检索结果-内蒙古大学图书馆

IEEE Workshop on Spoken Language Technology (SLT)

作者： Bagby, Tom Rao, Kanishka Sim, Khe Chai Google 1600 Amphitheatre Pkwy Mountain View CA 94043 USA

ISBN: (纸本)9781538643341

Recurrent neural network transducer (RNN-T) has been successfully applied to automatic speech recognition to jointly learn the acoustic and language model components. The RNN-T loss and its gradient with respect to the softmax outputs can be computed efficiently using a forward-backward algorithm. In this paper, we present an efficient implementation of the RNN-T forward-backward and Viterbi algorithms using standard matrix operations. This allows us to easily implement the algorithm in TensorFlow by making use of the existing hardware-accelerated implementations of these operations. This work is based on a similar technique used in our previous work for computing the connectionist temporal classification and lattice-free maximum mutual information losses, where the forward and backward recursions are viewed as a bi-directional RNN whose states represent the forward and backward probabilities. Our benchmark results on graphic processing unit (GPU) and tensor processing unit (TPU) show that our implementation can achieve better throughput performance by increasing the batch size to maximize parallel computation. Furthermore, our implementation is about twice as fast on TPU compared to GPU for batch

关键词： recurrent neural network transducer forward-backward algorithm TensorFlow GPU TPU

来源：评论

学校读者我要写书评

暂无评论

Aggregation for low complexity recovery: Oracle inequalities and algorithms

Aggregation for low complexity recovery: Oracle inequalities...

引用

作者： Duy Tung Luu Universite de Caen Normandie

学位级别：博士

In many areas of statistics, including signal and image processing, high-dimensional estimation is an important task to recover an object of interest. However, in the overwhelming majority of cases, the recovery problem is ill-posed. Fortunately, even if the ambient dimension of the object to be restored (signal, image, video) is very large, its intrinsic "complexity" is generally small. The introduction of this prior information can be done through two approaches: (i) penalization (very popular) and (ii) aggregation by exponential weighting (EWA). The penalized approach aims at finding an estimator that minimizes a data loss function penalized by a term promoting objects of low (simple) complexity. The EWA combines a family of pre-estimators, each associated with a weight exponentially promoting the same objects of low complexity. This manuscript consists of two parts: a theoretical part and an algorithmic part. In the theoretical part, we first propose the EWA with a new family of priors promoting analysis-group sparse signals whose performance is guaranteed by oracle inequalities. Next, we will analysis the penalized estimator and EWA, with a general prior promoting simple objects, in a unified framework for establishing some theoretical guarantees. Two types of guarantees will be established: (i) prediction oracle inequalities, and (ii) estimation bounds. We will exemplify them for particular cases some of which studied in the literature. In the algorithmic part, we will propose an implementation of these estimators by combining Monte-Carlo simulation (Langevin diffusion process) and proximal splitting algorithms, and show their guarantees of convergence. Several numerical experiments will be considered for illustrating our theoretical guarantees and our algorithms.

关键词： High-dimensional estimation low-complexity prior exponential weighted aggregation penalized estimation oracle inequality Langevin diffusion forward-backward algorithm consistency

来源：评论

学校读者我要写书评

暂无评论

Exponential weighted aggregation: oracle inequalities and algorithms

Exponential weighted aggregation: oracle inequalities and al...

引用

作者： Duy Tung Luu Normandie Universite

学位级别：博士

关键词： High-dimensional estimation low-complexity prior exponential weighted aggregation penalized estimation oracle inequality Langevin diffusion forward-backward algorithm consistency

来源：评论

学校读者我要写书评

暂无评论

Application of the principle of maximum conformality to the top-quark forward-backward asymmetry at the Tevatron

引用

Physical Review D 2012年第11期85卷 114040-114040页

作者： Stanley J. Brodsky Xing-Gang Wu SLAC National Accelerator Laboratory 2575 Sand Hill Road Menlo Park California 94025 USA Department of Physics Chongqing University Chongqing 401331 People’s Republic of China

The renormalization scale uncertainty can be eliminated by the principle of maximum conformality (PMC) in a systematic, scheme-independent way. Applying the PMC for the tt¯-pair hadroproduction at the next-to-next-to-leading-order level, we have found that the total cross sections σtt¯ at both the Tevatron and LHC remain almost unchanged when taking very disparate initial scales μRinit equal to mt, 10mt, 20mt and s, which is consistent with renormalization group invariance. As an important new application, we apply PMC scale setting to study the top quark forward-backward asymmetry. We observe that the more convergent perturbative series after PMC scale-setting leads to a more accurate top quark forward-backward asymmetry. The resulting PMC prediction on the asymmetry is also free from the initial renormalization scale dependence. Because the next-to-leading-order PMC scale has a dip behavior for the (qq¯) channel at small subprocess collision energies, the importance of this channel to the asymmetry is increased. We observe that the asymmetries AFBtt¯ and AFBpp¯ at the Tevatron will be increased by 42% in comparison to the previous estimates obtained by using conventional scale setting; i.e., we obtain AFBtt¯,PMC≃12.5% and AFBpp¯,PMC≃8.28%. Moreover, we obtain AFBtt¯,PMC(Mtt¯>450 GeV)≃35.0%. These predictions have a 1σ deviation from the present CDF and D0 measurements; the large discrepancies of the top quark forward-backward asymmetry between the standard model estimate and the CDF and D0 data are thus greatly reduced.

