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     

Sound localization in the median sagittal plane by listeners with presbyacusis

In Experiment 1, a group of listeners with substantial hearing loss due to presbyacusis and a group of listeners with normal hearing were given three localization tests: a frontal plane test in which they judged whether sounds came from the left, overhead, or the right; a sagittal plane test in which they judged whether sounds came from directly in front, overhead, or behind; and an elevation test in which they judged the vertical position of sounds coming from in front. The two groups performed similarly on the frontal plane test, which chiefly depended upon their ability to use binaural localization cues. They performed differently on the sagittal plane and elevation tests, for which the predominant localization cues were spectral. The listeners with presbyacusis were substantially less accurate than those with normal hearing in both of these instances. They had particular difficulty judging source elevation, rarely scoring much above chance. Follow-up testing of a group of subjects in the early stages of presbyacusis showed localization performance that was intermediate to the other two groups, but far more like that of the normal-hearing listeners. In Experiment 2, additional tests were run with the following conditions designed to encourage improved performance by listeners with presbyacusic hearing loss: (1) filtering of stimuli to preclude masking of more informative high-frequency components by low frequencies; (2) simplification of the elevation test and greater spatial separation of its loudspeaker sources; and (3) use of hearing aids. Conditions 1 and 2 had no appreciable effect on performance; condition 3 significantly improved presbyacusic listeners' ability to localize in the sagittal plane, particularly when sounds came from the front.     

Laser scattering instrument for real time in-vivo measurement of ciliary activity in human fallopian tubes

Based on a laser light scattering technique and fibre optic probe, we have developed and tested a simple and practical device for real time measurements of ciliary activity in human Fallopian tubes during laparoscopy and laparotomy. A further aim was to investigate the relationship between the ciliary beat frequency (CBF) and the morphology of the ciliary epithelium. The mean +/- SE of CBF in the fimbria and in the ampulla were 5.4 +/- 0.3 Hz and 5.0 +/- 0.1 Hz respectively. Small pieces of fimbria and ampulla epithelium were taken from the same sites at which the CBF was measured, and the percentage of ciliary cells was determined by scanning electron microscopy. A high positive correlation was found between CBF and the percentage of ciliary cells in the fimbria (r = 0.84) and in the ampulla (r = 0.88). The instrument presented in this study provided, for the first time, a quantitative examination of the CBF in intact human Fallopian tubes and may be used for the investigation of ciliary activity in patients with infertility.     

Curvilinear fatigue crack reliability analysis by stochastic boundary element method

In this paper, the stochastic boundary element method, which combines the mixed boundary integral equations method explored in Reference 1 with the first-order reliability method, is developed to study probabilistic fatigue crack growth. Due to the high degree of complexity and non-linearity of the response, direct differentiation coupied with the response-surface method is employed to determine the response gradient. Three random processes, the mode I and mode II. stress intensity factors and the crack direction angle, are included in the expression of the response gradient. The sensitivity of these random processes is determined using a first-order response model. An iteration scheme based on the HL-RF method² is applied to locate the most probable failure point on the limit-state surface. The accuracy and efficiency of the stochastic boundary element method are evaluated by comparing the cumulative distribution function of the fatigue life obtained with Monte Carlo simulation. The reliability index and the corresponding probability of failure are calculated for a fatigue crack growth problem with randomness in the crack geometry, defect geometry, fatigue parameters and external loads. The response sensitivity of each primary random variable at the design point is determined to show its role in the fatigue failure. The variation of each primary random variable at the design point with the change of probability of failure is also presented in numerical examples.     

Discussion of “an interpretation of the carbon redistribution process during aging of high-carbon martensite”

Y. OHMORI and I. TAMURA:Metall. Trans. A, 1992, vol. 23A,     

The forward muon detector of the DELPHI experiment at LEP

The layout and construction of the muon drift chambers equipping the end caps of the DELPHI detector are described. The performance of this forward muon detection system, determined from cosmic ray tests and data collected during the first year of LEP operation, is presented.     

