CLUSTERING VERY LARGE DATA SETS USING A LOW MEMORY MATRIX FACTORED REPRESENTATION

A scalable method to cluster data sets too large to fit in memory is presented. This method does not depend on random subsampling, but does scan every individual data sample in a deterministic way. The original data are represented in factored form by the product of two matrices, one or both of which is very sparse. This factored form avoids the need to multiply together these two matrices by using a variant of the Principal Direction Divisive Partitioning (PDDP) algorithm which does not depend on computing the distances between the individual samples. The resulting clustering algorithm is Piecemeal PDDP (PMPDDP), in which the original data are broken up into sections which will fit into memory and clustered. The cluster centers are used to create approximations to the original data items, and each original data item is represented by a linear combination of these centers. We evaluate the performance of PMPDDP on three real data sets, and observe that the quality of the clusters of PMPDDP is comparable to PDDP for the data sets examined.     

All-solid-state tunable continuous-wave ultraviolet source with high spectral purity and frequency stability

We present a novel approach for the generation of higly frequency-stable, widely tunable, single-frequency cw UV light that is suitable for high-resolution spectroscopy. Sum-frequency generation (SFG) of two solid-state sources with a single cavity resonant for both fundamental waves is employed. Using a highly stable, narrow-linewidth frequency-doubled cw Nd:YAG laser as a master laser and slaving to it the SFG cavity and the other fundamental wave from a Ti:sapphire laser, we generate UV radiation of 33-mW output power around 313 nm. Alternatively, we use a diode laser instead of the Ti:sapphire laser and produce an output power of 2.1 mW at 313 nm. With both setups we obtain a continuous tunability of >15 GHz, short-term frequency fluctuations in the submegahertz range, a long-term frequency drift below 100 MHz/h, and stable operation for several hours. The theory of optimized doubly resonant SFG is also given.     

Gabor-QR decomposition for image encoding

A new approach to Gabor transformation based on the QR decomposition, with potential applications to image sequence encoding, is proposed. This Gabor-QR scheme computes the exact Gabor coefficients by solving a system of linear equations Ax=b, which is more efficient than iterative schemes, and less complicated than other matrix-based methods. Experimental results from image encoding are presented.<>     

On the use of superposition in the approximate solution of heat conduction problems

The use of convolution integrals and of the method of images, applied to approximate basic solutions, is investigated. The value of these techniques in deriving approximate solutions of heat conduction problems is discussed, and their accuracy, relative to both known exact solutions and other approximate ones, is assessed.     

Denitrification with polycaprolactone as solid substrate in a laboratory-scale recirculated aquaculture system.

A denitrification system based on the biodegradable polymer Poly-epsilon-Caprolactone (PCL) was tested in a laboratory-scale recirculated aquaculture system with eels in comparison to a reference system without denitrification. The experiments were conducted with fluidized bed reactors in two parallel systems to examine the feasibility and performance of the process and to observe the condition of fishes by measuring weight gain during the test-period. The most evident effect of the system with denitrification was the low nitrate concentration compared with the untreated reference system. A further advantage was the stability of the pH in the systems with denitrification whereas pH of the untreated water decreased due to nitrification. All over the test-period the eels showed a similar weight gain in both systems.     

Direct Writing of Gallium‐Indium Alloy for Stretchable Electronics

In this paper, a direct writing method for gallium‐indium alloys is presented. The relationships between nozzle inner diameter, standoff distance, flow rate, and the resulting trace geometry are demonstrated. The interaction between the gallium oxide layer and the substrate is critically important in understanding the printing behavior of the liquid metal. The difference between receding and advancing contact angles demonstrates that the adhesion of the oxide layer to the substrate surface is stronger than the wetting of the surface by the gallium‐indium alloy. This further demonstrates why free‐standing structures such as the traces described herein can be realized. In addition to the basic characterization of the direct writing process, a design algorithm that is generalizable to a range of trace geometries is developed. This method is applied to the fabrication of an elastomer‐encapsulated strain gauge that displays an approximately linear behavior through 50% strain with a gauge factor of 1.5.     

Computing the Kalman decomposition: An optimal method

In this paper we describe a method for computing the complete Kalman decomposition of a linear dynamical system, and we show that this method is optimal in a certain sense. Specifically, we describe an algorithm to compute a mapping that, when applied to the original system, yields a complete Kalman decomposition. We show that, of all transformations that yield a Kalman decomposition, the one we construct here has the lowest condition number.