Structure Learning Algorithm Research Articles

Locality-Adaptive Structured Dictionary Learning for Cross-Domain Recognition

Dictionary learning has achieved remarkable success on a wide range of machine learning-based applications. In this paper, a locality-adaptive structured dictionary learning (LASDL) algorithm for cross-domain recognition is proposed. In the LASDL, a projective structured double reconstruction strategy is developed to train class-oriented sub-dictionaries from the specific classes of cross-domain samples. The strategy benefits to make full advantage of the discriminative information of cross-domain data and bridge the distribution divergence between two different domains. Meanwhile, an adaptive geometrical structure preserving function is designed to not only preserve the local manifold structures spanned by the representation coefficient spaces of the source and target domains, but also impose a constraint that the coefficients should keep closer to their class centers, which is propitious to reduce the distribution divergence and make the representation more accurate. With the structured linear coding technique, the final cross-domain recognition can be efficiently performed by determining class-specific reconstruction error. The optimization of the proposed LASDL model can be efficiently solved by simple least square method and alternating direction method of multipliers (ADMM) algorithm. Extensive experimental results validate the superiority of the proposed algorithm in contrast to other state-of-the-art predecessors.

161P Bayesian network structure for predicting two-year survival inpatients diagnosed with non-small cell lung cancer

The incidence of lung cancer has been increasing. Healthcare providers are trying to acquire more knowledge of the disease’s biology to treat their patients better. However, the information available is more than humans can comprehend. Predictive models such as Bayesian networks, which can intricately represent causal relations between variables, are suitable structures to model this information. However, the conventional methods for developing Bayesian network structures are limited in healthcare. 545 NSCLC patients treated primarily with (chemo)radiation therapy from Maastro clinic in the Netherlands between 2010 and 2013 were considered to develop this Bayesian network structure. All continuous variables were binned before learning the structure process. Patients with missing survival status and variables with more than 25% missing information were excluded from the structure-building process. The causal relationship (arcs) between variables in the data was determined using the hill-climbing algorithm with domain experts’ restrictions over bootstrap runs (B=400). The final structure consists of nodes and arcs, where arcs (relationships) are present in at least 70% of all bootstrap samples. Structural performance was assessed by computing the area under the curve (AUC) values and producing calibration plots based on a repeated 5-fold cross-validation. In addition, an adapted pre-specified expert structure was compared with a structure developed from the method in this study. Tumor load was removed from the final Bayesian network structure due to its high percentage (37%) of missing information. The final cohort used to develop the structure was reduced to 499, excluding 46 (08.4%) patients with missing survival status. The resulting structure’s mean AUC and confidence interval to predict two-year survival was 0.614 (0.499 - 0.730). The performance of the compared structures was only slightly above the chance level, but the structure based on the method in this study was clinically more correct. The aim of this study was attained and shows that Bayesian network structures which combine expert knowledge with a rigorous structure-learning algorithm produce a clinically valid structure.

Partitioned hybrid learning of Bayesian network structures

We develop a novel hybrid method for Bayesian network structure learning called partitioned hybrid greedy search (pHGS), composed of three distinct yet compatible new algorithms: Partitioned PC (pPC) accelerates skeleton learning via a divide-and-conquer strategy, p-value adjacency thresholding (PATH) effectively accomplishes parameter tuning with a single execution, and hybrid greedy initialization (HGI) maximally utilizes constraint-based information to obtain a high-scoring and well-performing initial graph for greedy search. We establish structure learning consistency of our algorithms in the large-sample limit, and empirically validate our methods individually and collectively through extensive numerical comparisons. The combined merits of pPC and PATH achieve significant computational reductions compared to the PC algorithm without sacrificing the accuracy of estimated structures, and our generally applicable HGI strategy reliably improves the estimation structural accuracy of popular hybrid algorithms with negligible additional computational expense. Our empirical results demonstrate the competitive empirical performance of pHGS against many state-of-the-art structure learning algorithms.

