Growing Neural Gas approach for obtaining homogeneous maps by restricting the insertion of new nodes

The Growing Neural Gas model is used widely in artificial neural networks. However, its application is limited in some contexts by the proliferation of nodes in dense areas of the input space. In this study, we introduce some modifications to address this problem by imposing three restrictions on the insertion of new nodes. Each restriction aims to maintain the homogeneous values of selected criteria. One criterion is related to the square error of classification and an alternative approach is proposed for avoiding additional computational costs. Three parameters are added that allow the regulation of the restriction criteria. The resulting algorithm allows models to be obtained that suit specific needs by specifying meaningful parameters.

Real time motion estimation using a neural architecture implemented on GPUs

This work describes a neural network based architecture that represents and estimates object motion in videos. This architecture addresses multiple computer vision tasks such as image segmentation, object representation or characterization, motion analysis and tracking. The use of a neural network architecture allows for the simultaneous estimation of global and local motion and the representation of deformable objects. This architecture also avoids the problem of finding corresponding features while tracking moving objects. Due to the parallel nature of neural networks, the architecture has been implemented on GPUs that allows the system to meet a set of requirements such as: time constraints management, robustness, high processing speed and re-configurability. Experiments are presented that demonstrate the validity of our architecture to solve problems of mobile agents tracking and motion analysis.

3D model reconstruction using neural gas accelerated on GPU

In this work, we propose the use of the neural gas (NG), a neural network that uses an unsupervised Competitive Hebbian Learning (CHL) rule, to develop a reverse engineering process. This is a simple and accurate method to reconstruct objects from point clouds obtained from multiple overlapping views using low-cost sensors. In contrast to other methods that may need several stages that include downsampling, noise filtering and many other tasks, the NG automatically obtains the 3D model of the scanned objects. To demonstrate the validity of our proposal we tested our method with several models and performed a study of the neural network parameterization computing the quality of representation and also comparing results with other neural methods like growing neural gas and Kohonen maps or classical methods like Voxel Grid. We also reconstructed models acquired by low cost sensors that can be used in virtual and augmented reality environments for redesign or manipulation purposes. Since the NG algorithm has a strong computational cost we propose its acceleration. We have redesigned and implemented the NG learning algorithm to fit it onto Graphics Processing Units using CUDA. A speed-up of 180× faster is obtained compared to the sequential CPU version.

Chiral, Three-Dimensional Supramolecular Compounds: Homo- and Bimetallic Oxalate- and 1,2-Dithiooxalate-Bridged Networks. A Structural and Photophysical Study.

In analogy to the [M(II)(bpy)(3)](2+) cations, where M(II) is a divalent transition-metal and bpy is 2,2'-bipyridine, the tris-chelated [M(III)(bpy)(3)](3+) cations, where M(III) is Cr(III) or Co(III), induce the crystallization of chiral, anionic three-dimensional (3D) coordination polymers of oxalate-bridged (&mgr;-ox) metal complexes with stoichiometries [M(II)(2)(ox)(3)](n)()(2)(n)()(-) or [M(I)M(III)(ox)(3)](n)()(2)(n)()(-). The tripositive charge is partially compensated by inclusion of additional complex anions like ClO(4)(-), BF(4)(-), or PF(6)(-) which are encapsulated in cubic shaped cavities formed by the bipyridine ligands of the cations. Thus, an elaborate structure of cationic and anionic species within a polymeric anionic network is realized. The compounds isolated and structurally characterized include [Cr(III)(bpy)(3)][ClO(4)] [NaCr(III)(ox)(3)] (1), [Cr(III)(bpy)(3)][ClO(4)][Mn(II)(2)(ox)(3)] (2), [Cr(III)(bpy)(3)][BF(4)] [Mn(II)(2)(ox)(3)] (3), [Co(III)(bpy)(3)][PF(6)][NaCr(III)(ox)(3)] (4). Crystal data: 1, cubic, P2(1)3, a = 15.523(4) Å, Z = 4; 2, cubic, P4(1)32, a = 15.564(3) Å, Z = 4; 3, cubic, P4(1)32, a = 15.553(3) Å, Z = 4; 4, cubic, P2(1)3, a = 15.515(3) Å, Z = 4. Furthermore, it seemed likely that 1,2-dithiooxalate (dto) could act as an alternative to the oxalate bridging ligand, and as a result the compound [Ni(II)(phen)(3)][NaCo(III)(dto)(3)].C(3)H(6)O (5) has successfully been isolated and structurally characterized. Crystal data: 5, orthorhombic, P2(1)2(1)2(1), a = 16.238(4) Å, b = 16.225(4) Å, c = 18.371(5) Å, Z = 4. In addition, the photophysical properties of compound 1 have been investigated in detail. In single crystal absorption spectra of [Cr(III)(bpy)(3)][ClO(4)][NaCr(III)(ox)(3)] (1), the spin-flip transitions of both the [Cr(bpy)(3)](3+) and the [Cr(ox)(3)](3)(-) chromophores are observed and can be clearly distinguished. Irradiating into the spin-allowed (4)A(2) --> (4)T(2) absorption band of [Cr(ox)(3)](3)(-) results in intense luminescence from the (2)E state of [Cr(bpy)(3)](3+) as a result of rapid energy transfer processes.

DARPA neural network study final report /

Mode of access: Internet.