Tight bounds on Rician-type error probabilities and someapplications

Consider the classic problem of evaluating the probability that one Rician random variable exceeds another, possibly correlated, Rician random variable. This probability is given by Stein (1964) in terms of the Marcum's Q-function, which requires numerical integration on the computer for its evaluation. To facilitate application in many digital communication problems, we derive here tight upper and lower bounds on this probability. The bounds are motivated by a classic result in communication theory, namely, the error probability performance of binary orthogonal signaling over the Gaussian channel with unknown carrier phase. Various applications of the bounds are reported, including the evaluation of the bit error probabilities of MDPSK and MPSK with differential detection and generalized differential detection, respectively. The bounds prove to be tight in all cases. Further applications will be reported in the future     

Approach to the patient with pulmonary disease

The approach to the HIV-infected patient with pulmonary disease is summarized by the algorithms in Figures 3 and 4. These are not intended to be followed in a rigid step-wise fashion. Rather, the practitioner's knowledge of the patient with his or her accompanying medical risks influences the path taken, including the depth and the speed of the evaluation. For example, the patient with cough who is afebrile and breathing at 18 breaths a minute, with a normal chest radiograph and a CD4 count of 350 cells/mm3, is reasonably treated with a macrolide or cephalosporin for bacterial bronchitis and clinical follow-up while awaiting cultures (see Fig. 4). A febrile patient with a cough productive of thin mucus, but known to have a CD4 count of 60 cells/mm3 should be started on anti-PCP therapy while being evaluated for PCP with an induced sputum and if nondiagnostic, a bronchoscope despite a normal chest radiograph. Screening can be as simple as placing an oximeter on the patient's finger in the clinic. If the oxygen saturation of a patient with a normal chest radiograph is low, then the patient should be hospitalized and begun on treatment for PCP while diagnostic evaluation is initiated. If the oxygen saturation is normal, the patient can be exercised to elicit desaturation. If there is no desaturation, PCP is unlikely. If the results are equivocal (i.e., a decrease in saturation, but less than 3%), rest and exercise arterial blood gases can be performed, along with a Dlco-Gallium scanning can be done in patients known to have abnormal Dlco or those who cannot exercise. Patients with focal infiltrates who have acute onset of symptoms (see Fig. 4) commonly have bacterial infections, but the possibility of PCP or TB should not be dismissed. Induced sputum should be examined if TB or PCP is suspected. Patients who are severely ill might go quickly to bronchoscopy without awaiting improvement on empiric therapy. The patient with diffuse infiltrates (see Fig. 4) needs no screening because the presence of disease is apparent from the radiograph. The diagnostic part quickly leads to bronchoscopy for these patients and the initiation of therapy for PCP when suspected. In patients with known pulmonary KS, gallium scanning can be helpful to rule out acute infection, but bronchoscopy is warranted if the patient is severely ill, or at high risk for PCP. This approach should avoid unnecessary procedures in patients with simple bacterial infections, without missing opportunistic infections and tumors.     

On a Mixture GARCH Time‐Series Model

Abstract. Recently, there has been a lot of interest in modelling real data with a heavy‐tailed distribution. A popular candidate is the so‐called generalized autoregressive conditional heteroscedastic (GARCH) model. Unfortunately, the tails of GARCH models are not thick enough in some applications. In this paper, we propose a mixture generalized autoregressive conditional heteroscedastic (MGARCH) model. The stationarity conditions and the tail behaviour of the MGARCH model are studied. It is shown that MGARCH models have tails thicker than those of the associated GARCH models. Therefore, the MGARCH models are more capable of capturing the heavy‐tailed features in real data. Some real examples illustrate the results.     

Isolation of the trail recruitment pheromone ofSolenopsis invicta

TheSolenopsis invicta trail pheromone is synthesized by the Dufour's gland and is released through the sting apparatus. The recruitment subcategory of theS. invicta trail pheromone was shown to be composed of a mixture of the orientation pheromone, (Z,E)--farnesene and an unidentified homosesquiterpene consisting of three rings and one double bond (C-1). C-1 is present in worker Dufour's glands at only 75 pg per worker equivalent. This is the first report that demonstrates that different exocrine products from the same gland control different subcategories of behavior related to mass recruitment.