关键词： QUARKS forward-backward algorithm SYMMETRY (Physics) CROSS sections (Nuclear physics) PERTURBATION (Quantum dynamics) DEPENDENCE (Statistics) PREDICTION (Logic) STANDARD model (Nuclear physics)

来源：评论

学校读者我要写书评

暂无评论

A Multi-Parameter Optimization Approach for Complex Continuous Sparse Modelling 19

A Multi-Parameter Optimization Approach for Complex Continuo...

引用

19th International Conference on Digital Signal Processing (DSP)

作者： Chouzenoux, Emilie Pesquet, Jean-Christophe Florescu, Anisia Univ Paris Est LIGM CNRS UMR 8049 Champs Sur Marne France Univ Galatzi Elect & Telecommun Dept Galati Romania

ISBN: (纸本)9781479946129

The main focus of this work is the estimation of a complex valued signal assumed to have a sparse representation in an uncountable dictionary of signals. The dictionary elements are parameterized by a real-valued vector and the available observations are corrupted with an additive noise. By applying a linearization technique, the original model is recast as a constrained sparse perturbed model. The problem of the computation of the involved multiple parameters is addressed from a nonconvex optimization viewpoint. A cost function is defined including an arbitrary Lipschitz differentiable data fidelity term accounting for the noise statistics, and an l(0)-like penalty. A proximal algorithm is then employed to solve the resulting nonconvex and nonsmooth minimization problem. Experimental results illustrate the good practical performance of the proposed approach when applied to 2D spectrum analysis.

关键词： sparse modelling continuous compressive sensing proximity operator hard thresholding multivariate estimation 2D spectrum estimation forward-backward algorithm nonconvex optimization

来源：评论

学校读者我要写书评

暂无评论

Electroweak contribution to the top quark forward-backward asymmetry at the Tevatron

引用

Physical Review D 2011年第9期84卷 093003-093003页

作者： Wolfgang Hollik Davide Pagani Max-Planck-Institut für Physik (Werner-Heisenberg-Institut) Föhringer Ring 6 80805 München Germany

The electroweak contributions to the forward-backward asymmetry in the production of top-quark pairs at the Tevatron are evaluated at O(α2) and O(ααs2). We perform a detailed analysis of all partonic channels that produce an asymmetry and combine them with the QCD contributions. They provide a non-negligible addition to the QCD-induced asymmetry with the same overall sign, thus enlarging the standard model prediction and diminishing the observed deviation. For the observed mass-dependent forward-backward asymmetry a 3σ deviation still remains at an invariant-mass cut of Mtt¯>450 GeV.

关键词： ELECTROWEAK interactions forward-backward algorithm QUARKS STANDARD model (Nuclear physics) PARTICLES (Nuclear physics)

来源：评论

学校读者我要写书评

暂无评论

Enhance the Voltage Profile and Minimize Power Loss of Radial Power System 10

Enhance the Voltage Profile and Minimize Power Loss of Radia...

引用

10th IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering, UPCON 2023

作者： Bhengra, Ivan Carol Rai, J.N. G. B. Pant Dseu Okhla-III Campus Delhi Skill and Entrepreneurship University New Delhi India Delhi Technological University Department of Electrical Engineering New Delhi India

ISBN: (纸本)9798350382471

In order to analyze the power flow in radial distribution systems, this paper suggests a backward/forward sweep method. High R/X ratios and radial structure characterize the distribution system as a result, the distribution system causes the Newton-Raphson and fast-decoupled methods to fail. One of the best methods for the load-flow analysis of the radial distribution system, the proposed method presents a load-flow study using the backward/forward sweep method. This method can be used to calculate voltage magnitudes at each bus node as well as power losses for each bus branch. The IEEE 33-bus Radial Distribution System has been used to test this method, and MATLAB has been used to produce efficient results. A small power plant known as distributed generation (DG) is attached to a distribution system load in order to enhance the voltage profile, voltage regulation, stability, lower power losses, and increase economic return. The DG can be placed in the best possible way to get the aforementioned benefits. The ideal DG placement and size are determined using a forward-backward algorithm. The algorithm has been put to the test on 33 bus test systems, and the outcomes have been contrasted with those of another evolutionary algorithm. The suggested algorithm with the DG unit operational produced the best results. © 2023 IEEE.

关键词： Breadth-first search (BFS) Distributed generation (DG) forward-backward algorithm IEEE 33 bus system IEEE 69 bus system

来源：评论

学校读者我要写书评

暂无评论

On-line handwritten signature verification using hidden semi-Markov model

On-line handwritten signature verification using hidden semi...

