Beyond dataset bias: Multi-task unaligned shared knowledge transfer

Many visual datasets are traditionally used to analyze the performance of different learning techniques. The evaluation is usually done within each dataset, therefore it is questionable if such results are a reliable indicator of true generalization ability. We propose here an algorithm to exploit the existing data resources when learning on a new multiclass problem. Our main idea is to identify an image representation that decomposes orthogonally into two subspaces: a part specific to each dataset, and a part generic to, and therefore shared between, all the considered source sets. This allows us to use the generic representation as un-biased reference knowledge for a novel classification task. By casting the method in the multi-view setting, we also make it possible to use different features for different databases. We call the algorithm MUST, Multitask Unaligned Shared knowledge Transfer. Through extensive experiments on five public datasets, we show that MUST consistently improves the cross-datasets generalization performance. © 2013 Springer-Verlag.

Drop speeds from drop-on-demand ink-jet print heads

Measured drop speeds from a range of industrial drop-on-demand (DoD) ink-jet print head designs scale with the predictions of very simple physical models and results of numerical simulations. The main drop/jet speeds at a specified stand-off depend on fluid properties, nozzle exit diameter, and print head drive amplitude for fixed waveform timescales. Drop speeds from the Xaar, Spectra Dimatix, and MicroFab DoD print heads tested with (i) Newtonian, (ii) weakly elastic, and (iii) highly shear-thinning fluids all show a characteristic linear rise with drive voltage (setting) above an apparent threshold drive voltage. Jetting, simple modeling approaches, and numerical simulations of Newtonian fluids over the typical DoD printing range of surface tensions and viscosities were studied to determine how this threshold drive value and the slope of the characteristic linear rise depend on these fluid properties and nozzle exit area. The final speed is inversely proportional to the nozzle exit area, as expected from volume conservation. These results should assist specialist users in the development and optimization of DoD applications and print head design. For a given density, the drive threshold is determined primarily by viscosity, and the constant of proportionality k linking speed with drive above a drive threshold becomes independent of viscosity and surface tension for more viscous DoD fluid jetting. © 2013 Society for Imaging Science and Technology.

A 40 Gb/s scalable optical polymer backplane using a regenerative shared bus architecture

The first multi-channel optical backplane demonstrator using on-board multimode polymer waveguides and a scalable shared-bus regenerative architecture is reported. The system allows bus extension by cascading multiple polymeric bus modules, and enables error-free 4×10 Gb/s interconnection between any two card interfaces on the bus.

Evidence of print gap airflow affecting web printing quality

The current study extends our earlier investigation on the real-time dynamics of print gap airflow around a single jetted drop over a moving substrate. In the present work, simulated web press printing was performed using a stationary grey-scale commercial inkjet print head to print full-width block of solid colour images onto a paper substrate with extended print gaps. The resultant printed images exhibit patterns or 'wood-graining' effects which become more prevalent as the relevant Reynolds number (Re) increases. The high-resolution scans of the printed images revealed that the patterns are created by oscillation and coalescence of neighboring printed tracks across the web. The phenomenon could be a result of drop stream perturbations caused by unsteady print gap airflow of the type similar to that observed in the previous study. ©2013; Society for Imaging Science and Technology.

On the performance analysis and design of a novel shared-layer integrated devices using RCE-p-i-n-PD/SHBT - art. no. 67820J

We have explored the shared-layer integration fabrication of an resonant-cavity-enhanced p-i-n photodector (RCE- p-i-n-PD) and a single heterojunction bipolar transistor (SHBT) with the same epitaxy grown layer structure. MOCVD growth of the different layer structure for the GaAs based RCE- p-i-n-PD/SHBT require compromises to obtain the best performance of the integrated devices. The SHBT is proposed with super-lattice in the collector, and the structure of the base and the collector of the SHBT is used for the RCE. Up to now, the DC characteristics of the integrated device have been obtained.

A Parallel Crossbar Routing Chip for a Shared Memory Multiprocessor

This thesis describes the design and implementation of an integrated circuit and associated packaging to be used as the building block for the data routing network of a large scale shared memory multiprocessor system. A general purpose multiprocessor depends on high-bandwidth, low-latency communications between computing elements. This thesis describes the design and construction of RN1, a novel self-routing, enhanced crossbar switch as a CMOS VLSI chip. This chip provides the basic building block for a scalable pipelined routing network with byte-wide data channels. A series of RN1 chips can be cascaded with no additional internal network components to form a multistage fault-tolerant routing switch. The chip is designed to operate at clock frequencies up to 100Mhz using Hewlett-Packard's HP34 $1.2\\mu$ process. This aggressive performance goal demands that special attention be paid to optimization of the logic architecture and circuit design.