Isolation kernel: the X factor in efficient and effective large scale online kernel learning

Large scale online kernel learning aims to build an efficient and scalable kernel-based predictive model incrementally from a sequence of potentially infinite data points. Current state-of-the-art large scale online kernel learning focuses on improving efficiency. Two key approaches to gain efficiency through approximation are (1) limiting the number of support vectors, and (2) using an approximate feature map. They often employ a kernel with a feature map with intractable dimensionality. While these approaches can deal with large scale datasets efficiently, this outcome is achieved by compromising predictive accuracy because of the approximation. We offer an alternative approach that puts the kernel used at the heart of the approach. It focuses on creating a sparse and finite-dimensional feature map of a kernel called Isolation Kernel. Using this new approach, to achieve the above aim of large scale online kernel learning becomes extremely simple—simply use Isolation Kernel instead of a kernel having a feature map with intractable dimensionality. We show that, using Isolation Kernel, large scale online kernel learning can be achieved efficiently without sacrificing accuracy.

     

Development of a high pressure automated lag time apparatus for experimental studies and statistical analyses of nucleation and growth of gas hydrates

Nucleation in a supercooled or a supersaturated medium is a stochastic event, and hence statistical analyses are required for the understanding and prediction of such events. The development of reliable statistical methods for quantifying nucleation probability is highly desirable for applications where control of nucleation is required. The nucleation of gas hydrates in supercooled conditions is one such application. We describe the design and development of a high pressure automated lag time apparatus (HP-ALTA) for the statistical study of gas hydrate nucleation and growth at elevated gas pressures. The apparatus allows a small volume (≈150 μl) of water to be cooled at a controlled rate in a pressurized gas atmosphere, and the temperature of gas hydrate nucleation, T(f), to be detected. The instrument then raises the sample temperature under controlled conditions to facilitate dissociation of the gas hydrate before repeating the cooling-nucleation cycle again. This process of forming and dissociating gas hydrates can be automatically repeated for a statistically significant (>100) number of nucleation events. The HP-ALTA can be operated in two modes, one for the detection of hydrate in the bulk of the sample, under a stirring action, and the other for the detection of the formation of hydrate films across the water-gas interface of a quiescent sample. The technique can be applied to the study of several parameters, such as gas pressure, cooling rate and gas composition, on the gas hydrate nucleation probability distribution for supercooled water samples.     

Stiction issues and actuation of RF LIGA-MEMS variable capacitors

High aspect ratio variable capacitors have been fabricated using deep X-ray lithography and electroplating. Stiction phenomena applicable to high aspect ratio devices are presented, including the conditions for stiction to occur and the critical dimensions of structures. Actuation tests at 3 GHz are also presented and show a maximum capacitance of 0.86 pF with no actuation voltage and a minimum capacitance of 0.70 pF with an actuation voltage of 20 V just before pull-in, which gives a tuning range of 1.23:1. Corresponding Q-factor values are 49.3 and 70.8 respectively. After pull-in, the measured capacitance is 0.61 pF, corresponding to a tuning range of 1.41:1, with a maximum Q-factor of 102.9.     

Modeling the microstructural changes during hot tandem rolling of AA5XXX aluminum alloys: Part I. Microstructural evolution

A comprehensive mathematical model of the hot tandem rolling process for aluminum alloys has been developed. Reflecting the complex thermomechanical and microstructural changes effected in the alloys during rolling, the model incorporated heat flow, plastic deformation, kinetics of static recrystallization, final recrystallized grain size, and texture evolution. The results of this microstructural engineering study, combining computer modeling, laboratory tests, and industrial measurements, are presented in three parts. In this Part I, laboratory measurements of static recrystallization kinetics and final recrystallized grain size are described for AA5182 and AA5052 aluminum alloys and expressed quantitatively by semiempirical equations. In Part II, laboratory measurements of the texture evolution during static recrystallization are described for each of the alloys and expressed mathematically using a modified form of the Avrami equation. Finally, Part III of this article describes the development of an overall mathematical model for an industrial aluminum hot tandem rolling process which incorporates the microstructure and texture equations developed and the model validation using industrial data. The laboratory measurements for the microstructural evolution were carried out using industrially rolled material and a state-of-the-art plane strain compression tester at Alcan International. Each sample was given a single deformation and heat treated in a salt bath at 400 °C for various lengths of time to effect different levels of recrystallization in the samples. The range of hot-working conditions used for the laboratory study was chosen to represent conditions typically seen in industrial aluminum hot tandem rolling processes, i.e., deformation temperatures of 350 °C to 500 °C, strain rates of 0.5 to 100 seconds and total strains of 0.5 to 2.0. The semiempirical equations developed indicated that both the recrystallization kinetics and the final recrystallized grain size were dependent on the deformation history of the material i.e., total strain and Zener-Hollomon parameter (Z), where and time at the recrystallization temperature.     

