In this paper, we propose a new method for computing and applying language model look-ahead in a dynamic network decoder, exploiting the sparseness of backing-off n-gram language models. Only partial (sparse) look-ahe...
详细信息
ISBN:
(纸本)9781457705397
In this paper, we propose a new method for computing and applying language model look-ahead in a dynamic network decoder, exploiting the sparseness of backing-off n-gram language models. Only partial (sparse) look-ahead tables are computed, with a size that depends on the number of words that have an n-gram score in the language model for a specific context, rather than a constant, vocabulary dependent size. Since high order backing-off language models are inherently sparse, this mechanism reduces the runtime- and memory effort of computing the look-ahead tables by magnitudes. A modified decoding algorithm is required to apply these sparse LM look-ahead tables efficiently. We show that sparse LM look-ahead is much more efficient than the classical method, and that full n-gram look-ahead becomes favorable over lower order look-ahead even when many distinct LM contexts appear during decoding.
Computerized human face recognition is a complex task of deformable patternrecognition. The principal source of complexities lies in the significant inter-class overlapping of faces due to the variations caused by di...
详细信息
We propose a new method for shape recognition and retrieval based on dynamic programming. Our approach uses the dynamic programming algorithm to compute the optimal score and to find the optimal alignment between two ...
详细信息
The proceedings contain 39 papers. The topics discussed include: planar implementation of butler matrix feed network for a switched muitibeam antenna: a survey;comparative performance analysis of various training algo...
ISBN:
(纸本)9781467301329
The proceedings contain 39 papers. The topics discussed include: planar implementation of butler matrix feed network for a switched muitibeam antenna: a survey;comparative performance analysis of various training algorithms for control of CSTR process using NARMA-L2 control;a novel approach for face recognition and age estimation using local binary pattern, discriminative approach using two layered back propagation network;a comparative study of fuzzy classifiers on heart data;rough set analysis for uncertain data classification;shape, texture and local movement hand gesture features for Indian sign language recognition;efficient web service discovery model based on QOS and meta data instances;3D modeling of human faces- a survey;a novel image retrieval technique for enhanced telemedical applications;and a novel image retrieval technique for enhanced telemedical applications.
The traditional set covering models did not consider the equity of users that carried by each set. We developed a new method to solve that for the Integrated Delivering System (IDS) in medical service in Taichung city...
详细信息
The traditional set covering models did not consider the equity of users that carried by each set. We developed a new method to solve that for the Integrated Delivering System (IDS) in medical service in Taichung city. This method requires grouping the clinics in equal market share in a way that satisfy the equity conditions given. It uses Voronoi blocks as market share for each clinic within the study area and calculates the house and clinic ratios as the equity conditions. For each scan, the selected block merges with one of the adjacent blocks using five conditions with different weights. This process continues until the objective functions are met. This model opens up a new way of thinking for set covering by grouping the block objects to obtain optimal result. The result is workable and can be reproduced by computer program if the same data set existed.
We have developed a parameterized (fast) simulation for detector optimization and physics reach studies of the proposed SuperB Flavor Factory in Italy. Detector components are modeled as thin sections of planes, cylin...
详细信息
We have developed a parameterized (fast) simulation for detector optimization and physics reach studies of the proposed SuperB Flavor Factory in Italy. Detector components are modeled as thin sections of planes, cylinders, disks or cones. Particle-material interactions are modeled using simplified cross-sections and formulas. Active detectors are modeled using parameterized response functions. Geometry and response parameters are configured using xml files with a custom-designed schema. Reconstruction algorithms adapted from BaBar are used to build tracks and clusters. Multiple sources of background signals can be merged with primary signals. patternrecognition errors are modeled statistically by randomly misassigning nearby tracking hits. Standard BaBar analysis tuples are used as an event output. Hadronic B meson pair events can be simulated at roughly 10Hz.
Analogies play an important role in many reasoning tasks. The presentation will introduce a recently proposed modeling for analogical proportions, i.e. statements of the form A is to B as C is to D, which put four sit...
详细信息
Analogies play an important role in many reasoning tasks. The presentation will introduce a recently proposed modeling for analogical proportions, i.e. statements of the form A is to B as C is to D, which put four situations into comparative relations. The logical representation used for encoding such proportions takes both into account what the four situations have in common and how they differ. Thanks to the use of a propositional logic modeling extended with suitable fuzzy logic connectives, the approach can deal with situations described by features that may be binary- or multiple-valued. Moreover, it will be shown that analogical proportions are related to other logically expressible proportions that are also of interest. The presentation will emphasize the interest of the approach for handling a large variety of forms of reasoning ranging from the solving of IQ tests, to case-based reasoning, interpolative and extrapolative reasoning, and transductive reasoning for classification. Potential for analogical argumentation will be also briefly discussed. The approach does not just rely on the exploitation of similarities between two cases, but rather takes advantage of the parallel made between a situation to evaluate or complete and three other situations.
In this paper we develop two methods that are able to analyze and recognize patterns in time series. The first model is based on analytic programming (AP), which belongs to softcomputing. AP is based as well as geneti...
详细信息
The proceedings contain 87 papers. The topics discussed include: era of cloud computing;an overview on softcomputing techniques;a novel approach for task processing through NEST network in a grid;TCP/IP security prot...
ISBN:
(纸本)9783642225765
The proceedings contain 87 papers. The topics discussed include: era of cloud computing;an overview on softcomputing techniques;a novel approach for task processing through NEST network in a grid;TCP/IP security protocol suite for grid computing architecture;security issues in cloud computing;classification of software quality attributes for service oriented architecture;energy efficiency for software and services on the cloud;evaluation of grid middleware frameworks for execution of MPI applications;virtualization as an engine to drive cloud computing security;efficient task scheduling algorithms for cloud computing environment;efficient grid scheduling with clouds;security concerns in cloud computing;cloud computing: a need for a regulatory body;clustering dynamic class coupling data to measure class reusability pattern;cloud computing in education: make India better with the emerging trends;and QoS for grid systems.
This paper describes the various aspects of the BEDMOND AAL Joint Programme project, which aims at the development of an ICT-based system for the early detection of Alzheimer's disease and other neurodegenerative ...
详细信息
暂无评论