A NON-PARAMETER BAYESIAN CLASSIFIER FOR FACE RECOGNITION

A non-parameter Bayesian classifier based on Kernel Density Estimation (KDE)is presented for face recognition, which can be regarded as a weighted Nearest Neighbor (NN)classifier in formation. The class conditional density is estimated by KDE and the bandwidthof the kernel function is estimated by Expectation Maximum (EM) algorithm. Two subspaceanalysis methods-linear Principal Component Analysis (PCA) and Kernel-based PCA (KPCA)are respectively used to extract features, and the proposed method is compared with ProbabilisticReasoning Models (PRM), Nearest Center (NC) and NN classifiers which are widely used in facerecognition systems. The experiments are performed on two benchmarks and the experimentalresults show that the KDE outperforms PRM, NC and NN classifiers.

Hybrid Chinese Information Retrieval Model Based on the Combination of Keyword and Concept

A hybrid model that is based on the Combination of keywords and concept was put forward. The hybrid model is built on vector space model and probabilistic reasoning network. It not only can exert the advantages of keywords retrieval and concept retrieval but also can compensate for their shortcomings. Their parameters can be adjusted according to different usage in order to accept the best information retrieval result, and it has been proved by our experiments.

Calculate Joint Probability Distribution of Steady Directed Cyclic Graph with Local Data and Domain Casual Knowledge

It is desired to obtain the joint probability distribution(JPD) over a set of random variables with local data, so as to avoid the hard work to collect statistical data in the scale of all variables. A lot of work has been done when all variables are in a known directed acyclic graph(DAG). However, steady directed cyclic graphs(DCGs) may be involved when we simply combine modules containing local data together, where a module is composed of a child variable and its parent variables. So far, the physical and statistical meaning of steady DCGs remain unclear and unsolved. This paper illustrates the physical and statistical meaning of steady DCGs, and presents a method to calculate the JPD with local data, given that all variables are in a known single-valued Dynamic Uncertain Causality Graph(S-DUCG), and thus defines a new Bayesian Network with steady DCGs. The so-called single-valued means that only the causes of the true state of a variable are specified, while the false state is the complement of the true state.

Experts' Knowledge Fusion in Model-Based Diagnosis Based on Bayes Networks

In previous researches on a model-based diagnostic system, the components are assumed mutually independent. Howerver , the assumption is not always the case because the information about whether a component is faulty or not usually influences our knowledge about other components. Some experts may draw such a conclusion that 'if component m 1 is faulty, then component m 2 may be faulty too'. How can we use this experts' knowledge to aid the diagnosis? Based on Kohlas's probabilistic assumption-based reasoning method, we use Bayes networks to solve this problem. We calculate the posterior fault probability of the components in the observation state. The result is reasonable and reflects the effectiveness of the experts' knowledge.

Group Similarity and Social Influence Analysis in Online Communities

A fundamental open question in the analysis of social networks was to understand the evolution between similarity and group social ties.In general,two groups are similar for two distinct reasons:first,they grow to change their behaviors to the same group due to social influence;second,they tend to merge a group due to similar behaviors,where a process often is termed selection by sociologists.It was important to understand why two groups could merge and what led to high similarities for members in a group,influence or selection.In this paper,the techniques for identifying and modeling interactions between social influence and selection for different groups were developed.Different similarities were computed in three phases where groups came into being,before or after according to the number of common edits in Wikipedia.Experimental results showed selection played a more important role in two group merging.

Dynamic Uncertain Causality Graph for Knowledge Representation and Reasoning: Discrete DAG Cases

Developed from the dynamic causality diagram （DCD） model, a new approach for knowledge representation and reasoning named as dynamic uncertain causality graph （DUCG） is presented, which focuses on the compact representation of complex uncertain causalities and efficient probabilistie inference. It is pointed out that the existing models of compact representation and inference in Bayesian Network （BN） is applicable in single-valued cases, but may not be suitable to be applied in multi-valued cases. DUCG overcomes this problem and beyond. The main features of DUCG are： 1） compactly and graphically representing complex conditional probability distributions （CPDs）, regardless of whether the cases are single-valued or multi-valued; 2） able to perform exact reasoning in the case of the incomplete knowledge representation; 3） simplifying the graphical knowledge base conditional on observations before other calculations, so that the scale and complexity of problem can be reduced exponentially; 4） the efficient two-step inference algorithm consisting of （a） logic operation to find all possible hypotheses in concern for given observations and （b） the probability calculation for these hypotheses; and 5） much less relying on the parameter accuracy. An alarm system example is provided to illustrate the DUCG methodology.

BIC-based node order learning for improving Bayesian network structure learning

Node order is one of the most important factors in learning the structure of a Bayesian network(BN)for probabilistic reasoning.To improve the BN structure learning,we propose a node order learning algorithmbased on the frequently used Bayesian information criterion(BIC)score function.The algorithm dramatically reduces the space of node order and makes the results of BN learning more stable and effective.Specifically,we first find the most dependent node for each individual node,prove analytically that the dependencies are undirected,and then construct undirected subgraphs UG.Secondly,the UG-is examined and connected into a single undirected graph UGC.The relation between the subgraph number and the node number is analyzed.Thirdly,we provide the rules of orienting directions for all edges in UGC,which converts it into a directed acyclic graph(DAG).Further,we rank the DAG’s topology order and describe the BIC-based node order learning algorithm.Its complexity analysis shows that the algorithm can be conducted in linear time with respect to the number of samples,and in polynomial time with respect to the number of variables.Finally,experimental results demonstrate significant performance improvement by comparing with other methods.

Model Failure and Context Switching Using Logic-Based Stochastic Models

This paper addresses parameter drift in stochastic models. We define a notion of context that represents invariant, stable-over-time behavior and we then propose an algorithm for detecting context changes in processing a stream of data. A context change is seen as model failure, when a probabilistic model representing current behavior is no longer able to ＂fit＂ newly encountered data. We specify our stochastic models using a first-order logic-based probabilistic modeling language called Generalized Loopy Logic （GLL）. An important component of GLL is its learning mechanism that can identify context drift. We demonstrate how our algorithm can be incorporated into a failure-driven context-switching probabilistic modeling framework and offer several examples of its application.