PSL: An Algorithm for Partial Bayesian Network Structure Learning

Learning partial Bayesian network (BN) structure is an interesting and challenging problem. In this challenge, it is computationally expensive to use global BN structure learning algorithms, while only one part of a BN structure is interesting, local BN structure learning algorithms are not a favourable solution either due to the issue of false edge orientation. To address the problem, this article first presents a detailed analysis of the false edge orientation issue with local BN structure learning algorithms and then proposes PSL, an efficient and accurate P artial BN S tructure L earning (PSL) algorithm. Specifically, PSL divides V-structures in a Markov blanket (MB) into two types: Type-C V-structures and Type-NC V-structures, then it starts from the given node of interest and recursively finds both types of V-structures in the MB of the current node until all edges in the partial BN structure are oriented. To further improve the efficiency of PSL, the PSL-FS algorithm is designed by incorporating F eature S election (FS) into PSL. Extensive experiments with six benchmark BNs validate the efficiency and accuracy of the proposed algorithms.

The ecology of human behavior: A network perspective

There is a broad agreement that some of the most relevant problems in the social and behavioral sciences are fundamentally structural, and as a consequence require structural explanations. Yet researchers disagree on what a structural explanation is, and what are the specific questions that can only be answered through a structural lens. In this study, I shed some light on the nature of structural explanations by distinguishing between three types of structural questions related to structural proximity, structural cohesion and structural importance. In addition, I show how graphical methods can be used to answer these questions. In particular, I argue that structure learning algorithms can help us gain some understanding regarding causal structures, and network science can help us understand the organization of these structures. I provide an empirical application of these methods using a nationally representative dataset with a wide range of factors related to child development.

A novel discrete firefly algorithm for Bayesian network structure learning

As an effective tool for the representation and reasoning of uncertain theories, Bayesian networks are widely used in various fields of artificial intelligence. However, learning the optimal Bayesian network structure is an NP-hard problem. The score-and-search approach is a common method for solving the problems associated with Bayesian network structure learning. This paper presents a novel method for learning the structure of a Bayesian network using a discrete firefly optimization algorithm, which has been successfully applied to solve various optimization problems. In the proposed algorithm, each firefly moves in a discrete space based on a redefined movement strategy. Then, the mutation operator is employed to prevent the algorithm from stopping prematurely and falling into a local optimum. Finally, a local optimizer is used to enhance the exploitation ability of the firefly to obtain the best feasible solution. We compared the proposed algorithm with state-of-the-art algorithms on well-known benchmark networks. The experimental results show that the proposed algorithm has better convergence accuracy and higher scores in most cases, compared to other algorithms, indicating that the proposed algorithm can be used as an effective and feasible method for learning Bayesian network structures.

A data-driven Bayesian network model integrating physical knowledge for prioritization of risk influencing factors

The coupling of multiple factors stemming from propagation effects and interdependency relationships among risks is prone to generate major accidents. It is of necessity to develop a feasible model with limited cases, which can generate reliable causal relationship evolution. To prioritize risk-influencing factors (RIFs) and investigate their relationships, we proposed a data-driven Bayesian Network (BN) model integrating physical information for risk analysis. Based on collected data, we combined prior knowledge with structure learning and parameter learning to obtain a BN model. In structure learning, we compared three structure learning algorithms including Bayesian search (BS), greedy thick thinning (GTT), and PC algorithm to obtain a robust directed acyclic graph (DAG). In parameter learning, we selected the expectation maximization (EM) algorithm to quantify the dependence and determine the probability distribution of node variables. This study provides a method to capture crucial factors and their interdependent relationships. To illustrate the applicability of the model, we developed a data-driven BN by taking the blowout accident as the case study. Eventually, we introduced vulnerability and resilience metrics for prioritizing risks through network propagation to conduct emergency plans and mitigation strategies.

Analysis on Dynamic Evolution of the Cost Risk of Prefabricated Building Based on DBN

Prefabricated building constitutes the development trend of the construction industry in the future. However, many uncertainties in the construction process will surely lead to a higher cost. Therefore, it is necessary to study the cost risk evolution and transfer mechanism in the implementation process of this project. A dynamic evolution model for the cost risk of prefabricated buildings has been established in this paper. First of all, a matrix for cost risk of prefabricated buildings was established based on the WSR (Wuli-Shili-Renli) model, and all risk factors in the implementation stage were classified in accordance with the WSR principle. Second, a DBN-based dynamic evolution model was established based on the risk matrix, and the structure and node parameters of the Dynamic Bayesian Network were determined with the aid of the K2 structure learning algorithm and parameter learning method. In view of the probability change process of risks over time, the dynamic evolution path of risks was predicted in different cases through causal reasoning and diagnostic reasoning. Eventually, the model was applied into construction projects. The research results show that: because prefabricated components need to be made by prefabricated component factories, the management systems of prefabricated component factories are usually not perfect, and the probability of management risks is higher. The occurrence of management risks not only has an impact on other risks at the current time node, but also causes other risks to occur in the subsequent transportation and construction phases at the next moment, which eventually leads to the occurrence of risk events.

