A hybrid generative/discriminative framework to train a semantic parser from an un-annotated corpus

We propose a hybrid generative/discriminative framework for semantic parsing which combines the hidden vector state (HVS) model and the hidden Markov support vector machines (HM-SVMs). The HVS model is an extension of the basic discrete Markov model in which context is encoded as a stack-oriented state vector. The HM-SVMs combine the advantages of the hidden Markov models and the support vector machines. By employing a modified K-means clustering method, a small set of most representative sentences can be automatically selected from an un-annotated corpus. These sentences together with their abstract annotations are used to train an HVS model which could be subsequently applied on the whole corpus to generate semantic parsing results. The most confident semantic parsing results are selected to generate a fully-annotated corpus which is used to train the HM-SVMs. The proposed framework has been tested on the DARPA Communicator Data. Experimental results show that an improvement over the baseline HVS parser has been observed using the hybrid framework. When compared with the HM-SVMs trained from the fully-annotated corpus, the hybrid framework gave a comparable performance with only a small set of lightly annotated sentences. © 2008. Licensed under the Creative Commons.

Transition of planar Couette flow at infinite Reynolds numbers

An outline of the state space of planar Couette flow at high Reynolds numbers (Re<105) is investigated via a variety of efficient numerical techniques. It is verified from nonlinear analysis that the lower branch of the hairpin vortex state (HVS) asymptotically approaches the primary (laminar) state with increasing Re. It is also predicted that the lower branch of the HVS at high Re belongs to the stability boundary that initiates a transition to turbulence, and that one of the unstable manifolds of the lower branch of HVS lies on the boundary. These facts suggest HVS may provide a criterion to estimate a minimum perturbation arising transition to turbulent states at the infinite Re limit. © 2013 American Physical Society.

Semi-supervised learning of statistical models for natural language understanding

Natural language understanding is to specify a computational model that maps sentences to their semantic mean representation. In this paper, we propose a novel framework to train the statistical models without using expensive fully annotated data. In particular, the input of our framework is a set of sentences labeled with abstract semantic annotations. These annotations encode the underlying embedded semantic structural relations without explicit word/semantic tag alignment. The proposed framework can automatically induce derivation rules that map sentences to their semantic meaning representations. The learning framework is applied on two statistical models, the conditional random fields (CRFs) and the hidden Markov support vector machines (HM-SVMs). Our experimental results on the DARPA communicator data show that both CRFs and HM-SVMs outperform the baseline approach, previously proposed hidden vector state (HVS) model which is also trained on abstract semantic annotations. In addition, the proposed framework shows superior performance than two other baseline approaches, a hybrid framework combining HVS and HM-SVMs and discriminative training of HVS, with a relative error reduction rate of about 25% and 15% being achieved in F-measure.

Altered resting-state EEG source functional connectivity in schizophrenia:the effect of illness duration

Despite the increasing body of evidence supporting the hypothesis of schizophrenia as a disconnection syndrome, studies of resting-state EEG Source Functional Connectivity (EEG-SFC) in people affected by schizophrenia are sparse. The aim of the present study was to investigate resting-state EEG-SFC in 77 stable, medicated patients with schizophrenia (SCZ) compared to 78 healthy volunteers (HV). In order to study the effect of illness duration, SCZ were divided in those with a short duration of disease (SDD; n = 25) and those with a long duration of disease (LDD; n = 52). Resting-state EEG recordings in eyes closed condition were analyzed and lagged phase synchronization (LPS) indices were calculated for each ROI pair in the source-space EEG data. In delta and theta bands, SCZ had greater EEG-SFC than HV; a higher theta band connectivity in frontal regions was observed in LDD compared with SDD. In the alpha band, SCZ showed lower frontal EEG-SFC compared with HV whereas no differences were found between LDD and SDD. In the beta1 band, SCZ had greater EEG-SFC compared with HVs and in the beta2 band, LDD presented lower frontal and parieto-temporal EEG-SFC compared with HV. In the gamma band, SDD had greater connectivity values compared with LDD and HV. This study suggests that resting state brain network connectivity is abnormally organized in schizophrenia, with different patterns for the different EEG frequency components and that EEG can be a powerful tool to further elucidate the complexity of such disordered connectivity.

