Latent Tree Models and Approximate Inference in Bayesian Networks

We propose a novel method for approximate inference in Bayesian networks (BNs). The idea is to sample data from a BN, learn a latent tree model (LTM) from the data offline, and when online, make inference with the LTM instead of the original BN. Because LTMs are tree-structured, inference takes linear time. In the meantime, they can represent complex relationship among leaf nodes and hence the approximation accuracy is often good. Empirical evidence shows that our method can achieve good approximation accuracy at low online computational cost.

Download Full-text

A Survey on Latent Tree Models and Applications

Journal of Artificial Intelligence Research ◽

10.1613/jair.3879 ◽

2013 ◽

Vol 47 ◽

pp. 157-203 ◽

Cited By ~ 24

Author(s):

R. Mourad ◽

C. Sinoquet ◽

N. L. Zhang ◽

T. Liu ◽

P. Leray

Keyword(s):

Graphical Models ◽

Latent Variables ◽

Simple Structure ◽

Tree Model ◽

The Past ◽

Latent Tree Models ◽

Tree Models ◽

Complex Relationships ◽

Latent Tree Model ◽

Comprehensive Study

In data analysis, latent variables play a central role because they help provide powerful insights into a wide variety of phenomena, ranging from biological to human sciences. The latent tree model, a particular type of probabilistic graphical models, deserves attention. Its simple structure - a tree - allows simple and efficient inference, while its latent variables capture complex relationships. In the past decade, the latent tree model has been subject to significant theoretical and methodological developments. In this review, we propose a comprehensive study of this model. First we summarize key ideas underlying the model. Second we explain how it can be efficiently learned from data. Third we illustrate its use within three types of applications: latent structure discovery, multidimensional clustering, and probabilistic inference. Finally, we conclude and give promising directions for future researches in this field.

Download Full-text

Approximate inference in Bayesian networks using binary probability trees

International Journal of Approximate Reasoning ◽

10.1016/j.ijar.2010.05.006 ◽

2011 ◽

Vol 52 (1) ◽

pp. 49-62 ◽

Cited By ~ 9

Author(s):

Andrés Cano ◽

Manuel Gémez-Olmedo ◽

Serafén Moral

Keyword(s):

Bayesian Networks ◽

Approximate Inference ◽

Probability Trees

Download Full-text

Interactive Modeling of 3D Tree with Ball B-Spline Curves

International Journal of Virtual Reality ◽

10.20870/ijvr.2009.8.2.2731 ◽

2009 ◽

Vol 8 (2) ◽

pp. 101-107 ◽

Cited By ~ 2

Author(s):

Zhongke Wu ◽

Mingquan Zhou ◽

Xingce Wang

Keyword(s):

Iterated Function Systems ◽

Interactive Methods ◽

Tree Model ◽

B Spline ◽

Spline Curves ◽

Novel Approach ◽

Iterated Function ◽

Tree Models ◽

Generating Tree ◽

L Systems

A novel approach to modeling realistic tree easily through interactive methods based on ball B-Spline Curves (BBSCs) and an efficient graph based data structure of tree model is proposed in the paper. As BBSCs are flexible for modifying, deforming and editing, these methods provide intuitive interaction and more freedom for users to model trees. If conjuncted with other methods like generating tree models through L-systems or iterated function systems (IFS), the models are more realistic and natural through modifying and editing. The method can be applied to the design of bonsai tree models.

Download Full-text

A Robust Procedure to Implement Dynamic Stall Models Into Actuator Line Methods for the Simulation of Vertical-Axis Wind Turbines

Journal of Engineering for Gas Turbines and Power ◽

10.1115/1.4051909 ◽

2021 ◽

Author(s):

Pier Francesco Melani ◽

Francesco Balduzzi ◽

Alessandro Bianchini

Keyword(s):

Flow Field ◽

Wind Turbines ◽

Signal To Noise Ratio ◽

Computational Cost ◽

Vertical Axis ◽

Dynamic Stall ◽

Lumped Parameter ◽

Low Pass ◽

Unsteady Rans ◽

Novel Method

Abstract The Actuator Line Method (ALM), combining a lumped-parameter representation of the rotating blades with the CFD resolution of the turbine flow field, stands out among the modern simulation methods for wind turbines as probably the most interesting compromise between accuracy and computational cost. Being however a method relying on tabulated coefficients for modeling the blade-flow interaction, the correct implementation of the sub-models to account for higher order aerodynamic effects is pivotal. Inter alia, the introduction of a dynamic stall model is extremely challenging: first, it is important to extrapolate a correct value of the angle of attack (AoA) from the solved flow field; second, the AoA history needed to calculate the rate of dynamic variation of the angle itself is characterized by a low signal-to-noise ratio, leading to severe numerical oscillations of the solution. The study introduces a robust procedure to improve the quality of the AoA signal extracted from an ALM simulation. It combines a novel method for sampling the inflow velocity from the numerical flow field with a low-pass filtering of the corresponding AoA signal based on Cubic Spline Smoothing. Such procedure has been implemented in the Actuator Line module developed by the authors for the commercial ANSYS® FLUENT® solver. To verify the reliability of the methodology, two-dimensional unsteady RANS simulations of a test 2-blade Darrieus H-rotor, for which high-fidelity experimental and numerical blade loading data were available, have been performed for a selected unstable operation point.