A Novel Framework for Real-time Traffic Flow Parameter Estimation from Aerial Videos

Thesis (Master's)--University of Washington, 2016-06

Cardiotrophin-like cytokine/cytokine-like factor 1 is an essential trophic factor for lumbar and facial motoneurons in vivo

The ciliary neurotrophic factor alpha-receptor(CNTFRalpha) is required for motoneuron survival during development, but the relevant ligand(s) has not been determined. One candidate is the heterodimer formed by cardiotrophin-like cytokine (CLC) and cytokine-like factor 1 (CLF). CLC/CLF binds to CNTFRalpha and enhances the survival of developing motoneurons in vitro; whether this novel trophic factor plays a role in neural development in vivo has not been tested. We examined motor and sensory neurons in embryonic chicks treated with CLC and in mice with a targeted deletion of the clf gene. Treatment with CLC increased the number of lumbar spinal cord motoneurons that survived the cell death period in chicks. However, this effect was regionally specific, because brachial and thoracic motoneurons were unaffected. Similarly, newborn clf -/- mice exhibited a significant reduction in lumbar motoneurons, with no change in the brachial or thoracic cord. Clf deletion also affected brainstem motor nuclei in a regionally specific manner; the number of motoneurons in the facial but not hypoglossal nucleus was significantly reduced. Sensory neurons of the dorsal root ganglia were not affected by either CLC treatment or clf gene deletion. Finally, mRNA for both clc and clf was found in skeletal muscle fibers of embryonic mice during the motoneuron cell death period. These findings support the view that CLC/CLF is a target-derived factor required for the survival of specific pools of motoneurons. The in vivo actions of CLC and CLF can account for many of the effects of CNTFRalpha on developing motoneurons.

Generalization by symbolic abstraction in cascaded recurrent networks

Generalization performance in recurrent neural networks is enhanced by cascading several networks. By discretizing abstractions induced in one network, other networks can operate on a coarse symbolic level with increased performance on sparse and structural prediction tasks. The level of systematicity exhibited by the cascade of recurrent networks is assessed on the basis of three language domains. (C) 2004 Elsevier B.V. All rights reserved.

Neural Stem Cells and Neurospheres - Re-evaluating the Relationship

For most of the past century, the prospect of replacing lost or damaged cells in the central nervous system (CNS) was hampered by the opinion that the adult mammalian CNS was incapable of generating new nerve cells. This belief, Like most dogmas, was essentially founded on a lack of experimental evidence to the contrary. The overturning of this 'no new neuron' hypothesis began midway through the twentieth century with a series of reports documenting neurogenesis in the postnatal and adult brain(1), continued with the isolation and in vitro culture of neurogenic cells from the adult mammalian brain(2,3), and culminated in the discovery of a population of muttipotent, selfrenewing cells in the adult CNS (that is, bona fide neural stem cells)(3-5). Although a variety of techniques were initially used, the neurosphere assay (NSA)(3,6) rapidly emerged as the assay of choice and has since become a valuable toot for isolating, and understanding the biology of, embryonic and adult CNS stem cells. Like all technologies, it is not without its limitations. In this article we will hightight several shortcomings of the assay related to its application and interpretation that we believe have led to a significant body of research whose conclusions may well be misleading.

Prediction of subcellular localization using sequence-biased recurrent networks

Motivation: Targeting peptides direct nascent proteins to their specific subcellular compartment. Knowledge of targeting signals enables informed drug design and reliable annotation of gene products. However, due to the low similarity of such sequences and the dynamical nature of the sorting process, the computational prediction of subcellular localization of proteins is challenging. Results: We contrast the use of feed forward models as employed by the popular TargetP/SignalP predictors with a sequence-biased recurrent network model. The models are evaluated in terms of performance at the residue level and at the sequence level, and demonstrate that recurrent networks improve the overall prediction performance. Compared to the original results reported for TargetP, an ensemble of the tested models increases the accuracy by 6 and 5% on non-plant and plant data, respectively.

Neural pathways mediating bilateral interactions between the upper limbs

The ease with which we perform tasks such as opening the lid of a jar, in which the two hands execute quite different actions, belies the fact that there is a strong tendency for the movements of the upper limbs to be drawn systematically towards one another. Mirror movements, involuntary contractions during intended unilateral engagement of the opposite limb, are considered pathological, as they occur in association with specific disorders of the CNS. Yet they are also observed frequently in normally developing children, and motor irradiation, an increase in the excitability of the (opposite) homologous motor pathways when unimanual movements are performed, is a robust feature of the mature motor system. The systematic nature of the interactions that occur between the upper limbs has also given rise to the expectation that functional improvements in the control of a paretic limb may occur when movements are performed in a bimanual context. In spite of the ubiquitous nature of these phenomena, there is remarkably little consensus concerning the neural basis of their mediation. In the present review, consideration is given to the putative roles of uncrossed corticofugal fibers, branched bilateral corticomotoroneuronal projections, and segmental networks. The potential for bilateral interactions to occur in various brain regions including the primary motor cortex, the supplementary motor area, non-primary motor areas, the basal ganglia, and the cerebellum is also explored. This information may provide principled bases upon which to evaluate and develop task and deficit-specific programs of movement rehabilitation and therapy. (c) 2005 Elsevier B.V. All rights reserved.