Pattern recognition using discriminative feature extraction

We propose a new design method, called discriminative feature extraction for practical modular pattern recognizers. A key concept of discriminative feature extraction is the design of an overall recognizer in a manner consistent with recognition error minimization. The utility of the method is demonstrated in a Japanese vowel recognition task     

Dynamics of the linear chain with arbitrary potential applied to quantum crystals

In one dimension, many-body matrix elements can be calculated exactly by the transfer integral (TI) method if the wave functions are assumed to be products of single-particle functions and nearest-neighbor correlation functions. In the frame of this TI approximation the phonons can be calculated without any further approximation. The theory is first developed for the case of a harmonic potential. Here the TI phonons are an excellent approximation of the well-known exact results, except for very long wavelengths. However, a Bogoliubov transformation finally yields the exact phonons. The whole procedure can be generalized for arbitrary two-body interactions, including those containing a hard core. In the generalized theory the hard-core problem has vanished completely.     

Self-consistent phonons in configurationally disordered quantum crystals. I. Theory

A generalization of the coherent potential approximation (CPA) to strongly anharmonic systems is presented that takes into account the static relaxations in the vicinity of each defect and the changes in the averaged mean lattice constant and in the mean square fluctuations. In the limits of a perturbed classical, harmonic crystal or an ideal, anharmonic crystal the theory reduces to the conventional CPA and the self-consistent phonon results, respectively. For the numerical solution an approximation is used that starts from the renormalized harmonic phonons in ideal crystals. As an example, some characteristic results are discussed for the systems ofoD₂ orpH₂ in Ne.Deceased.     

Bronchoscopy versus empirical therapy in HIV-infected patients with presumptive Pneumocystis carinii pneumonia. A decision analysis

The outcomes of alternative strategies for the management of pulmonary complications in patients infected with the human immunodeficiency virus (HIV) and with suspected Pneumocystis carinii pneumonia were compared using a decision analysis model. A decision tree was constructed using baseline probabilities derived from published data and expert opinion. The case scenario analyzed was that of a patient not currently receiving anti-Pneumocystis prophylaxis who presents with moderate pulmonary symptoms and fulfills the Centers for Disease Control (CDC) criteria for presumptive P. carinii pneumonia. Two strategies were compared: (1) early bronchoscopy with appropriate therapy based on the results, and (2) empiric treatment for P. carinii (trimethoprim/sulfamethoxazole or pentamidine, and steroids) with delayed bronchoscopy in those not responding to 5 days of empiric therapy. The expected 1-month survival rate (with and without quality of life adjustment) was found to be essentially the same for the two strategies using the baseline probabilities, and the decision remained a toss-up within the clinically relevant range of published probabilities for P. carinii pneumonia in patients fulfilling the CDC criteria. Because early bronchoscopy does not offer any additional survival benefits and is associated with greater costs and disutility, empiric therapy would appear to be the superior management strategy in this scenario.     

Maximization of mutual information for offline Thai handwriting recognition

This paper aims to improve the performance of an HMM-based offline Thai handwriting recognition system through discriminative training and the use of fine-tuned feature extraction methods. The discriminative training is implemented by maximizing the mutual information between the data and their classes. The feature extraction is based on our proposed block-based PCA and composite images, shown to be better at discriminating Thai confusable characters. We demonstrate significant improvements in recognition accuracies compared to the classifiers that are not discriminatively optimized.     

Minimum classification error training for online handwriting recognition

This paper describes an application of the minimum classification error (MCE) criterion to the problem of recognizing online unconstrained-style characters and words. We describe an HMM-based, character and word-level MCE training aimed at minimizing the character or word error rate while enabling flexibility in writing style through the use of multiple allographs per character. Experiments on a writer-independent character recognition task covering alpha-numerical characters and keyboard symbols show that the MCE criterion achieves more than 30 percent character error rate reduction compared to the baseline maximum likelihood-based system. Word recognition results, on vocabularies of 5k to 10k, show that MCE training achieves around 17 percent word error rate reduction when compared to the baseline maximum likelihood system.