A comparative analysis of methods for pruning decision trees

In this paper, we address the problem of retrospectively pruning decision trees induced from data, according to a top-down approach. This problem has received considerable attention in the areas of pattern recognition and machine learning, and many distinct methods have been proposed in literature. We make a comparative study of six well-known pruning methods with the aim of understanding their theoretical foundations, their computational complexity, and the strengths and weaknesses of their formulation. Comments on the characteristics of each method are empirically supported. In particular, a wide experimentation performed on several data sets leads us to opposite conclusions on the predictive accuracy of simplified trees from some drawn in the literature. We attribute this divergence to differences in experimental designs. Finally, we prove and make use of a property of the reduced error pruning method to obtain an objective evaluation of the tendency to overprune/underprune observed in each method     

Microbeam RBS on flat panel displays

We have demonstrated the utility of microbeam-Rutherford BackScattering (μ-RBS) in spatially resolved studies of operational plasma effects on the interior surfaces of plasma flat panel displays manufactured by Photonics Imaging. The experiments were performed at the Sandia Nuclear microprobe using a 2.8 MeV He beam with an average beam spot size of less than 8 μm. The interior surface of the top panes of the flat panels is composed of approximately 800 nm of MgO on top of a 2000 nm thick PbO layer. μ-RBS of sample panels operated under varying conditions measured changes in the surface MgO film thickness due to plasma erosion and redeposition as accurately as ± 1.5 nm. The high accuracy in the MgO thickness measurement was achieved by inferring the MgO thickness from the shift of the Pb front edge in the RBS spectrum. An estimate for the thickness accuracy as a function of the acquired statistics is presented. The surface of the flat panels' bottom panes is also comprised of MgO on top of PbO. However, troughs 100 μm wide by 10 μm deep were partially filled with phosphor and cover the entire width of the surface. This leaves only 100 μm long sections of MgO within the trough exposed. Using μ-RBS, we were able to analyze the surface composition of these regions.     

Selective employment of chemokine receptors as human immunodeficiency virus type 1 coreceptors determined by individual amino acids within the envelope V3 loop

The chemokine receptor CCR5 acts as an essential cofactor for cell entry by macrophage-tropic human immunodeficiency virus type 1 (HIV-1) strains, whereas CXCR4 acts as an essential cofactor for T-cell-line-adapted strains. We demonstrated that the specific amino acids in the V3 loop of the HIV-1 envelope protein that determine cellular tropism also regulate chemokine coreceptor preference for cell entry by the virus. Further, a strong correlation was found between HIV-1 strains classified as syncytium inducing in standard assays and those using CXCR4 as a coreceptor. These data support the hypothesis that progressive adaptation to additional coreceptors is a key molecular basis for HIV-1 phenotypic evolution in vivo.     

The held out wings (how) Drosophila gene encodes a putative RNA-binding protein involved in the control of muscular and cardiac activity

In an attempt to identify genes that are involved in Drosophila embryonic cardiac development, we have cloned and characterized a gene whose function is required late in embryogenesis to control heart rate and muscular activity. This gene has been named held out wings (how) because hypomorphic mutant alleles produce adult animals that have lost their ability to fly and that keep their wings horizontal at a 90 degree angle from the body axis. In contrast to the late phenotype observed in null mutants, the How protein is expressed early in the invaginating mesoderm and this expression is apparently under the control of twist. When the different mesodermal lineages segregate, the expression of How becomes restricted to the myogenic lineage, including the cardioblasts and probably all the myoblasts. Antibodies directed against the protein demonstrate that How is localized to the nucleus. how encodes a protein containing one KH-domain which has been implicated in binding RNA. how is highly related to the mouse quaking gene which plays a role at least in myelination and that could serve to link a signal transduction pathway to the control of mRNA metabolism. The properties of the how gene described herein suggest that this gene participates in the control of expression of as yet unidentified target mRNAs coding for proteins essential to cardiac and muscular activity.     