Musculoskeletal symptoms among mobile hand-held device users and their relationship to device use: A preliminary study in a Canadian university population

The study aims were, in a population of university students, staff, and faculty (n = 140), to: 1) determine the distribution of seven measures of mobile device use; 2) determine the distribution of musculoskeletal symptoms of the upper extremity, upper back and neck; and 3) assess the relationship between device use and symptoms. 137 of 140 participants (98%) reported using a mobile device. Most participants (84%) reported pain in at least one body part. Right hand pain was most common at the base of the thumb. Significant associations found included time spent internet browsing and pain in the base of the right thumb (odds ratio 2.21, 95% confidence interval 1.02–4.78), and total time spent using a mobile device and pain in the right shoulder (2.55, 1.25–5.21) and neck (2.72, 1.24–5.96). Although this research is preliminary, the observed associations, together with the rising use of these devices, raise concern for heavy users.     

Extremes in the complexity of computing metric distances between partitions

Day [3] describes an analytical model of minimum-length sequence (MLS) metrics measuring distances between partitions of a set. By selecting suitable values of model coordinates, a user may identify within the model that metric most appropriate to his classification application. Users should understand that within the model similar metrics may nevertheless exhibit extreme differences in their computational complexities. For example, the asymptotic time complexities of two MLS metrics are known to be linear in the number of objects being partitioned; yet we establish below that the computational problem for a closely related MLS metric is NP-complete.     

Cryogenic and elevated temperature hypervelocity impact facility

The hypervelocity impact facility at Space Research Institute (SRI), Auburn University has recently completed a facility upgrade that permits the impact testing of space materials within the cryogenic and elevated temperature range. Sample temperatures within the range of 40–450 K have been achieved for polymer films. These wide temperature range capabilities add to the facilities current testing experience with impact initiated plasma discharge testing for solar cell arrays. The facility utilizes a plasma drag gun to accelerate a variety simulated micrometeorite materials in the 50–150 μm range to velocities between 5 and 12 km/s. For each test 5–50 particles impact the surface of the target sample within an impact area of approximately 15 cm in diameter. The test chamber can accommodate samples up to a meter wide for ambient and heated tests, and 48 cm for cryogenic samples. The gun and test chambers are evacuated by He cryopumps and dry roughing pumps to produce a clean, oil free environment. Utilizing a streak camera and PMT detection system, the correspondence between individual particle size, speed and impact site can be determined. Standard post-analysis yields: micrographs of each impact site, dimensions of the pertinent impact characteristics, individual particle velocity and size estimates.     

Effect of subsurface thermocouple installation on the discrepancy of the measured thermal history and predicted surface heat flux during a quench operation

Water quenching plays an important role in metallurgical and materials manufacturing operations to control both the temperature of the product during processing and its final microstructure. In order to control a water-quench process, the surface heat-transfer coefficient or heat flux must be quantified accurately. A common procedure to do this is to use an inverse heat conduction (IHC) model to estimate the heat-transfer boundary condition (heat flux or heat- transfer coefficient) based on the measured thermal history during the quench operation at a known interior location in the sample. Traditionally, thermocouples (TCs) have been extensively used during quench tests to measure the sample temperature history. This article will examine the effect of the hole used to insert the thermocouple into the sample and its orientation with respect to the quenched surface, on the perturbation in the thermal field around the TC measurement point during water-quench operations characterized by boiling heat transfer. The effect of some other factors on the perturbation of the thermal field at the TC measurement point during water-quench operations such as the diameter of the thermocouple hole, thermocouple distance from the quench surface, sample thermal conductivity, and quench intensity were also investigated. A two-dimensional (2-D) axisymmetric IHC model developed at the University of British Columbia is used to estimate the error in the predicted heat fluxes based on the thermal history measured at the thermocouple measurement point. The study showed, for some quench conditions, that the thermocouple hole must be included in the IHC analysis as an independent body with its own thermophysical and geometrical characteristics. Validation of these model-predicted results was done using water-quench experiments performed on samples of steel and aluminum plates at the University of British Columbia. Using the Biot number (Bi), a simple criterion is developed to determine when the TC hole needs to be included in the heat-transfer analysis.