inhibited these changes. Permeability changes by histones were blocked by both TLR-2 and TLR4 blocking antibodies. Histones upregulated the expression of TLR-2, but not TLR-4, in HUVEC cells, and WT–SPD–FSAP abolished the upregulation of TLR-2 expression. An inactive variant, Marburg I (MI)–SPD–FSAP, did not have any of these effects. The inhibition of histone-mediated permeability may be an important function of FSAP with relevance to sepsis, trauma, and stroke and the need to be investigated further in in vivo experiments.