Large-scale simulation studies in image pattern recognition

Many obstacles to progress in image pattern recognition result from the fact that per-class distributions are often too irregular to be well-approximated by simple analytical functions. Simulation studies offer one way to circumvent these obstacles. We present three closely related studies of machine-printed character recognition that rely on synthetic data generated pseudo-randomly in accordance with an explicit stochastic model of document image degradations. The unusually large scale of experiments - involving several million samples that makes this methodology possible have allowed us to compute sharp estimates of the intrinsic difficulty (Bayes risk) of concrete image recognition problems, as well as the asymptotic accuracy and domain of competency of classifiers     

Design and performance evaluation of a banyan network basedinterconnection structure for ATM switches

This paper presents a new self-routing packet network called the plane interconnected parallel network (PIPN). In the proposed design, the traffic arriving at the network is shaped and routed through two banyan network based interconnected planes. The interconnections between the planes distribute the incoming load more homogeneously over the network. The throughput of the network under uniform and heterogeneous traffic requirements is studied analytically and by simulation. The results are compared with the results of the baseline network and another banyan network based parallel interconnection network. It is shown that, for the proposed design, a higher degree of heterogeneity results in better performance     

A fully soft-switched extended-period quasi-resonantpower-factor-correction circuit

This paper presents the design criteria, procedure, and implementation of a soft-switched power-factor-correction (PFC) circuit based on the extended-period quasi-resonant (EPQR) principles. All power electronic devices including switches and diodes in the circuit are fully soft switched. The design method is demonstrated in a prototype circuit. The operating principles are confirmed with computer simulation and experimental results. A comparison of the EP-QR operation and zero-voltage-transition (ZVT) pulse-width modulation (PWM) method     

Malignant bone pain: pathophysiology and treatment

2-acetyl aminofluorene (AAF) reacts in acidic conditions with nitrous fume yielding N-nitroso-AAF (N-NO-AAF), as previously described, that exerts more toxic and mutagenic effects than its parental compound. In this study, the effect of sodium nitrite (NaNO2) on the tumorigenicity of AAF in rats fed with AAF and NaNO2 was observed. Wistar rats were divided into five groups: group I served as control; group II were treated with NaNO2 (0.3%); group III was given 0.02% AAF alone; groups IV and V received both AAF and NaNO2 (0.2 and 0.3% respectively) in their diet for 12 weeks. At the end of the experiment, all rats in groups III, IV and V developed early stage phenomena of hepatocellular carcinoma, including hepatomegaly with variable-sized foci and neoplastic nodules. Severe damage was observed in the rats treated with AAF and NaNO2. Feeding of AAF (0.02%) for 3 months elevated the levels of c-Fos, c-Jun and c-Myc proteins in the rat livers. The AAF-induced c-Jun, c-Fos and c-Myc expressions were significantly magnified (P < 0.001) by NaNO2. These data confirmed that the strengthening of AAF-induced hepatocarcinogenesis by NaNO2 should be associated with its enhancing effect on the AAF-induced increases in the expressions of c-Jun, c-Fos and c-Myc.     

Error in the UNITY substitution rule for subscripted operators

UNITY, introduced by Chandy and Misra [ChM88], is a programming logic intended to reason about temporal properties of distributed programs. Despite the fact that UNITY does not have the full power of, for example, linear temporal logic, it enjoys popularity due to its simplicity.There was however a serious problem with the Substitution Rule. The logic is incomplete without the rule, and with the rule it is inconsistent.     

Prolog/Rex-a way to extend Prolog for better knowledgerepresentation

Prolog/Rex represents a powerful amalgamation of the latest techniques for knowledge representation and processing, rich in semantic features that ease the difficult task of encoding heterogeneous knowledge of real-world applications. The Prolog/Rex concept mechanism lets a user represent domain entities in terms of their structural and behavioral properties, including multiple inheritance, arbitrary user-defined relations among entities, annotated values (demons), incomplete knowledge, etc. A flexible rule language helps the knowledge engineer capture human expertise and provide flexible control of the reasoning process. Additional Prolog/Rex strength that cannot be found in any other hybrid language made on top of Prolog is language level support for keeping many potentially contradictory solutions to a problem, allowing possible solutions and their implications to be automatically generated and completely explored before they are committed. The same mechanism is used to model time-states, which are useful in planning and scheduling applications of Prolog/Rex     

Weight smoothing to improve network generalization

A weight smoothing algorithm is proposed in this paper to improve a neural network's generalization capability. The algorithm can be used when the data patterns to be classified are presented on an n-dimensional grid (n>/=1) and there exists some correlations among neighboring data points within a pattern. For a fully-interconnected feedforward net, no such correlation information is embedded into the architecture. Consequently, the correlations can only be extracted through sufficient amount of network training. With the proposed algorithm, a smoothing constraint is incorporated into the objective function of backpropagation to reflect the neighborhood correlations and to seek those solutions that have smooth connection weights. Experiments were performed on problems of waveform classification, multifont alphanumeric character recognition, and handwritten numeral recognition. The results indicate that (1) networks trained with the algorithm do have smooth connection weights, and (2) they generalize better.     

Network information criterion-determining the number of hiddenunits for an artificial neural network model

The problem of model selection, or determination of the number of hidden units, can be approached statistically, by generalizing Akaike's information criterion (AIC) to be applicable to unfaithful (i.e., unrealizable) models with general loss criteria including regularization terms. The relation between the training error and the generalization error is studied in terms of the number of the training examples and the complexity of a network which reduces to the number of parameters in the ordinary statistical theory of AIC. This relation leads to a new network information criterion which is useful for selecting the optimal network model based on a given training set.