Non-redundant data clustering

Data clustering is a popular approach for automatically finding classes, concepts, or groups of patterns. In practice, this discovery process should avoid redundancies with existing knowledge about class structures or groupings, and reveal novel, previously unknown aspects of the data. In order to deal with this problem, we present an extension of the information bottleneck framework, called coordinated conditional information bottleneck, which takes negative relevance information into account by maximizing a conditional mutual information score subject to constraints. Algorithmically, one can apply an alternating optimization scheme that can be used in conjunction with different types of numeric and non-numeric attributes. We discuss extensions of the technique to the tasks of semi-supervised classification and enumeration of successive non-redundant clusterings. We present experimental results for applications in text mining and computer vision.     

Subnanometer-scale chemistry and structure of α-iron/molybdenum nitride heterophase interfaces

The local chemistry and structure of α-iron/molybdenum nitride heterophase interfaces is studied on a subnanometer scale by atom-probe field-ion microscopy (APFIM), three-dimensional atom-probe microscopy (3DAPM) and both conventional transmission electron microscopy (CTEM) and highresolution electron microscopy (HREM). Molybdenum nitride precipitates are generated by annealing Fe-2 at. pct Mo-X, where X=0.4 at. pct Sb or 0.5 at. pct Sn, at 550 °C or 600 °C, in an ammonia/hydrogen mixture. Internal nitridation at 550 °C produces thin, coherent platelet-shaped molybdenum nitride precipitates. Nitridation at 600 °C generates a much coarser structure with semicoherent thick plate-shaped and spheroidal precipitates in addition to the thin-platelet structure. The APFIM and 3DAPM analyses of the heterophase interfaces show substantial segregation of the solute species Sn and Sb only at the coarse precipitates, with Gibbsian interfacial excesses of up to 7±3 nm⁻², whereas the broad faces of the thin platelets have no detectable segregation. The TEM and HREM analyses show that the coarse precipitates are semicoherent, whereas the thin platelets are either coherent or have much fewer misfit dislocations than geometrically necessary. This demonstrates that Sn and Sb segregation is related to the presence of misfit dislocations at the interfaces of the coarse precipitates. This article is based on a presentation made at the symposium entitled “The Mechanisms of the Massive Transformation,” a part of the Fall 2000 TMS Meeting held October 16–19, 2000, in St. Louis, Missouri, under the auspices of the ASM Phase Transformations Committee.     

Dicyanate semi IPNs—A new class of high performance,high temperature plastics

Dicyanate Semi-Interpenetrating Polymer Networks (semi IPNs) are made by dissolving a thermoplastic in crosslinking dicyanates and then curing the resulting mixture. The semi IPNs produced are strong, with tensile strengths of 10,000 to 12,000 psi, and flexible, with elongations to break of 10 to 17 percent. Dicyanate semi IPNs also have good thermal stability and the softening temperatures of the IPNs are significantly, higher than those of the corresponding thermoplastics.     

The attrition dilemma: Toward a new strategy for psychotherapy research.

Presents an analysis of the attrition dilemma in psychotherapy research and reviews the methods used to compensate for data loss. It is argued that attrition is not ultimately a problem of bias but a problem of lack of information. A reformulation is offered that integrates single-case studies with traditional group comparison methodology in an attempt to find optimal causal relations of treatments to outcomes. (10 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Giga Security

The flow of information, within organizations, between networks, and from single users to other individuals and networks, is commonly at rates that only a few years ago were dreams. Earlier methods for the detection and prevention of malicious activities are anywhere from inefficient to unworkable with transmissions at the giga speeds that are prevalent today. This writing sets forth the problems and threats associated with these new high speed transmissions, and presents methodologies and systems for treating them.     

Diel Vertical Migration of Round Goby Larvae in the Great Lakes

One hypothesis for the transcontinental and intra-Great Lakes basin transfer of round gobies (Neogobius melanostomus) has been that round gobies were pumped into the ballast water of ships. During June 2005 in Lake Erie, we obtained evidence of a vertical migration of round goby larvae, when we collected 167 round goby larvae in surface ichthyoplankton net tows at night and zero during day. These results complemented similar findings from the Muskegon River estuary of Lake Michigan during 2003 and 2004, documenting diel vertical migration for the first time in larval round gobies. We suggest vertical migration behavior may have allowed larval round gobies to be transported to and within the Great Lakes via ballast water and dispersed in the Great Lakes via advection of 6.5–8.5-mm long larvae at the surface. Based on our results, if ballast water was only taken on near the surface during daylight hours from May through September when larval round gobies were present, it would have mitigated the spread of round gobies throughout the Great Lakes.     

PROBABILISTIC MODELLING OF LIGNIN LIQUEFACTION

The results of the thermal and catalytic reactions of lignin model compounds are used in the a priori prediction of its liquefaction. Lignin was defined probabilistically as an ensemble of single ring aromatics with substituents on each of their six ring positions. Elements in a substituent vector x¹, for each ring position i, were matched into all possible combinations to describe lignin. Initial number fractions, or probabilities, for each vector element were ascertained through inspection of lignin structural data. The reaction of lignins altered the probabilities within each substituent vector, and reaction products were also calculated by matching all combinations of vector elements for each ring position. Model compound results provided the identities and rates of change of vector elements.

This simulation of lignin liquefaction predicted the yields of permanent gases, aqueous liquids, single- and two-ring phenolic products, and a carbonaceous residue. The impact of the distribution of oligomeric sizes on catalyst effectiveness was probed.     

THE LOW CYCLE IMPACT FATIGUE PROPERTIES OF PEARLITIC PLAIN CARBON STEELS

The results of Stan ton and Bairstow (1908) have been reanalyzed to yield information on the composition dependence of the low cycle impact fatigue properties of pearlitic plain carbon steels. It has been found that the energy absorbed per impact, E_i, and the number of impacts are failure, N_f, are related by the equation: In this equation q is determined by the carbon, manganese and silicon contents of the steel; E_o is the impact endurance limit; and C and D are constants. The curve represented by this equation intersects that representing the Johnson-Keller high cycle impact fatigue equation at a value of log_eN_f equal to 1/C. Thus, the transition from low to high cycle impact fatigue occurs at this value of log_eN_f.