A model for compressing probabilities in belief networks

Probabilistic reasoning with belief (Bayesian) networks is based on conditional probability matrices. Thus it suffers from NP-hard implementations. In particular, the amount of probabilistic information necessary for the computations is often overwhelming. So, compressing the conditional probability table is one of the most important issues faced by the probabilistic reasoning community. Santos suggested an approach (called linear potential functions) for compressing the information from a combinatorial amount to roughly linear in the number of random variable assignments. However, much of the information in Bayesian networks, in which there are no linear potential functions, would be fitted by polynomial approximating functions rather than by reluctantly linear functions. For this reason, we construct a polynomial method to compress the conditional probability table in this paper. We evaluated the proposed technique, and our experimental results demonstrate that the approach is efficient and promising.

Evaluation of Bayesian belief networks for diagnosis of psychotic disorders

The thesis examined the inter-rater reliability and procedural validity of four computerised Bayesian belief networks (BBNs) which were developed to assist with the diagnosis of psychotic disorders. The results of this research indicated that BBNs can significantly improve diagnostic reliability and may represent an important advance over current diagnostic methods. The professional portfolio investigated, through the presentation of case studies and review of literature relevant to each case study, how comorbidity and context of depression may impact on cognitive behavioural therapy treatment.

Incorporating fairness into development of an integrated multi-agent online dispute resolution environment

The paper describes the development of an integrated multi-agent online dispute resolution environment called IMODRE that was designed to assist parties involved in Australian family law disputes achieve legally fairer negotiated outcomes. The system extends our previous work in developing negotiation support systems Family_Winner and AssetDivider. In this environment one agent uses a Bayesian Belief Network expertly modeled with knowledge of the Australian Family Law domain to advise disputants of their Best Alternatives to Negotiated Agreements. Another agent incorporates the percentage split of marital property into an integrative bargaining process and applies heuristics and game theory to equitably distribute marital property assets and facilitate further trade-offs. We use this system to add greater fairness to Family property law negotiations.

High-dimensional and large-scale anomaly detection using a linear one-class SVM with deep learning

High-dimensional problem domains pose significant challenges for anomaly detection. The presence of irrelevant features can conceal the presence of anomalies. This problem, known as the '. curse of dimensionality', is an obstacle for many anomaly detection techniques. Building a robust anomaly detection model for use in high-dimensional spaces requires the combination of an unsupervised feature extractor and an anomaly detector. While one-class support vector machines are effective at producing decision surfaces from well-behaved feature vectors, they can be inefficient at modelling the variation in large, high-dimensional datasets. Architectures such as deep belief networks (DBNs) are a promising technique for learning robust features. We present a hybrid model where an unsupervised DBN is trained to extract generic underlying features, and a one-class SVM is trained from the features learned by the DBN. Since a linear kernel can be substituted for nonlinear ones in our hybrid model without loss of accuracy, our model is scalable and computationally efficient. The experimental results show that our proposed model yields comparable anomaly detection performance with a deep autoencoder, while reducing its training and testing time by a factor of 3 and 1000, respectively.

Traffic matrix prediction and estimation based on deep learning in large-scale IP backbone networks

Network traffic analysis has been one of the most crucial techniques for preserving a large-scale IP backbone network. Despite its importance, large-scale network traffic monitoring techniques suffer from some technical and mercantile issues to obtain precise network traffic data. Though the network traffic estimation method has been the most prevalent technique for acquiring network traffic, it still has a great number of problems that need solving. With the development of the scale of our networks, the level of the ill-posed property of the network traffic estimation problem is more deteriorated. Besides, the statistical features of network traffic have changed greatly in terms of current network architectures and applications. Motivated by that, in this paper, we propose a network traffic prediction and estimation method respectively. We first use a deep learning architecture to explore the dynamic properties of network traffic, and then propose a novel network traffic prediction approach based on a deep belief network. We further propose a network traffic estimation method utilizing the deep belief network via link counts and routing information. We validate the effectiveness of our methodologies by real data sets from the Abilene and GÉANT backbone networks.