Directional features in online handwriting recognition

The selection of valuable features is crucial in pattern recognition. In this paper we deal with the issue that part of features originate from directional instead of common linear data. Both for directional and linear data a theory for a statistical modeling exists. However, none of these theories gives an integrated solution to problems, where linear and directional variables are to be combined in a single, multivariate probability density function. We describe a general approach for a unified statistical modeling, given the constraint that variances of the circular variables are small. The method is practically evaluated in the context of our online handwriting recognition system frog on hand and the so-called tangent slope angle feature. Recognition results are compared with two alternative modeling approaches. The proposed solution gives significant improvements in recognition accuracy, computational speed and memory requirements.     

Comparative analysis of urban reflectance and surface temperature

Urban environmental conditions are strongly dependent on the biophysical properties and radiant thermal field of the land cover elements in the urban mosaic. Observations of urban reflectance and surface temperature provide valuable constraints on the physical properties that are determinants of mass and energy fluxes in the urban environment. Consistencies in the covariation of surface temperature with reflectance properties can be parameterized to represent characteristics of the surface energy flux associated with different land covers and physical conditions. Linear mixture models can accurately represent Landsat ETM+ reflectances as fractions of generic spectral endmembers that correspond to land surface materials with distinct physical properties. Modeling heterogeneous land cover as mixtures of rock and/or soil Substrate, Vegetation and non-reflective Dark surface (SVD) generic endmembers makes it possible to quantify the dependence of aggregate surface temperature on the relative abundance of each physical component of the land cover, thereby distinguishing the effects of vegetation abundance, soil exposure, albedo and shadowing. Comparing these covariations in a wide variety of urban settings and physical environments provides a more robust indication of the global variability in these parameter spaces than could be inferred from a single study area. A comparative analysis of 24 urban areas and their non-urban peripheries illustrates the variability in the urban thermal fields and its dependence on biophysical land surface components. Contrary to expectation, moderate resolution intra-urban variations in surface temperature are generally as large as regional surface heat island signatures in these urban areas. Many of the non-temperate urban areas did not have surface heat island signatures at all. However, the multivariate distributions of surface temperature and generic endmember fractions reveal consistent patterns of thermal fraction covariation resulting from land cover characteristics. The Thermal-Vegetation (TV) fraction space illustrates the considerable variability in the well-known inverse correlation between surface temperature and vegetation fraction at moderate (< 100 m) spatial resolutions. The Thermal-Substrate (TS) fraction space reveals energetic thresholds where competing effects of albedo, illumination and soil moisture determine the covariation of maximum and minimum temperature with illuminated substrate fraction. The dark surface endmember fraction represents a fundamental ambiguity in the radiance signal because it can correspond to either absorptive (e.g. low albedo asphalt), transmissive (e.g. deep clear water) or shadowed (e.g. tree canopy shadow) surfaces. However, in areas where dark surface composition can be inferred from spatial context, the different responses of these surfaces may still allow them to be distinguished in the thermal fraction space.     

Non-redundant data clustering

Data clustering is a popular approach for automatically finding classes, concepts, or groups of patterns. In practice, this discovery process should avoid redundancies with existing knowledge about class structures or groupings, and reveal novel, previously unknown aspects of the data. In order to deal with this problem, we present an extension of the information bottleneck framework, called coordinated conditional information bottleneck, which takes negative relevance information into account by maximizing a conditional mutual information score subject to constraints. Algorithmically, one can apply an alternating optimization scheme that can be used in conjunction with different types of numeric and non-numeric attributes. We discuss extensions of the technique to the tasks of semi-supervised classification and enumeration of successive non-redundant clusterings. We present experimental results for applications in text mining and computer vision.     

Hazardous waste management problem: The case for incineration

We define the hazardous waste management problem as the combined decisions of selecting the disposal method, siting the disposal plants and deciding on the waste flow structure. The hazardous waste management problem has additional requirements depending on the selected disposal method. In this paper we focus on incineration, for which the main additional requirement is to satisfy the air pollution standards imposed by the governmental restrictions. We propose a cost-based mathematical model in which the satisfaction of air pollution standards is also incorporated. We used the Gaussian Plume equation in measuring the air pollution concentrations at population centers. A large-scale implementation of the proposed model within Turkey is provided.     

