An approach to the sensitivity analysis of local a posteriori inference equations in algebraic Bayesian networks is proposed in this paper. Some basic definitions and formulations are briefly given and the development...
详细信息
An approach to the sensitivity analysis of local a posteriori inference equations in algebraic Bayesian networks is proposed in this paper. Some basic definitions and formulations are briefly given and the development of the matrix-vector a posteriori inference approach is considered. Some cases of the propagation of deterministic and stochastic evidence in a knowledge pattern with scalar estimates of component truth probabilities over quantum propositions are described. For each of the considered cases, the necessary metrics are introduced, and some transformations resulting in four linear programming problems are performed. The solution of these problems gives the required estimates. In addition, two theorems postulating the covering estimates for the considered parameters are formulated. The results obtained in this work prove the correct application of models and create a basis for the sensitivity analysis of local and global probabilistic-logic inference equations.
Credal networks lift the precise probability assumption of Bayesian networks, enabling a richer representation of uncertainty in the form of closed convex sets of probability measures. The increase in expressiveness c...
详细信息
ISBN:
(纸本)9783902652409
Credal networks lift the precise probability assumption of Bayesian networks, enabling a richer representation of uncertainty in the form of closed convex sets of probability measures. The increase in expressiveness comes at the expense of higher computational costs. In this paper we present a new algorithm which is an extension of the well-known variable elimination algorithm for computing posterior inferences in extensively specified credal networks. The algorithm efficiency is empirically shown to outperform a state-of-the-art algorithm. We then provide the first fully polynomial time approximation scheme for inference in credal networks with bounded treewidth and number of states per variable.
Reliability analysis is an integral part of system design and operating. Moreover, it can be an input to optimize maintenance policies. Recently, Bayesian Networks (BN) and Dynamic Bayesian Networks (DBN) have been pr...
详细信息
ISBN:
(纸本)9781424441358
Reliability analysis is an integral part of system design and operating. Moreover, it can be an input to optimize maintenance policies. Recently, Bayesian Networks (BN) and Dynamic Bayesian Networks (DBN) have been proved relevant to represent complex systems and perform reliability studies. The major drawback of this approach comes from the constraint on the state sojourn times which are necessarily exponentially distributed, as in usual markovian approaches. Therefore, a new formalism was previously introduced in the literature to avoid this constraint: the graphical Duration models (GDM). This paper aims to quantify the reliability estimation error due to an exponential approximation when the system follows other kinds of sojourn time's distributions. Finally results obtained by DBN and GDM will be compared.
When acquiring labels from crowdsourcing platforms, a task may be designed to include multiple labels and the values of each label may belong to a set of various distinct options, which is the so-called multi-class mu...
详细信息
Image segmentation of very large and complex microscopy images are challenging due to variability in the images and the need for algorithms to be robust, fast and able to incorporate various types of information and c...
详细信息
ISBN:
(纸本)9781457718588
Image segmentation of very large and complex microscopy images are challenging due to variability in the images and the need for algorithms to be robust, fast and able to incorporate various types of information and constraints in the segmentation model. In this paper we propose a graphical model based image segmentation framework that combines the information in images regions with the information in their boundary in a unified probabilistic formulation.
probabilistic Circuits (PCs) are prominent tractable probabilisticmodels, allowing for a wide range of exact inferences. This paper focuses on the main algorithm for training PCs, LearnSPN, a gold standard due to its...
详细信息
probabilistic Circuits (PCs) are prominent tractable probabilisticmodels, allowing for a wide range of exact inferences. This paper focuses on the main algorithm for training PCs, LearnSPN, a gold standard due to its efficiency, performance, and ease of use, in particular for tabular data. We show that LearnSPN is a greedy likelihood maximizer under mild assumptions. While inferences in PCs may use the entire circuit structure for processing queries, LearnSPN applies a hard method for learning them, propagating at each sum node a data point through one and only one of the children/edges as in a hard clustering process. We propose a new learning procedure named SoftLearn, that induces a PC using a soft clustering process. We investigate the effect of this learning-inference compatibility in PCs. Our experiments show that SoftLearn outperforms LearnSPN in many situations, yielding better likelihoods and arguably better samples. We also analyze comparable tractable models to highlight the differences between soft/hard learning and model querying.
All data and information are not always available at the time of a system design and implementation. Especially in knowledge-based systems, training data could be limited at the early stage and more training data migh...
