The classification, fusion, and parallelization of array languageprimitives

We present a classification scheme for array language primitives that quantifies the variation in parallelism and data locality that results from the fusion of any two primitives. We also present an algorithm based on this scheme that efficiently determines when it is beneficial to fuse any two primitives. Experimental results show that five LINPACK routines report 50% performance improvement from the fusion of array operators     

Unicast-based multicast communication in wormhole-routed networks

Multicast communication, in which the same message is delivered from a source node to an arbitrary number of destination nodes, is being increasingly demanded in parallel computing. System supported multicast services can potentially offer improved performance, increased functionality, and simplified programming, and may in turn be used to support various higher-level operations for data movement and global process control. This paper presents efficient algorithms to implement multicast communication in wormhole-routed direct networks, in the absence of hardware multicast support, by exploiting the properties of the switching technology. Minimum-time multicast algorithms are presented for n-dimensional meshes and hypercubes that use deterministic, dimension-ordered routing of unicast messages. Both algorithms can deliver a multicast message to m-1 destinations in [log ₂ m] message passing steps, while avoiding contention among the constituent unicast messages. Performance results of implementations on a 64-node nCUBE-2 hypercube and a 168-node Symult 2010 2-D mesh are given     

Quantization effects in digitally behaving circuit implementationsof Kohonen networks

Implementing a neural network on a digital or mixed analog and digital chip yields the quantization of the synaptic weights dynamics. This paper addresses this topic in the case of Kohonen's self-organizing maps. We first study qualitatively how the quantization affects the convergence and the properties, and deduce from this analysis the way to choose the parameters of the network (adaptation gain and neighborhood). We show that a spatially decreasing neighborhood function is far more preferable than the usually rectangular neighborhood function, because of the weight quantization. Based on these results, an analog nonlinear network, integrated in a standard CMOS technology, and implementing this spatially decreasing neighborhood function is then presented. It can be used in a mixed analog and digital circuit implementation.     

Connectivity and performance tradeoffs in the cascade correlationlearning architecture

The cascade correlation is a very flexible, efficient and fast algorithm for supervised learning. It incrementally builds the network by adding hidden units one at a time, until the desired input/output mapping is achieved. It connects all the previously installed units to the new unit being added. Consequently, each new unit in effect adds a new layer and the fan-in of the hidden and output units keeps on increasing as more units get added. The resulting structure could be hard to implement in VLSI, because the connections are irregular and the fan-in is unbounded. Moreover, the depth or the propagation delay through the resulting network is directly proportional to the number of units and can be excessive. We have modified the algorithm to generate networks with restricted fan-in and small depth (propagation delay) by controlling the connectivity. Our results reveal that there is a tradeoff between connectivity and other performance attributes like depth, total number of independent parameters, and learning time.     

The design, fabrication, and testing of corrugated silicon nitridediaphragms

Silicon nitride corrugated diaphragms of 2 mm×2 mm×1 μm have been fabricated with 8 circular corrugations, having depths of 4, 10, or 14 μm. The diaphragms with 4-μm-deep corrugations show a measured mechanical sensitivity (increase in the deflection over the increase in the applied pressure) which is 25 times larger than the mechanical sensitivity of flat diaphragms of equal size and thickness. Since this gain in sensitivity is due to reduction of the initial stress, the sensitivity can only increase in the case of diaphragms with initial stress. A simple analytical model has been proposed that takes the influence of initial tensile stress into account. The model predicts that the presence of corrugations increases the sensitivity of the diaphragms, because the initial diaphragm stress is reduced. The model also predicts that for corrugations with a larger depth the sensitivity decreases, because the bending stiffness of the corrugations then becomes dominant. These predictions have been confirmed by experiments. The application of corrugated diaphragms offers the possibility to control the sensitivity of thin diaphragms by geometrical parameters, thus eliminating the effect of variations in the initial stress, due to variations in the diaphragm deposition process and/or the influence of temperature changes and packaging stress     

First break refraction event picking using fuzzy logic systems

First break picking is a pattern recognition problem in seismic signal processing, one that requires much human effort and is difficult to automate. The authors' goal is to reduce the manual effort in the picking process and accurately perform the picking. Feedforward neural network first break pickers have been developed using backpropagation training algorithms applied either to an encoded version of the raw data or to derived seismic attributes which are extracted from the raw data. The authors summarize a study in which they applied a backpropagation fuzzy logic system (BPFLS) to first break picking. The authors use derived seismic attributes as features, and take lateral variations into account by using the distance to a piecewise linear guiding function as a new feature. Experimental results indicate that the BPFLS achieves about the same picking accuracy as a feedforward neural network that is also trained using a backpropagation algorithm; however, the BPFLS is trained in a much shorter time, because there is a systematic way in which the initial parameters of the BPFLS can be chosen, versus the random way in which the weights of the neural network are chosen     

Assessment of annihilation anxiety from projective tests

This report details procedures to measure annihilation anxiety, a concept derived from Freud's 1926 formulation of traumatic anxiety. A 25-item pencil-and-paper inventory administered to patient and to nonpatient samples is described, along with a brief summary of earlier findings. The delineation of nine interrelated experiential components of annihilation anxiety provides the background for the construction of Rorschach and TAT measures of the concept. Findings comparing the pencil-and-paper inventory and the projective test measures are presented as well as examples of responses judged to reflect annihilation anxiety from Rorschach and TAT protocols.