A comparative evaluation of nonlinear dynamics methods for time series prediction

A key problem in time series prediction using autoregressive models is to fix the model order, namely the number of past samples required to model the time series adequately. The estimation of the model order using cross-validation may be a long process. In this paper, we investigate alternative methods to cross-validation, based on nonlinear dynamics methods, namely Grassberger–Procaccia, Kégl, Levina–Bickel and False Nearest Neighbors algorithms. The experiments have been performed in two different ways. In the first case, the model order has been used to carry out the prediction, performed by a SVM for regression on three real data time series showing that nonlinear dynamics methods have performances very close to the cross-validation ones. In the second case, we have tested the accuracy of nonlinear dynamics methods in predicting the known model order of synthetic time series. In this case, most of the methods have yielded a correct estimate and when the estimate was not correct, the value was very close to the real one.     

Protection of low density lipoprotein oxidation at chemical and cellular level by the antioxidant drug dipyridamole

1. The oxidative modification of low density lipoprotein (LDL) is thought to be an important factor in the initiation and development of atherosclerosis. Natural and synthetic antioxidants have been shown to protect LDL from oxidation and to inhibit atherosclerosis development in animals. Synthetic antioxidants are currently being tested, by they are not necessarily safe for human use. 2. We have previously reported that dipyridamole, currently used in clinical practice, is a potent scavenger of free radicals. Thus, we tested whether dipyridamole could affect LDL oxidation at chemical and cellular level. 3. Chemically induced LDL oxidation was made by Cu(II), Cu(II) plus hydrogen peroxide or peroxyl radicals generated by thermolysis of 2,2'-azo-bis(2-amidino propane). Dipyridamole, (1-10 microM), inhibited LDL oxidation as monitored by diene formation, evolution of hydroperoxides and thiobarbituric acid reactive substances, apoprotein modification and by the fluorescence of cis-parinaric acid. 4. The physiological relevance of the antioxidant activity was validated by experiments at the cellular level where dipyridamole inhibited endothelial cell-mediated LDL oxidation, their degradation by monocytes, and cytotoxicity. 5. In comparison with ascorbic acid, alpha-tocopherol and probucol, dipyridamole was the more efficient antioxidant with the following order of activity: dipyridamole > probucol > ascorbic acid > alpha-tocopherol. The present study shows that dipyridamole inhibits oxidation of LDL at pharmacologically relevant concentrations. The inhibition of LDL oxidation is unequivocally confirmed by use of three different methods of chemical oxidation, by several methods of oxidation monitoring, and the pharmacological relevance is demonstrated by the superiority of dipyridamole over the naturally occurring antioxidants, ascorbic acid and alpha-tocopherol and the synthetic antioxidant probucol.     

Enhanced lipid peroxidation in hepatic cirrhosis

> Objective: This paper describes the deliberations of an interdisciplinary group of clinical and basic scientists who met at the National Institute of Child Health and Human Development to discuss the potential role of fetal behavior in assessing fetal well being and predicting neonatal outcome. The conference focused on three aspects of fetal behavior: 1) habituation; 2) state transitions; and 3) movement. Methods: The participants consisted of 25 leaders in the fields of obstetrics, perinatal medicine, neonatology, developmental psychobiology, developmental neuroscience, developmental psychology, ethology, and mathematics. The meeting was divided into three parts. In each of these a plenary speaker (a recognized expert in his field) began the session with an overview of the scientific theme. Two respondents, with research expertise in fetal research (animal models or human fetuses) followed with remarks on the plenary talk and comments based on their own studies. At the conclusion of these comments, the participants met in small groups to discuss the plenary proceedings and their implications for assessing human fetal well being and predicting outcome. At the conclusion of the small group deliberations all of the participants reconvened in a plenary session. During this part of the meeting a rapporteur from each small group summarized their discussions. Results and Conclusions: 1) Fetal habituation: there was a general consensus that research on this aspect of fetal behavior may have a high payoff for assessing human fetal well being and predicting neonatal outcome. 2) Behavioral state transitions: participants agreed that transitions afford investigators with an indication of when (timing) and how (models) behavior changes within and between developmental periods. Knowledge of transitions during development allows for tracking of behaviors that may be necessary for the fetus to adapt to its in utero environment or prepare for its postnatal life. 3) Chaos theory and fetal movement: participants concluded that non-linear dynamics systems analysis models could be useful to analyze "noise" within a measurement system; better define time scales; and increase resolution and thereby better identify "signals."     

Linear SVM-based recognition of elementary juggling movements using correlation dimension of Euler Angles of a single arm

Recognizing the human arm movements has several applications, and it can be performed in a number of ways through the use of one or more sensor devices that the technology offers. This paper aims to exploit the exercises performed by jugglers in order to recognize the arm movements on the basis of the only information on the arm orientation provided by the Euler Angles. The proposed recognizer has two modules, i.e., a feature extractor and a classifier. The former reconstructs the dynamics of the system and estimates three correlation dimensions, each associated with a given Euler Angle. The latter is formed by a Linear Support Vector Machine. Extensive experimentations show the effectiveness of the proposed approach.

