Bounds on the number of samples needed for neural learning.

The relationship between the number of hidden nodes in a neural network, the complexity of a multiclass discrimination problem, and the number of samples needed for effect learning are discussed. Bounds for the number of samples needed for effect learning are given. It is shown that Omega(min (d,n) M) boundary samples are required for successful classification of M clusters of samples using a two-hidden-layer neural network with d-dimensional inputs and n nodes in the first hidden layer.     

Intravaginal Controlled Administration of Flurogestone Acetate: (III) Development of Rate-Control Vaginal Devices

A Rate-Control vaginal device was developed which overcomes the low bioavailability and unpredictable Q - t^1/2 type release and absorption rate profiles of flurogestone acetate delivered by the currently marketed Syncro-Mate pessary.

The in vitro release and vaginal absorption profiles from the Rate-Control vaginal device were run simultaneously, a linear Q - t relationship was obtained with a significant improvement in bioavailability. A mathematical model was developed to correlate the in vitro drug release and the vaginal absorption profiles of flurogestone acetate from the vaginal devices.

The design, development and the simultaneous release and absorption profiles of flurogestone acetate from this new vaginal device were outlined and discussed.     

Distributed memory systems for simulating artificial neural networks

In executing tasks involving intelligent information processing, the human brain performs better than the digital computer. The human brain derives its power from a large number [O(10¹¹)] of neurons which are interconnected by a dense interconnection network [O(10⁵) connections per neuron]. Artificial neural network (ANN) paradigms adopt the structure of the brain to try to emulate the intelligent information processing methods of the brain. ANN techniques are being employed to solve problems in areas such as pattern recognition, and robotic processing. Simulation of ANNs involves implementation of large number of neurons and a massive interconnection network. In this paper, we discuss various simulation models of ANNs and their implementation on distributed memory systems. Our investigations reveal that communication-efficient networks of distributed memory systems perform better than other topologies in implementing ANNs.     

Adaptive Spatial Partitioning for Multidimensional Data Streams

We propose a space-efficient scheme for summarizing multidimensional data streams. Our sketch can be used to solve spatial versions of several classical data stream queries efficiently. For instance, we can track ε-hot spots, which are congruent boxes containing at least an ε fraction of the stream, and maintain hierarchical heavy hitters in d dimensions. Our sketch can also be viewed as a multidimensional generalization of the ε-approximate quantile summary. The space complexity of our scheme is O((1/ε) log R) if the points lie in the domain [0, R]^d, where d is assumed to be a constant. The scheme extends to the sliding window model with a log (ε n) factor increase in space, where n is the size of the sliding window. Our sketch can also be used to answer ε-approximate rectangular range queries over a stream of d-dimensional points.     

A systematic design method to obtain broadband characteristics for singly-fed electromagnetically coupled patch antennas for circular polarization

We present a systematic, empirical design technique to obtain optimum broadband impedance, axial-ratio (AR) and gain bandwidths for a singly-fed electromagnetically coupled patch antenna for circular polarization. Our investigation has also revealed tradeoffs amongst obtainable AR, impedance bandwidth and AR bandwidth. Using two design examples at different frequency bands and for different senses of circular polarization, we have demonstrated the effectiveness of the proposed knowledge-based tuning method. We have obtained at C-band measured values of impedance bandwidth (VSWR/spl les/2) equal to 43%, 3-dB AR bandwidth of 8%, AR of less than 0.3 dB and a mean gain level of 7 dB. For the Ku-band element, a 40% impedance bandwidth and a 17.3% of 3-dB AR bandwidth have been obtained with a peak gain of 7.2 dBic.     

An efficient TDMA start-up and restart synchronization approach for distributed embedded systems

A desired attribute in safety-critical embedded real-time systems is a system time and event synchronization capability on which predictable communication can be established. Focusing on bus-based communication protocols, we present a novel, efficient, and low-cost start-up and restart synchronization approach for TDMA environments. This approach utilizes information about a node's message length that forms a unique sequence to achieve synchronization such that communication overhead can be avoided. We present a fault-tolerant initial synchronization protocol with a bounded start-up time. The protocol avoids start-up collisions by deterministically postponing retries after a collision. We also present a resynchronization strategy that incorporates recovering nodes into synchronization.     

Defect-free sintering of two material powder injection molded components Part II Model

A defect-free, two-material component can be obtained via co-sintering by suitably altering the powder characteristics or compositions, as demonstrated in Part I. In this paper, a model to ascertain the suitability of material systems to be co-sintered without defects such as delamination or interface pores is presented. The model is based on the management of the stress induced due to the difference in shrinkage and an analysis of the in situ strength of the weaker material during sintering. Tool steel in combination with stainless steel admixed with boron and Fe-2Ni admixed with boron are two systems used to validate the model. The predictions of the model are in good agreement with the observations.