Download Full-text

Feedback-Driven Refinement of Mandarin Speech Recognition Result Based on Lattice Modification and Rescoring

Cognitive Analytics ◽

10.4018/978-1-7998-2460-2.ch062 ◽

2020 ◽

pp. 1237-1247

Author(s):

Xiangdong Wang ◽

Yang Yang ◽

Hong Liu ◽

Yueliang Qian ◽

Duan Jia

Keyword(s):

Speech Recognition ◽

Real World ◽

Computational Cost ◽

User Feedback ◽

Recognition Result ◽

Novel Method ◽

Mandarin Speech Recognition ◽

Recognition Errors ◽

Dynamic Updating ◽

Manual Correction

In real world applications of speech recognition, recognition errors are inevitable, and manual correction is necessary. This paper presents an approach for the refinement of Mandarin speech recognition result by exploiting user feedback. An interface incorporating character-based candidate lists and feedback-driven updating of the candidate lists is introduced. For dynamic updating of candidate lists, a novel method based on lattice modification and rescoring is proposed. By adding words with similar pronunciations to the candidates next to the corrected character into the lattice and then performing rescoring on the modified lattice, the proposed method can improve the accuracy of the candidate lists even if the correct characters are not in the original lattice, with much lower computational cost than that of the speech re-recognition methods. Experimental results show that the proposed method can reduce 24.03% of user inputs and improve average candidate rank by 25.31%.

Download Full-text

Biquadratic Digital Phase-Compensator Design with Stability-Margin Controllability

Journal of Circuits System and Computers ◽

10.1142/s0218126619500683 ◽

2019 ◽

Vol 28 (04) ◽

pp. 1950068 ◽

Cited By ~ 2

Author(s):

Tian-Bo Deng

Keyword(s):

Stability Margin ◽

Transformation Method ◽

Second Order ◽

Approximation Accuracy ◽

Parameter Transformation ◽

Practical Applications ◽

Compensator Design ◽

Novel Method ◽

Design Idea ◽

The Stability

This paper proposes a novel method for the design of a recursive second-order (biquadratic) all-pass phase compensator with controllable stability margin. The design idea stems from the generalized stability triangle (GST) derived by the author for the second-order biquadratic digital filter. Based on the GST, a parameter-transformation method is proposed on the transformations of the denominator coefficients of the transfer function of the biquadratic phase compensator. The transformations convert the original denominator coefficients to other new parameters, and any values of those new parameters can guarantee that the GST condition is always satisfied. Optimizing the new parameters yields a biquadratic phase compensator that definitely meets a prespecified stability margin. That is, a biquadratic all-pass phase compensator can be designed to have an arbitrarily specified stability margin. This in turn avoids the occurrence that a recursive phase compensator may become unstable in the practical applications. Thus, the resulting biquadratic phase compensator has robust stability, which is extremely important during the practical filtering operations. A design example is given to show the stability margin guarantee as well as the approximation accuracy.

Download Full-text

Estonian traces in the tree of life concept and in the language family tree theory

Eesti ja soome-ugri keeleteaduse ajakiri Journal of Estonian and Finno-Ugric Linguistics ◽

10.12697/jeful.2012.3.1.15 ◽

2012 ◽

Vol 3 (1) ◽

pp. 297-326

Author(s):

Urmas Sutrop

Keyword(s):

Charles Darwin ◽

Tree Of Life ◽

Family Tree ◽

Language Family ◽

Animal Life ◽

Tree Model ◽

Tree Models

In this paper the tree model – a well-formed tree is shortly described. After that the language family tree model by August Schleicher is treated and compared with the Charles Darwin’s tree of life diagram and metaphor. The development of the idea of the linguistic trees and the tree of life is considered historically. Earlier models – scala naturae – and tree models, both well-formed and not-well-formed are introduced. Special attention is paid to the scholars connected to Estonia who developed the idea of tree models: Georg Stiernhielm was the first who pictured a language tree already in 1671; Karl Eduard Eichwald published an early tree of animal life in 1829; and Karl Ernst von Baer influenced the tree of life models and diagrams of Charles Darwin.

Download Full-text

Approximate inference in Bayesian networks: Parameterized complexity results

International Journal of Approximate Reasoning ◽

10.1016/j.ijar.2017.10.029 ◽

2018 ◽

Vol 93 ◽

pp. 119-131 ◽

Cited By ~ 6

Author(s):

Johan Kwisthout

Keyword(s):

Bayesian Networks ◽

Parameterized Complexity ◽

Approximate Inference ◽

Complexity Results

Download Full-text

Evaluation of Data Mining Methods

Encyclopedia of Data Warehousing and Mining ◽

10.4018/978-1-59140-557-3.ch088 ◽

2011 ◽

pp. 464-468

Author(s):

Paolo Giudici

Keyword(s):

Data Mining ◽

Linear Regression ◽

Bayesian Networks ◽

Statistical Methods ◽

Linear Regression Models ◽

Explanatory Variables ◽

Tree Models ◽

Number Of Leaves ◽

Mining Methods ◽

Evaluation Of Data

Several classes of computational and statistical methods for data mining are available. Each class can be parameterised so that models within the class differ in terms of such parameters (see, for instance, Giudici, 2003; Hastie et al., 2001; Han & Kamber, 2000; Hand et al., 2001; Witten & Frank, 1999): for example, the class of linear regression models, which differ in the number of explanatory variables; the class of Bayesian networks, which differ in the number of conditional dependencies (links in the graph); the class of tree models, which differ in the number of leaves; and the class multi-layer perceptrons, which differ in terms of the number of hidden strata and nodes. Once a class of models has been established the problem is to choose the “best” model from it.

Download Full-text