详细信息
ISBN:
(纸本)9781479903863
All data and information are not always available at the time of a system design and implementation. Especially in knowledge-based systems, training data could be limited at the early stage and more training data might be acquired after the system deployment. This paper is concerned with a method to keep track of knowledge evolution and to detect the changes in the knowledge as more training data are provided. The method assumes that the knowledge is expressed in Bayesian networks and makes use of an agent framework for autonomous processing of knowledge evolution and change detection. It maintains sufficient statistics using a tiled sliding window structure. In order to flexibly encode the strategy for detecting the changes in the joint probability distributions, a set of fuzzy rules are used with which application domains specify their own strategy.
Discriminating probabilistic graphical models are reliable tools for a sequence labeling task. Conditional Random Fields (CRFs) are discriminative models which will enable us to label a sequence of input data. Other v...
详细信息
ISBN:
(纸本)9781509058204
Discriminating probabilistic graphical models are reliable tools for a sequence labeling task. Conditional Random Fields (CRFs) are discriminative models which will enable us to label a sequence of input data. Other variations of CRFs have been proposed. Hidden Conditional Random Fields (HCRFs) incorporate hidden states to the CRF model and assign a label for the whole input sequence as the model's output. Latent-Dynamic Conditional Random Fields (LDCRFs) also incorporate hidden variable states to the CRFs, in addition, these models are able to label each output variables separately. These models can capture subtle changes among different classes which will help us to achieve better recognition results. In this work we experiment various models and settings in order to achieve better results in facial expression recognition from sequence of videos. We use CRF and LDCRF models and train them with Limited-memory BFGS and Conjugate Gradient parameter learning methods. For each model we use various feature vectors in order to achieve better recognition results. We use Active Appearance Model (AAM) landmark points, Histogram of Oriented Gradients (HOG) and Uniform Local Binary Pattern (U-LBP) as our feature vectors in our models. We show which combination of learning methods and feature vectors are suitable for CRF and LDCRF discriminative models.
Sparse graph recovery methods work well where the data follows their assumptions, however, they are not always designed for doing downstream probabilistic queries. This limits their adoption to only identifying connec...
详细信息
ISBN:
(纸本)9783031476785;9783031476792
Sparse graph recovery methods work well where the data follows their assumptions, however, they are not always designed for doing downstream probabilistic queries. This limits their adoption to only identifying connections among domain variables. On the other hand, probabilistic graphical models (PGMs) learn an underlying base graph together with a distribution over the variables (nodes). PGM design choices are carefully made such that the inference and sampling algorithms are efficient. This results in certain restrictions and simplifying assumptions. In this work, we propose Neural Graph Revealers (NGRs) which attempt to efficiently merge the sparse graph recovery methods with PGMs into a single flow. The task is to recover a sparse graph showing connections between the features and learn a probability distribution over them at the same time. NGRs use a neural network as a multi-task learning framework. We introduce graph-constrained path norm that NGRs leverage to learn a graphical model that captures complex non-linear functional dependencies between features in the form of an undirected sparse graph. NGRs can handle multimodal inputs like images, text, categorical data, embeddings etc. which are not straightforward to incorporate in the existing methods. We show experimental results on data from Gaussian graphicalmodels and a multimodal infant mortality dataset by CDC (Software: https://***/harshs27/neural-graph-revealers).
This paper proposes an approach to learn robust behavior representations in online platforms by addressing the challenges of user behavior skew and sparse participation. Latent behavior models are important in a wide ...
详细信息
ISBN:
(纸本)9781450360142
This paper proposes an approach to learn robust behavior representations in online platforms by addressing the challenges of user behavior skew and sparse participation. Latent behavior models are important in a wide variety of applications: recommender systems;prediction;user profiling;community characterization. Our framework is the first to jointly address skew and sparsity across graphical behavior models. We propose a generalizable bayesian approach to partition users in the presence of skew while simultaneously learning latent behavior profiles over these partitions to address user-level sparsity. Our behavior profiles incorporate the temporal activity and links between participants, although the proposed framework is flexible to introduce other definitions of participant behavior. Our approach explicitly discounts frequent behaviors and learns variable size partitions capturing diverse behavior trends. The partitioning approach is data-driven with no rigid assumptions, adapting to varying degrees of skew and sparsity. A qualitative analysis indicates our ability to discover niche and informative user groups on large online platforms. Results on User Characterization (+6-22% AUC);Content Recommendation (+6-43% AUC) and Future Activity Prediction (+12-25% RMSE) indicate significant gains over state-of-the-art baselines. Furthermore, user cluster quality is validated with magnified gains in the characterization of users with sparse activity.
暂无评论