     

Estimating the intrinsic dimension of data with a fractal-based method

In this paper, the problem of estimating the intrinsic dimension of a data set is investigated. A fractal-based approach using the Grassberger-Procaccia algorithm is proposed. Since the Grassberger-Procaccia algorithm (1983) performs badly on sets of high dimensionality, an empirical procedure that improves the original algorithm has been developed. The procedure has been tested on data sets of known dimensionality and on time series of Santa Fe competition.     

Handy: A real-time three color glove-based gesture recognizer with learning vector quantization

This paper presents Handy, a real-time hand gesture recognizer based on a three color glove. The recognizer is formed by three modules. The first module, fed by the frame acquired by a webcam, identifies the hand image in the scene. The second module, a feature extractor, represents the image by a nine-dimensional feature vector. The third module, the classifier, is performed by means of learning vector quantization. The recognizer, tested on a dataset of 907 hand gestures, has shown very high recognition rate.     

Intrinsic Dimension Estimation of Data: An Approach Based on Grassberger–Procaccia's Algorithm

In this paper the problem of estimating the intrinsic dimension of a data set is investigated. An approach based on the Grassberger–Procaccia's algorithm has been studied. Since this algorithm does not yield accurate measures in high-dimensional data sets, an empirical procedure has been developed. Grassberger–Procaccia's algorithm was tested on two different benchmarks and was compared to a TRN-based method.     

A novel kernel method for clustering

Kernel methods are algorithms that, by replacing the inner product with an appropriate positive definite function, implicitly perform a nonlinear mapping of the input data into a high-dimensional feature space. In this paper, we present a kernel method for clustering inspired by the classical k-means algorithm in which each cluster is iteratively refined using a one-class support vector machine. Our method, which can be easily implemented, compares favorably with respect to popular clustering algorithms, like k-means, neural gas, and self-organizing maps, on a synthetic data set and three UCI real data benchmarks (IRIS data, Wisconsin breast cancer database, Spam database).     

Autonomic and hemodynamic responses to insulin in lean and obese humans

To study the acute effects of insulin on autonomic control of cardiac function, we performed spectral analysis of heart rate variability and measured cardiac dynamics (by two-dimensional echocardiography) in 18 obese (BMI = 35 +/- 1 kg.m-2) and 14 lean (BMI = 24 +/- 1 kg.m-2) subjects in the basal state and in response to physiological hyperinsulinemia (1 mU.min-1.kg-1 insulin clamp). In the lean group, insulin promptly (within 20 min) and consistently depressed spectral powers, both in the low-frequency and high-frequency range. These changes were twice as large as accounted for by the concomitant changes in heart rate (68 +/- 2 to 70 +/- 2 beats/min). At the end of the 2-h clamp, stroke volume (67 +/- 4 to 76 +/- 9 ml.min-1) and cardiac output (4.45 +/- 0.21 to 5.06 +/- 0.55 l.min-1) rose, whereas peripheral vascular resistance fell. The low-to-high frequency ratio increased from 1.7 +/- 0.2 to 2.3 +/- 0.3 (P < 0.01), indicating sympathetic shift of autonomic balance. In the obese group, all basal spectral powers were significantly lower (by 40% on average) than in the lean group, and were further reduced by insulin administration. The low-to-high frequency ratio was higher than in controls at baseline (2.4 +/- 0.4, P < 0.03), and failed to increase after insulin (2.2 +/- 0.3, P = ns). Furthermore, obesity was associated with higher resting stroke volume (89 +/- 5 vs. 67 +/- 4 ml.min-1, P < 0.01) and cardiac output (6.01 +/- 0.31 vs. 4.45 +/- 0.21 l.min-1, P = 0.001) but lower peripheral vascular resistance (15.1 +/- 0.8 vs. 19.2 +/- 1.1 mmHg.min.L-1, P = 0.002), whereas mean arterial blood pressure was similar to control (90 +/- 2 vs. 86 +/- 2 mmHg, P = not significant). We conclude that physiological hyperinsulinemia causes acute desensitization of sinus node activity to both sympathetic and para-sympathetic stimuli, sympathetic shift of autonomic balance, and a high-output, low-resistance hemodynamic state. In the obese, these changes are already present in the basal state, and may therefore be linked with chronic hyperinsulinemia.     

A survey of kernel and spectral methods for clustering

Clustering algorithms are a useful tool to explore data structures and have been employed in many disciplines. The focus of this paper is the partitioning clustering problem with a special interest in two recent approaches: kernel and spectral methods. The aim of this paper is to present a survey of kernel and spectral clustering methods, two approaches able to produce nonlinear separating hypersurfaces between clusters. The presented kernel clustering methods are the kernel version of many classical clustering algorithms, e.g., K-means, SOM and neural gas. Spectral clustering arise from concepts in spectral graph theory and the clustering problem is configured as a graph cut problem where an appropriate objective function has to be optimized. An explicit proof of the fact that these two paradigms have the same objective is reported since it has been proven that these two seemingly different approaches have the same mathematical foundation. Besides, fuzzy kernel clustering methods are presented as extensions of kernel K-means clustering algorithm.