A Bayesian network structural learning algorithm for calculating the failure probabilities of complex engineering systems with limited data

The failure of complex engineering systems is easy to lead to disastrous consequences. To prevent the failure, it is necessary to model complex engineering systems using probabilistic techniques with limited data which is a major feature of complex engineering systems. It is a good choice to perform such modeling using Bayesian network because of its advantages in probabilistic modeling. However, few Bayesian network structural learning algorithms are designed for complex engineering systems with limited data. Therefore, an algorithm for learning the Bayesian network structure of them should be developed. Based on the process of self-purification of water, a complex engineering system is segmented into three components according to the degree of difficulty in solving them. And then a Bayesian network learning algorithm with three components (TC), including PC algorithm, MIK algorithm which is originated by the paper through combining Mutual Information and K2 algorithm, and the Hill-Climbing method, is developed, i.e. TC algorithm. To verify its effectiveness, TC algorithm, K2 algorithm, and Max-Min Hill-Climbing are respectively used to learn Alarm network with different sizes of samples. The results imply that TC algorithm has the best performance. Finally, TC algorithm is applied to study tank spill accidents with 220 samples.

A decomposition structure learning algorithm in Bayesian network based on a two-stage combination method

Decomposition hybrid algorithms with the recursive framework which recursively decompose the structural task into structural subtasks to reduce computational complexity are employed to learn Bayesian network (BN) structure. Merging rules are commonly adopted as the combination method in the combination step. The direction determination rule of merging rules has problems in using the idea of keeping v-structures unchanged before and after combination to determine directions of edges in the whole structure. It breaks down in one case due to appearances of wrong v-structures, and is hard to operate in practice. Therefore, we adopt a novel approach for direction determination and propose a two-stage combination method. In the first-stage combination method, we determine nodes, links of edges by merging rules and adopt the idea of permutation and combination to determine directions of contradictory edges. In the second-stage combination method, we restrict edges between nodes that do not satisfy the decomposition property and their parent nodes by determining the target domain according to the decomposition property. Simulation experiments on four networks show that the proposed algorithm can obtain BN structure with higher accuracy compared with other algorithms. Finally, the proposed algorithm is applied to the thickening process of gold hydrometallurgy to solve the practical problem.

Causal Structure Learning Algorithm Based on Partial Rank Correlation under Additive Noise Model

ABSTRACT Aiming at the structural learning problem of the additive noise model in causal discovery and the challenge of massive data processing in the era of artificial intelligence, this paper combines partial rank correlation coefficients and proposes two new Bayesian network causal structure learning algorithms: PRCB algorithm based on threshold selection and PRCS algorithm based on hypothesis testing. We mainly made three contributions. First, we proved that the partial rank correlation coefficient can be used as the standard of independence test, and explored the distribution of corresponding statistics. Second, the partial rank correlation coefficient is associated with the correlation, and a causal discovery algorithm PRCB based on partial rank correlation and an improved PRCS algorithm based on hypothesis testing are proposed. Finally, comparing with the existing technology on seven classic Bayesian networks, it proves the superiority of the algorithm in low-dimensional networks; the processing of millions of data on three high-dimensional Bayesian networks verifies the high-efficiency performance of the algorithm in high-dimensional large sample data; the application performance of the algorithm is tested by performing fault prediction on the real power plant equipment measurement point data set. Theoretical analysis and experimental results have proved the superiority of the algorithm.

Learning Bayesian Network for Rainfall Prediction Modeling in Urban Area using Remote Sensing Satellite Data (Case Study: Jakarta, Indonesia)