Impaired brainstem and thalamic high-frequency oscillatory EEG activity in migraine between attacks

INTRODUCTION: We investigated whether interictal thalamic dysfunction in migraine without aura (MO) patients is a primary determinant or the expression of its functional disconnection from proximal or distal areas along the somatosensory pathway. METHODS: Twenty MO patients and twenty healthy volunteers (HVs) underwent an electroencephalographic (EEG) recording during electrical stimulation of the median nerve at the wrist. We used the functional source separation algorithm to extract four functionally constrained nodes (brainstem, thalamus, primary sensory radial, and primary sensory motor tangential parietal sources) along the somatosensory pathway. Two digital filters (1-400 Hz and 450-750 Hz) were applied in order to extract low- (LFO) and high- frequency (HFO) oscillatory activity from the broadband signal. RESULTS: Compared to HVs, patients presented significantly lower brainstem (BS) and thalamic (Th) HFO activation bilaterally. No difference between the two cortical HFO as well as in LFO peak activations between the two groups was seen. The age of onset of the headache was positively correlated with HFO power in the right brainstem and thalamus. CONCLUSIONS: This study provides evidence for complex dysfunction of brainstem and thalamocortical networks under the control of genetic factors that might act by modulating the severity of migraine phenotype.

Image usefulness of compressed surveillance footage with different scene contents

The police use both subjective (i.e. police staff) and automated (e.g. face recognition systems) methods for the completion of visual tasks (e.g person identification). Image quality for police tasks has been defined as the image usefulness, or image suitability of the visual material to satisfy a visual task. It is not necessarily affected by any artefact that may affect the visual image quality (i.e. decrease fidelity), as long as these artefacts do not affect the relevant useful information for the task. The capture of useful information will be affected by the unconstrained conditions commonly encountered by CCTV systems such as variations in illumination and high compression levels. The main aim of this thesis is to investigate aspects of image quality and video compression that may affect the completion of police visual tasks/applications with respect to CCTV imagery. This is accomplished by investigating 3 specific police areas/tasks utilising: 1) the human visual system (HVS) for a face recognition task, 2) automated face recognition systems, and 3) automated human detection systems. These systems (HVS and automated) were assessed with defined scene content properties, and video compression, i.e. H.264/MPEG-4 AVC. The performance of imaging systems/processes (e.g. subjective investigations, performance of compression algorithms) are affected by scene content properties. No other investigation has been identified that takes into consideration scene content properties to the same extend. Results have shown that the HVS is more sensitive to compression effects in comparison to the automated systems. In automated face recognition systems, `mixed lightness' scenes were the most affected and `low lightness' scenes were the least affected by compression. In contrast the HVS for the face recognition task, `low lightness' scenes were the most affected and `medium lightness' scenes the least affected. For the automated human detection systems, `close distance' and `run approach' are some of the most commonly affected scenes. Findings have the potential to broaden the methods used for testing imaging systems for security applications.

The fastest unbound star in our Galaxy ejected by a thermonuclear supernova

Hypervelocity stars (HVSs) travel with velocities so high that they exceed the escape velocity of the Galaxy. Several acceleration mechanisms have been discussed. Only one HVS (US 708, HVS 2) is a compact helium star. Here we present a spectroscopic and kinematic analysis of US 708. Traveling with a velocity of ∼1200 kilometers per second, it is the fastest unbound star in our Galaxy. In reconstructing its trajectory, the Galactic center becomes very unlikely as an origin, which is hardly consistent with the most favored ejection mechanism for the other HVSs. Furthermore, we detected that US 708 is a fast rotator. According to our binary evolution model, it was spun-up by tidal interaction in a close binary and is likely to be the ejected donor remnant of a thermonuclear supernova.

Mecanismos de optimización en el procesado para H.265/HEVC

En esta memoria se ha implementado una etapa de preprocesado que sirva como primera fase en el proceso de codificación de vídeo. Esta etapa integra dos variedades del filtro de mediana (3×3 y 5×5) y un operador. Dicho operador lleva a cabo el cálculo del gradiente de los píxeles que conforman una imagen o fotograma con objeto de filtrar después aquellos que están por debajo de un determinado valor (threshold). El cálculo de dicho threshold se realiza de manera empírica mediante dos procesos distintos. En el primero se obtienen valores de luminancia y crominancia de píxeles que integran bordes para encontrar aquel que tenga el valor mínimo, mientras que en el segundo se calcula la tasa de píxeles que forman parte de bordes. Una vez se ha realizado el cálculo anterior, se han utilizado distintos valores de threshold, distintas variedades de filtro de mediana y distintos valores de QP (calidad) con objeto de parametrizar las codificaciones que hacen uso de esta nueva etapa. Posteriormente a dichas codificaciones, se han obtenido los tamaños de los bitstreams de salida y se ha evaluado la calidad de los vídeos decodificados o reconstruidos mediante dos métricas objetivas: PSNR y SSIM. Las codificaciones que no utilizan etapa de preprocesado también han sido evaluadas mediante dichas métricas y comparadas con aquellas que sí integran dicha etapa. Los resultados obtenidos dejan patente el compromiso existente entre tamaño de bitstream y calidad, siendo más representativos los de la métrica SSIM, estando esta última más relacionada con la percepción de la imagen por parte del HVS (sistema visual humano). Como resultado, se obtiene para esta métrica tasas de compresión mayores que las alcanzadas sin preprocesamiento, con pérdidas de calidad prácticamente inapreciables.