Learning Recursive Bayesian Multinets for Data Clustering by Means of Constructive Induction

This paper introduces and evaluates a new class of knowledge model, the recursive Bayesian multinet (RBMN), which encodes the joint probability distribution of a given database. RBMNs extend Bayesian networks (BNs) as well as partitional clustering systems. Briefly, a RBMN is a decision tree with component BNs at the leaves. A RBMN is learnt using a greedy, heuristic approach akin to that used by many supervised decision tree learners, but where BNs are learnt at leaves using constructive induction. A key idea is to treat expected data as real data. This allows us to complete the database and to take advantage of a closed form for the marginal likelihood of the expected complete data that factorizes into separate marginal likelihoods for each family (a node and its parents). Our approach is evaluated on synthetic and real-world databases.     

The mirror effect and mixture signal detection theory.

The mirror effect for word frequency refers to the finding that low-frequency words have higher hit rates and lower false alarm rates than high-frequency words. This result is typically interpreted in terms of conventional signal detection theory (SDT), in which case it indicates that the order of the underlying old item distributions mirrors the order of the new item distributions. However, when viewed in terms of a mixture version of SDT, the order of hits and false alarms does not necessarily imply the same order in the underlying distributions because of possible effects of mixing. A reversal in underlying distributions did not appear for fits of mixture SDT models to data from 4 experiments. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

基于人工神经网络的混合气体定量检测方法

将气体传感器阵列与人工神经网络相结合，设计了用于混合气体定量检测的人工嗅觉系统；构造了局部连接的前馈神经网络的结构和基于递推预报误差的网络训练算法；通过实验对CO，H2，CH4三种气体的混合物进行了分析，实验结果证明气体传感器阵列和人工神经网络用于混合气体定量检测是可行的。     

Extensions of LDA by PCA mixture model and class-wise features

Linear discriminant analysis (LDA) is a data discrimination technique that seeks transformation to maximize the ratio of the between-class scatter and the within-class scatter. While it has been successfully applied to several applications, it has two limitations, both concerning the underfitting problem. First, it fails to discriminate data with complex distributions since all data in each class are assumed to be distributed in the Gaussian manner. Second, it can lose class-wise information, since it produces only one transformation over the entire range of classes. We propose three extensions of LDA to overcome the above problems. The first extension overcomes the first problem by modelling the within-class scatter using a PCA mixture model that can represent more complex distribution. The second extension overcomes the second problem by taking different transformation for each class in order to provide class-wise features. The third extension combines these two modifications by representing each class in terms of the PCA mixture model and taking different transformation for each mixture component. It is shown that all our proposed extensions of LDA outperform LDA concerning classification errors for synthetic data classification, hand-written digit recognition, and alphabet recognition.     

Question hard, answer simply: A comment on Storms et al. (2003).

It seems clear that, for whatever reasons, the dementia of the Alzheimer type patient group (as well as other patient groups) exhibits behavior that is different from the normal control group. G. Storms, T. Dirikx, J. Saerens, S. Verstraeten, and P. P. De Deyn (2003) rightfully argue that the observed behavior (similarity judgments) does not tell us the source (cause) of the differences between the 2 groups. Rather, the focus of the study should be placed more on finding the ways the 2 groups are different. They also point out various methodological problems in some of the previous attempts to characterize the nature of the differences. Further methodological issues in G. Storms et al.'s study are examined. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

一种应用于二次布局的有效划分方法

提出了一种基于二次布局的结合MFFC结群和hMETIS划分的算法.实验表明:这种方法能得到很好的布局结果,但是运行消耗的时间比较长.为了缩短划分在二次布局中运行的时间,提出了一种改进的结群算法IMFFC,用它在二次布局中做划分.与前者相比较,这种方法虽然布局质量稍差,但速度更快.