Rainfall modeling is one of the most critical factors in agricultural monitoring and statistics, transportation schedules, and urban flood prevention. Weather anomaly during the dry season in urban coastal areas of tropical countries such as Jakarta, Indonesia has become a challenging issue that causes unexpected changes in rain patterns. In this paper, we propose the Bayesian Network (BN) approach to model the probabilistic nature of rain patterns in urban areas and causal relationships among its predictor variables. Rain occurrences are predicted using temperature, relative humidity, mean-sea level (MSL) pressure, cloud cover, and precipitation variables. Data are obtained from the remote sensing sources of the National Oceanic and Atmospheric Administration (NOAA) satellite in Jakarta 2020-2021. We compare both of the score-based, i.e., Hill Climbing (HC), and hybrid structure learning algorithms of Bayesian Network including the techniques of Max-Min Hill Climbing (MMHC), General 2-Phase Restricted Maximization (RSMAX2), and Hybrid-Hybrid Parents & Children (H2PC). Further, we also compare the performance of score-based model (Hill Climbing) under five different popular scorings: Bayesian Information Criterion (BIC), K2, Log-Likelihood, Bayesian Dirichlet Equivalent (BDE), and Akaike Information Criterion (AIC) methods. The main contributions of this study are as follows: (1) insights that the hybrid structure learning algorithms of Bayesian Network models are either superior in performance or at least comparable to its score-based counterparts (2) our proposed best performed Bayesian Network model that is able to predict the rain occurrences in Jakarta with a promising overall accuracy of more than 81 percent.

Semi-Supervised Structured Subspace Learning for Multi-View Clustering.

Multi-view clustering aims at simultaneously obtaining a consensus underlying subspace across multiple views and conducting clustering on the learned consensus subspace, which has gained a variety of interest in image processing. In this paper, we propose the Semi-supervised Structured Subspace Learning algorithm for clustering data points from Multiple sources (SSSL-M). We explicitly extend the traditional multi-view clustering with a semi-supervised manner and then build an anti-block-diagonal indicator matrix with small amount of supervisory information to pursue the block-diagonal structure of the shared affinity matrix. SSSL-M regularizes multiple view-specific affinity matrices into a shared affinity matrix based on reconstruction through a unified framework consisting of backward encoding networks and the self-expressive mapping. The shared affinity matrix is comprehensive and can flexibly encode complementary information from multiple view-specific affinity matrices. An enhanced structural consistency of affinity matrices from different views can be achieved and the intrinsic relationships among affinity matrices from multiple views can be effectively reflected in this manner. Technically, we formulate the proposed model as an optimization problem, which can be solved by an alternating optimization scheme. Experimental results over seven different benchmark datasets demonstrate that better clustering results can be obtained by our method compared with the state-of-the-art approaches.

Meta Structural Learning Algorithm With Interpretable Convolutional Neural Networks for Arrhythmia Detection of Multisession ECG

Detection of arrhythmia of electrocardiogram (ECG) signals recorded within several sessions for each person is a challenging issue, which has not been properly investigated in the past. This arrhythmia detection is challenging since a classification model that is constructed and tested using ECG signals maintains generalization when dealing with unseen samples. This article has proposed a new interpretable meta structural learning algorithm for this challenging detection. Therefore, a compound loss function was suggested including the structural feature extraction fault and space label fault with GUMBEL-SOFTMAX distribution in the convolutional neural network (CNN) models. The collaboration between models was carried out to create learning to learn features in models by transferring the knowledge among them when confronted by unseen samples. One of the deficiencies of a meta-learning algorithm is the non-interpretability of its models. Therefore, to create an interpretability feature for CNN models, they are encoded as the evolutionary trees of the genetic programming (GP) algorithms in this article. These trees learn the process of extracting deep structural features in the course of the evolution in the GP algorithm. The experimental results suggested that the proposed detection model enjoys an accuracy of 98% regarding the classification of 7 types of arrhythmia in the samples of the Chapman ECG dataset recorded from 10646 patients in different sessions. Finally, the comparisons demonstrated the competitive performance of the proposed model concerning the other models based on the big deep models.

A Recursive Local Search Method of Separators for Bayesian Network Decomposition Structure Learning Algorithm

A Recursive Local Search Method of Separators for Bayesian Network Decomposition Structure Learning Algorithm

A Bayesian Network Structure Learning Algorithm Based on Probabilistic Incremental Analysis and Constraint

To address the problem of low efficiency of the existing hill-climbing algorithm in Bayesian network structure learning, this paper proposes a Bayesian network structure learning algorithm based on probabilistic incremental analysis and constraints. The algorithm constructs a suitable measure for representing the degree of node association in Bayesian networks based on the principle of random forest feature extraction; then uses the method to construct the initial Bayesian network structure and constrains the search space by setting a corresponding threshold for the probability increment centered on each node; finally takes the initial Bayesian network as the starting point and learns it by the forbidden hill-climbing search and BIC scoring method to obtain the optimal Bayesian network structure. Experimental results show that the correlation degree measure and mutual information proposed in this paper have an approximate correlation expression effect; compared with other Bayesian network structure learning algorithms of the same type, the method in this paper has a faster operation speed while ensuring the quality of the learned network. The experimental results show that the Bayesian network structure learning algorithm based on probabilistic incremental analysis and constraints is an effective and efficient Bayesian network structure learning algorithm.

