Search

A Bayesian network is a multivariate (potentially high dimensional) probabilistic model, which is often formed by combining lower dimensional components. Inference (computation of conditional probabilities) is based on message passing algorithms that utilize conditional independence structures. In Bayesian networks for discrete variables with finite state spaces, there is a fundamental problem in high dimensions: A discrete distribution is represented by a table of values, and in high dimensions such tables can become prohibitively large. In inference, such tables must be multiplied which can lead to even larger tables. The sparta package tries to meet this challenge by introducing new methods that efficiently handles multiplication and marginalization of sparse tables. The package was written in the R programming language and is freely available from the Comprehensive R Archive Network (CRAN). The companion package jti, also on CRAN, was developed to showcase the potential of sparta in connection to the Junction Tree Algorithm. We show that sparta outperforms existing methods for high-dimensional sparse probability tables. Furthermore, we demonstrate, that sparta can handle problems which are otherwise infeasible due to lack of computer memory. A graphical model is a multivariate (potentially very high dimensional) probabilistic model, which is formed by combining lower dimensional components. Inference (computation of conditional probabilities) is based on message passing algorithms that utilize conditional independence structures. In graphical models for discrete variables with finite state spaces, there is a fundamental problem in high dimensions: A discrete distribution is represented by a table of values, and in high dimensions such tables can become prohibitively large. In inference, such tables must be multiplied which can lead to even larger tables. The sparta package meets this challenge by implementing methods that efficiently handles multiplication and marginalization of sparse tables. The package was written in the R programming language and is freely available from the Comprehensive R Archive Network (CRAN). The companion package jti, also on CRAN, was developed to showcase the potential of sparta in connection to the Junction Tree Algorithm. We show, that jti is able to handle highly complex graphical models which are otherwise infeasible due to lack of computer memory, using sparta as a backend for table operations.

We propose an algorithm for estimating positions of devices in a sensor network using Kalman filtering techniques. The specific area of application is monitoring the movements of cows in a barn. The algorithm consists of two filters. The first filter enhances the signal-to-noise ratio of the observed signal strengths and gives interpolated values at specific timestamps. Information from the first filter is transferred to the second filter which estimates the positions. Methods for estimating the parameters of the filters are given and these provide a straightforward calibration of the system We propose an algorithm for estimating positions of devices in a sensor network using Kalman filtering techniques. The specific area of application is monitoring the movements of cows in a barn. The algorithm consists of two filters. The first filter enhances the signal-to-noise ratio of the observed signal strengths and gives interpolated values at specific timestamps. Information from the first filter is transferred to the second filter which estimates the positions. Methods for estimating the parameters of the filters are given and these provide a straightforward calibration of the system

This paper addresses detection of oestrus in dairy cows using methods from statistical change detection. The activity of the cows was measured by a necklace attached sensor. Statistical properties of the activity measure were investigated. Using data sets from 17 cows, diurnal activity variations were identified for the ensemble and for the individual cows. A diurnal filter was adapted to remove the daily variation of the individual. Change detection algorithms were designed for the actual probability densities, which were Rayleigh distributed with individual parameters for each cow. A generalized likelihood ratio algorithm was derived for the compensated activity signal and detection algorithm was tested on 2323 days of activity, which contained 42 oestruses on 12 cows in total. The application of statistical change detection methods is a new approach for detecting oestrus in dairy cows and the results are shown to outperform earlier approaches in respect to combined statistics of false alarms and missed detections.

Progesterone is a hormone linked to the reproductive status of dairy cows. Hence, with the increasing availability of on-line records of the concentration of progesterone in cow milk, there is a need for new tools to analyse such data. The aim is to find techniques for better determination of the time when cows are in oestrus to increase the rate of succesful inseminations. In this paper we propose a state space model for data with a continuous and cyclic trend in the mean. Furthermore a matching Kalman filter is developed. The model is tested on progesterone data from 112 cow-lactations with the purpose of evaluating the use of progesterone for detection of oestrus.

There is concern whether consuming products based on crop from Chlormequat-treated fields can cause reproduction problems in animals and humans. An experiment is presently being conducted to investigate this using the pig as a model. Considerations with regard to experimental design when investigating whether differently treated crop can affect animal/human biology is discussed. Only about half of the data are presently available. A preliminary survey of these data does not show clear differences between Chlormequat-treated and organic non-treated wheat with regard to reproduction performance of pigs.

We have developed a package, called , consisting of a number of classes and associated methods to support the analysis of data using graphical models. It is developed for the open source language, R, and is available for several platforms. The package is intended to be widely extendible and flexible so that package developers may implement further types of graphical models using the available methods. contains methods for representing data, specification of models using a formal language, and is linked to , an interactive graphical user interface for manipulating graphs. We show how these building blocks can be combined and integrated with inference engines in the special cases of hierarchical log-linear models (undirected models). gRbase gRbase dynamicGraph