A prototype 3D optically interconnected neural network

We report the implementation of a prototype three-dimensional (3D) optoelectronic neural network that combines free-space optical interconnects with silicon-VLSI-based optoelectronic circuits. The prototype system consists of a 16-node input, 4-neuron hidden, and a single-neuron output layer, where the denser input-to-hidden-layer connections are optical. The input layer uses PLZT light modulators to generate optical outputs which are distributed over an optoelectronic neural network chip through space-invariant holographic optical interconnects. Optical interconnections provide negligible fan-out delay and allow compact, purely on-chip electronic H-tree type fan-in structure. The small prototype system achieves a measured 8-bit electronic fan-in precision and a calculated maximum speed of 640 million interconnections per second. The system was tested using synaptic weights learned off system and was shown to distinguish any vertical line from any horizontal one in an image of 4×4 pixels. New, more efficient light detector and small-area analog synapse circuits and denser optoelectronic neuron layouts are proposed to scale up the system. A high-speed, feed-forward optoelectronic synapse implementation density of up to 10⁴/cm² seems feasible using new synapse design. A scaling analysis of the system shows that the optically interconnected neural network implementation can provide higher fan-in speed and lower power consumption characteristics than a purely electronic, crossbar-based neural network implementation     

Formulation Factors Affecting Thimerosal Stability

A high-performance liquid chromatographic (HPLC) assay has been developed for thimerosal and some of its degradation products: thiosalicylic and 2,2'-dithiosalicylic acids. Using this method, the influence of formulation factors as: isotonic agent, initial concentration, addition of tromethamine, pH and container material, over thimerosal stability was studied.     

D-lactic acidosis: a treatable encephalopathy in pediatric patients

Isolation of Salmonella from the feces of impounded dogs was carried out to make clear the recent microbiological condition of dogs introduced into our facilities. Salmonella was isolated from 10 out of 283 samples (3.5%). Inparticular, during the first week after introduction, Salmonella was isolated from nine out of 74 dogs (12.2%). The isolation rate during the first week after introduction was significantly higher than that for dogs introduced later. No isolates were detected from dogs reared for more than 3 weeks after introduction. This study indicates that impounded dogs had the highest risk of infecting a person with Salmonella during the first week after introduction into our laboratory animal facilities, and that we need to pay attention to this fact when handling them.     

The information-theoretic capacity of discrete-time queues

The information-theoretic capacity of continuous-time queues was analyzed recently by Anantharam and Verdu (see ibid. vol.42, p.4-18, 1996). Along similar lines, we analyze the information-theoretic capacity of two models of discrete-time queues. The first model has single packet arrivals and departures in a time slot and independent packet service times, and is the discrete-time analog of the continuous-time model analyzed by Anantharam and Verdu. We show that in this model, the geometric service time distribution plays a role analogous to that of the exponential distribution in continuous-time queues, in that, among all queues in this model with a given mean service time, the queue with geometric service time distribution has the least capacity. The second model allows multiple arrivals in each slot, and the queue is modeled as serving an independent random number of packets in each slot. We obtain upper and lower bounds on the capacity of queues with an arbitrary service distribution within this model, and show that the bounds coincide in the case of the queue that serves a geometrically distributed number of packets in each slot. We also discuss the extremal nature of the geometric service distribution within this model     

Adiabatic couplers: design rules and optimization

Explicit analytic design rules are derived for both 3 dB and full adiabatic couplers. The design rules are in excellent agreement with numerical calculations using the beam propagation method (BPM). It is shown that the length scaling for 3 dB couplers compared to full couplers makes the former more difficult to design. The design for each case is optimized to obtain the upper limit of performance and a comparison is carried out between two different design geometries for both 3 dB and full adiabatic couplers     

Successive interference cancellation for multiuser asynchronousDS/CDMA detectors in multipath fading links

There has been an increasing interest in the use of code-division multiple access (CDMA) in cellular mobile and wireless personal communications. The choice of such multiaccess technique is attractive because of its potential capacity increases and other technical factors such as privacy and multipath rejection capabilities. However, it is well known that the performance of CDMA can be significantly degraded due to cochannel interference (CI) and the near-far effects. We consider the performance of direct-sequence (DS)-based CDMA over fading channels that are modeled as slowly varying Rayleigh-fading discrete multipath channels. Specifically, we propose and analyze an adaptive multistage interference cancellation strategy for the demodulation of asynchronous DS spread-spectrum multiple-access signals. Numerical results show that the proposed multistage detector, which alleviates the detrimental effects of the near-far problem, can significantly improve the system performance