引用

14th International Conference on Human-Computer Interaction, HCI International 2011

作者： Lin, Daw-Tung Liao, Yu-Chia Department of Computer Science and Information Engineering National Taipei University 151 University Rd. San-Shia New Taipei City 237 Taiwan

ISBN: (纸本)9783642220975

Handwritten signature has been extensively adopted as biometric for identity verification in daily life, as it is the most widely accepted personal authentication method. Automatic signature recognition technologies can definitely facilitate the verification process. Many research attempts and advances have occurred in this field, automatic signature verification still is a challenging and important issue. This work presents a novel and robust on-line signature verification approach using Hidden Semi-Markov Model (HSMM). The proposed system comprises three stages. First, dynamic features are extracted according to the local statistical information of velocity, acceleration, azimuth, altitude, and pressure. Next, the extracted features are normalized into unified observation length. To improve the verification accuracy, features with slight variation are clustered into the same class using K-means classification algorithm. Furthermore, the forward-backward algorithm is utilized to accelerate the computation of HSMM parameters. Finally, the system builds a unique HSMM for each identity and estimates the signature baseline in corresponding to the features. To assess the recognition performance of the proposed algorithm, experiments were conducted using SVC2004 signature database. Analytical results reveal that the proposed method is very promising. © 2011 Springer-Verlag.

关键词： dynamic features forward-backward algorithm Hidden Semi-Markov Model On-line handwritten signature verification

来源：评论

学校读者我要写书评

暂无评论

OPTIMIZED PREDICTION MODEL OF DIABETIC PATIENTS WITH HIDDEN MARKOV MODELS

OPTIMIZED PREDICTION MODEL OF DIABETIC PATIENTS WITH HIDDEN ...

引用

作者： Vaibhav Gautam Delhi Technological University

学位级别：硕士

The extraordinary advances in biotechnology and wellbeing sciences have prompted a unique creation of information, for example, high throughput genetic information and clinical data, produced from expansive Electronic Health Records (EHRs). To this end, utilization of ma- chine learning and information mining techniques in biosciences are direct, like never before previously, essential and critical in endeavors to change brilliantly all accessible data into valu- able information. In this research work, we have used an all-female dataset [19] which comprises eight attributes, i. e., Pregnancies, Glucose, Blood Pressure, Skin Thickness, Insulin, BMI, Diabetes Pedigree Function, Age and two class labels, i. e., 1 or 0 which indicates if a person has diabetes or not. Firstly, we used supervised techniques on the given dataset and tested their performance, i. e., accuracy in the prediction of people having diabetes. The results for various supervised al- gorithms are: K-NN - 64.91%, SVM - 65.91%, Random Forest - 67.80%, Gaussian Naive Bayesian - 64.36%, Decision Tree - 63.86%, Logistic Regression - 60.71%. From these results, we conclude that these algorithms cannot predict with sufficient accuracy which stands to be at least 80%. The unique feature about the supervised technique is if these algorithms encounter a specific attribute which dominates our prediction models, these algorithms always comes out to be on top. The result above means that our dataset has many features essential for prediction. As a result, the supervised Hidden Markov Model as an alternate technique for prediction. In our prediction methodology, with the help of the Hidden Markov Model, we use Viterbi and BestFirst Search algorithms as decoding techniques. These decoding techniques come up with following subsequences after the forward-backward algorithm gives emission probabil- ities. These probabilities belong to different states of our data model. As a result of which learning by our model gets

关键词： Artificial Intelligence Machine Learning Diabetes Viterbi Best-First Search Hidden Markov Models forward-backward algorithm

来源：评论

学校读者我要写书评

暂无评论

Metody pořízení a zpracování obrazů založené na řídkých reprezentacích

Metody pořízení a zpracování obrazů založené na ří...

引用

作者： Talár, Ondřej Brno University of Technology

Práce se zabývá rekonstrukčními možnostmi, které poskytuje řídká reprezentace signálů. Tato reprezentace zmenšuje signál na pouhý vektor prvků, které ukazují... 详细信息

Práce se zabývá rekonstrukčními možnostmi, které poskytuje řídká reprezentace signálů. Tato reprezentace zmenšuje signál na pouhý vektor prvků, které ukazují na části signálu ve slovníkové matici. Je nastíněna problematika s kvantovaným signálem a připomenuty druhy modulací, při kterých dochází ke kvantizaci a její způsoby. K řešení je zvolen algoritmus Douglas-Rachford, který nám umožňuje přibližovat se postupně k množině všech přijatelných řešení. Na konci je předvedeno řešení problému a několik testů pro ukázku funkce vytvořeného programu.

关键词： Řídké reprezentace dekvantizace odšumování dopředně-zpětný algoritmus Douglas-Rachford algoritmus rekonstrukce signálu Sparse representation dequantization denoising forward-backward algorithm Douglas-Rachford algorithm signal reconstruction T

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：