This paper studies the principle of text categorization in which Jensen-Shannon Divergence is used to calculate text similarity, comparing its accuracy of classification and time taking to the ...traditional Cosine Similarity algorithm. Experimental research shows that Jensen-Shannon Divergence algorithm will reach better results when test materials remain unchanged.
This paper deals with empirical multiresolution linear models intended for image processing. Such models contain information about gray-level composition of regions in the image. First, a general ...method for building these models from samples of selected images is described. Then, a measure of their quality, based on the Jensen-Shannon divergence, is introduced. This divergence is also used as a distance measure for classifying images. Applications in non-linear image filtering are provided, giving better result than classical median filtering.
In this paper, we propose a framework to characterize and compare two search engine results. Typical user-queries are ambiguous and, consequentially, each search engine will compute ranks in ...different manners, attempting to answer them in the best possible way. Thus, each search engine will have its own bias. Given the importance of the first page results in Web Search Engines, in this paper we propose a framework to assess the information presented in the first page by measuring the information entropy and the correlations between two ranks. Employing the recently proposed Rank-Biased Overlap measure 2 we compare to which extent do Bing and Google rankings in fact differ. We also extend this measure and propose a measure for comparing the information entropy present in two ranks. The proposed measure is based on the correlation of two ranks and the application of Jensen-Shannon's divergence among two document sets. Our methodology starts with 40,000 user queries and crawls the search results for these queries on both search engines. The results allow us to determine the search engines correlations, crawling coverage, information overlap, and information entropy.
We propose a bottom-up analysis method of multi-modal dialogue interaction with a pattern and motif mining method to summarize such interviews as between doctors and patients for medical diagnosis. ...Our aim is to generate a hierarchical model of the interviewing behavior of such kinds as interaction corpora, consisting of primitive, pattern, motif, and pattern clusters from the given dialogue session data. We exploit a Jensen-Shannon Divergence measure to extract important patterns and motifs. Medical interview is chosen as an important application of such analysis because a doctor's multi-modal interviewing technique is essential to establish a reliable relationship and to conclude with a successful diagnosis.
An interaction corpus of example simulated medical interviews is constructed by the proposed method. The interviews are captured by a video camera and microphones. Based on the constructed indices in terms of given pattern notations and clusters, the interviews were summarized. Performance evaluation of the indices by a medical doctor was performed to confirm their plausibility and summary descriptions of the results.
The Jensen-Bregman divergence is a distortion measure defined by the Jensen difference provided by a strictly convex function. Jensen-Bregman divergences extend the well-known Jensen-Shannon ...divergence by allowing to choose an arbitrary convex function generator instead of the standard Shannon entropy. This class of Jensen-Bregman divergences notably includes the squared Euclidean distance. Although Jensen-Bregman divergences are symmetric distances by construction, they are not metrics as they violate the triangle inequality. We study the geometric properties and combinatorial complexities of both the Voronoi diagrams and the centroidal Voronoi diagrams induced by such as class of information-theoretic divergences. We show that those Jensen-Bregman divergences appear naturally into two contexts: (1) when symmetrizing Bregman divergences, and (2) when computing the Bhattacharyya distances of statistical distributions. The Bhattacharyya distance measures the amount of overlap of distributions, and is popularly used to provide both lower and upper bounds in machine learning: the Bayes' misclassification error. Since the Bhattacharyya distance of popular distributions in statistics called the exponential families (including familiar Gaussian, Poisson, multinomial, Beta/Gamma families, etc.) can be computed equivalently as Jensen-Bregman divergences, the Jensen-Bregman Voronoi diagrams allow one also to study statistical Voronoi diagrams induced by an entropic function.
Tsallis kernels on measures Martins, A.; Aguiarz, P.; Figueiredo, M.
2008 IEEE Information Theory Workshop
Conference Proceeding
Odprti dostop
Recent approaches to classification of text, images, and other types of structured data, launched the quest for positive definite (p.d.) kernels on probability measures. In particular, kernels based ...on the Jensen-Shannon (JS) divergence and other information-theoretic quantities have been proposed. We introduce new JS-type divergences, by extending its two building blocks: convexity and Shannonpsilas entropy. These divergences are then used to define new information-theoretic kernels on measures. In particular, we introduce a new concept of q-convexity, for which a Jensen q-inequality is proved. Based on this inequality, we introduce the Jensen-Tsallis q-difference, a nonextensive generalization of the Jensen-Shannon divergence. Furthermore, we provide denormalization formulae for entropies and divergences, which we use to define a family of nonextensive information-theoretic kernels on measures. This family, grounded in nonextensive entropies, extends Jensen-Shannon divergence kernels, and allows assigning weights to its arguments.
A novel image segmentation method using conditional random fields is presented in this paper. It dynamically fuses color, texture, spatial and edge information to implement image segmentation ...associated with interactive inputs. Most existing algorithms combine multiple cues statically with a constant ratio, it is hard for them to describe the inherent property of different images efficiently. The proposed method takes the full advantage of the user labeled information about the foreground and background, and builds a certain standard to measure the reliability of established probability distribution, with which to assemble energy terms in the conditional random field. This makes the related energy terms fusing dynamically in accordance with the image inner property, and improves the segmentation power of the model. From experiments we find that the conditional random field is able to capture the image edge characteristics and produce meaningful outputs. These results demonstrate that this method performs steady and works well on various natural images.
Standard properties of ϕ-divergences of probability measures are widely applied in various areas of information processing. Among the desirable supplementary properties facilitating employment of ...mathematical methods is the metricity of ϕ-divergences, or the metricity of their powers. This paper extends the previously known family of ϕ-divergences with these properties. The extension consists of a continuum of ϕ-divergences which are squared metric distances and which are mostly new but include also some classical cases like e. g. the Le Cam squared distance. The paper establishes also basic properties of the ϕ-divergences from the extended class including the range of values and the upper and lower bounds attained under fixed total variation.