Dynamic Programming Structure Learning Algorithm of Bayesian Network Integrating MWST and Improved MMPC

Dynamic programming is difficult to apply to large-scale Bayesian network structure learning. In view of this, this article proposes a BN structure learning algorithm based on dynamic programming, which integrates improved MMPC (maximum-minimum parents and children) and MWST (maximum weight spanning tree). First, we use the maximum weight spanning tree to obtain the maximum number of parent nodes of the network node. Second, the MMPC algorithm is improved by the symmetric relationship to reduce false-positive nodes and obtain the set of candidate parent-child nodes. Finally, with the maximum number of parent nodes and the set of candidate parent nodes as constraints, we prune the parent graph of dynamic programming to reduce the number of scoring calculations and the complexity of the algorithm. Experiments have proved that when an appropriate significance level α is selected, the MMPCDP algorithm can greatly reduce the number of scoring calculations and running time while ensuring its accuracy.

Modeling Semiarid River–Aquifer Systems with Bayesian Networks and Artificial Neural Networks

In semiarid areas, precipitations usually appear in the form of big and brief floods, which affect the aquifer through water infiltration, causing groundwater temperature changes. These changes may have an impact on the physical, chemical and biological processes of the aquifer and, thus, modeling the groundwater temperature variations associated with stormy precipitation episodes is essential, especially since this kind of precipitation is becoming increasingly frequent in semiarid regions. In this paper, we compare the predictive performance of two popular tools in statistics and machine learning, namely Bayesian networks (BNs) and artificial neural networks (ANNs), in modeling groundwater temperature variation associated with precipitation events. More specifically, we trained a total of 2145 ANNs with different node configurations, from one to five layers. On the other hand, we trained three different BNs using different structure learning algorithms. We conclude that, while both tools are equivalent in terms of accuracy for predicting groundwater temperature drops, the computational cost associated with the estimation of Bayesian networks is significantly lower, and the resulting BN models are more versatile and allow a more detailed analysis.

New Media Advertising Communication Analysis Model Based on Extension Neural Network

In order to improve the effect of new media advertising communication analysis, this paper combines the scalable neural network to construct the new media advertising communication analysis model. Moreover, this paper analyzes in detail the basic theories of fuzzy neural network and extension evaluation, the structure design and learning algorithm, and classification of fuzzy neural network. In particular, this paper summarizes the optimization algorithms and methods of neural network structure. In addition, this paper improves the algorithm to meet the needs of new media advertising data analysis and builds an intelligent system framework. The experimental verification shows that the new media advertising communication analysis model based on the extension neural network proposed in this paper meets the new media advertising communication analysis effect.

Efficient structural pseudoinverse learning-based hierarchical representation learning for skin lesion classification

The success of deep learning in skin lesion classification mainly depends on the ultra-deep neural network and the significantly large training data set. Deep learning training is usually time-consuming, and large datasets with labels are hard to obtain, especially skin lesion images. Although pre-training and data augmentation can alleviate these issues, there are still some problems: (1) the data domain is not consistent, resulting in the slow convergence; and (2) low robustness to confusing skin lesions. To solve these problems, we propose an efficient structural pseudoinverse learning-based hierarchical representation learning method. Preliminary feature extraction, shallow network feature extraction and deep learning feature extraction are carried out respectively before the classification of skin lesion images. Gabor filter and pre-trained deep convolutional neural network are used for preliminary feature extraction. The structural pseudoinverse learning (S-PIL) algorithm is used to extract the shallow features. Then, S-PIL preliminarily identifies the skin lesion images that are difficult to be classified to form a new training set for deep learning feature extraction. Through the hierarchical representation learning, we analyze the features of skin lesion images layer by layer to improve the final classification. Our method not only avoid the slow convergence caused by inconsistency of data domain but also enhances the training of confusing examples. Without using additional data, our approach outperforms existing methods in the ISIC 2017 and ISIC 2018 datasets.