Primary afferent depolarizations of sensory origin within contact-sensitive mechanoreceptive afferents of a crayfish leg

Recordings from the central branches of single identified dactyl sensory afferent (DSA) neurons in a crayfish in vitro preparation were performed to study modifications of the sensory message occurring before the first central synapse. These afferents comprised hairs and force-sensitive mechanoreceptors with phasic and phasotonic response characteristics in the terminal segment (dactyl) of the crayfish leg. More than one afferent spike size was often observed in intracellular recordings from these afferents, thus indicating the presence of electrical coupling between the central processes of DSA fibers. Additionally, in identified DSA fibers with large spike sizes, primary afferent depolarizations (PADs) of up to 15 mV were observed, which sometimes triggered antidromic spikes in the afferent. Nevertheless, PADs were clearly inhibitory, because they shunted the afferent spikes. They exhibited the following properties. First, each PAD was preceded by an afferent spike from a neighboring hair, indicating that the PADs had a sensory rather than central origin. Second, PADs could follow high frequencies of afferent discharges without failure, a property suggestive of monosynaptic connections, but because PAD latencies varied by +/-0.5 ms it is more likely that they were mediated by a disynaptic pathway. Third, although PADs were evoked in an extremely reliable manner, their amplitude varied in a quantal manner. Most unitary PADs were the result of the release of < 12 quanta, the mean quantal content lying between 4 and 5; quantal size was large, approximately 1 mV. Fourth, PADs showed facilitation in some fibers, whereas in others they became much smaller when occurring at brief intervals. We suggest that PADs may be an efficient and parsimonious way to limit sensory inflow in space and time, allowing the crayfish to identify precisely both weak and strong mechanical stimuli.     

Chebyshev multilevel absorber design concept

Pyramidal- and wedge-absorber materials are used extensively in anechoic measurement chambers to attenuate stray signals. Typical absorber layouts result in large absorber walls in which the absorber tips and bases are roughly aligned in the same plane. Such a quasi-periodic configuration produces a strong coherent specular reflection which dominates the absorber scattered field. Based on the multisection impedance transformer concept, one can divide absorber elements into different levels (layers) so that this coherence can be destroyed to reduce the specular absorber scattering level. The synthesis of this desired behavior can be implemented by the Chebyshev transformer technique, which provides the largest bandwidth given a passband ripple threshold. The resulting reflected field is then the product of the original absorber response times the Chebyshev reduction factor, which is independent of polarization and absorber properties. Various measured results are used to show that more than a 10-dB improvement can be achieved at the critical low end of the frequency band using this approach. This improvement cannot be achieved using conventional design concepts unless the absorber size is doubled     

Spotlight mode SAR stereo technique for height computation

This paper examines the feasibility of extracting three-dimensional (3-D) or topographic information in spotlight mode stereo synthetic aperture radar (SAR). A display of a SAR (intensity) image has two axes: range and cross-range. Elevated scatterers appear closer in range; this phenomenon is called radar image layover. How the height of each scatterer can be computed from the difference in its layover between two images is investigated. This is analogous to computing height from disparity distance (triangulation) in optical stereo. The same procedure can be applied on pixel by pixel basis for terrain elevation mapping. A general expression is derived for the accuracy of the height estimate as a function of the range resolution and the angular difference between the image planes. Accuracy increases as the angle between the image planes increases, but the bright scatterers in one image tend to fade in the other image. This limited angular persistence of radar scatterers is also discussed. Trajectories for data collection are examined that provide near-optimal height estimates while eliminating the scatterer persistency problem.     

Fullerene Reactivity in an Oxygen Plasma

Fullerene reactivity in an oxygen plasma was measured and compared with those determined under the same conditions for a comprehensive set of 50 carbon materials. The possible reasons for the high plasma reactivity of the fullerene sample in an oxygen plasma are discussed.