Semisupervised Dimensionality Reduction and Classification Through Virtual Label Regression

Semisupervised dimensionality reduction has been attracting much attention as it not only utilizes both labeled and unlabeled data simultaneously, but also works well in the situation of out-of-sample. This paper proposes an effective approach of semisupervised dimensionality reduction through label propagation and label regression. Different from previous efforts, the new approach propagates the label information from labeled to unlabeled data with a well-designed mechanism of random walks, in which outliers are effectively detected and the obtained virtual labels of unlabeled data can be well encoded in a weighted regression model. These virtual labels are thereafter regressed with a linear model to calculate the projection matrix for dimensionality reduction. By this means, when the manifold or the clustering assumption of data is satisfied, the labels of labeled data can be correctly propagated to the unlabeled data; and thus, the proposed approach utilizes the labeled and the unlabeled data more effectively than previous work. Experimental results are carried out upon several databases, and the advantage of the new approach is well demonstrated.

Calculating the equation of state parameters and predicting the spinodal curve of isotactic polypropylene/poly(ethylene-co-octene) blend by molecular dynamics simulations combined with Sanchez-Lacombe lattice fluid theory

Molecular dynamics simulations are adopted to calculate the equation of state characteristic parameters P*, rho*, and T* of isotactic polypropylene (iPP) and poly(ethylene-co-octene) (PEOC), which can be further used in the Sanchez-Lacombe lattice fluid theory (SLLFT) to describe the respective physical properties. The calculated T* is a function of the temperature, which was also found in the literature. To solve this problem, we propose a Boltzmann fitting of the data and obtain T* at the high-temperature limit. With these characteristic parameters, the pressure-volume-temperature (PVT) data of iPP and PEOC are predicted by the SLLFT equation of state. To justify the correctness of our results, we also obtain the PVT data for iPP and PEOC by experiments. Good agreement is found between the two sets of data. By integrating the Euler-Lagrange equation and the Cahn-Hilliard relation, we predict the density profiles and the surface tensions for iPP and PEOC, respectively. Furthermore, a recursive method is proposed to obtain the characteristic interaction energy parameter between iPP and PEOC. This method, which does not require fitting to the experimental phase equilibrium data, suggests an alternative way to predict the phase diagrams that are not easily obtained in experiments.

Effective descriptions of molecular structures and the quantitative structure - Activity relationship studies

In this research. we found CoMFA alone could not obtain sufficiently a strong equation to allow confident prediction for aminobenzenes. When some other parameter. such as heat of molecular formation of the compounds, was introduced into the CoMFA model, the results Were improved greatly. It gives us a hint that a better description for molecular structures will yield a better prediction model, and this hint challenged us to look for another method-the projection areas of molecules in 3D space for 3D-QSAR. It is surprising that much better results than that obtained by using CoMFA Were achieved. Besides the CoMFA analysis. multiregression analysis and neural network methods for building the models were used in this paper.

Selection for variables in quantitative structure-activity/property relationship studies - Comparison between methods of orthogonal descriptors and leaps-and-bounds regression analysis

In this paper, the comparison of orthogonal descriptors and Leaps-and-Bounds regression analysis is performed. The results obtained by using orthogonal descriptors are better than that obtained by using Leaps-and-Bounds regression for the data set of nitrobenzenes used in this study. Leaps-and-Bounds regression can be used effectively for selection of variables in quantitative structure-activity/property relationship(QSAR/QSPR) studies. Consequently, orthogonalisation of descriptors is also a good method for variable selection for studies on QSAR/QSPR.

The glass transition temperatures of PS/PPO blends: Couchman volume-based equation and its verification

The glass transition temperatures (T-g) of PS/PPO blends with different compositions were studied under various pressures by means of a PVT-100 analyzer. A general relation of T-g and pressure of the PS/PPO system was deduced by fitting the experimental T-g's. Couchman volume-based equation was testified with the aid of those data. It was found that the experimental T-g's do not obey the Couchman equation of glass transition temperature based on thermodynamic theory. According to our studies, the major reason of the deviation is caused by the neglect of DeltaV(mix). (C) 2001 Published by Elsevier Science Ltd.

Studies on the method of leaps and bounds regression

In this paper, we introduce the method of leaps and bounds regression which can be used to select variables quickly and obtain the best regression models. These models contain one variable, two variables, three variables and so on. The results obtained by using leaps and bounds regression were compared with those achieved by using stepwise regression to lead to the conclusion that leaps and bounds regression is an effective method.

Prediction of molar absorptivities of color reagents during their color reactions with cerium using multiple regression analysis and neural network

A quantitative structure-property study has been made on the relationship between molar absorptivities (epsilon) of asymmetrical phosphone bisazo derivatives of chromotropic acid and their color reactions with cerium by multiple regression analysis and neural network. The new topological indices A(x1) - A(x3) suggested in our laboratory and molecular connectivity indices of 43 compounds have been calculated. The results obtained from the two methods are compared. The neural network model is superior to the regression analysis technique and gave a prediction which was sufficiently accurate to estimate the molar absorptivities of color reagents during their color reactions with cerium.

Study on Phenol Compounds Using Multiple Regression and Neural Network Methods

In this paper, the molecular connectivity indices and the electronic charge parameters of forty-eight phenol compounds nave been calculated. and applied for studying the relationship between partition coefficients and structure of phenol compounds. The results demonstrate that the properties of compounds can be described better with selective parameters, and the results obtained by neural network are superior to that by multiplle regression.

Quantitative structure-property relationships for colour reagents and their colour reactions with ytterbium using regression analysis and computational neural networks

In this paper, the new topological indices A(x1)-A(x3) suggested in our laboratory and molecular connectivity indices have been applied to multivariate analysis in structure-property studies. The topological indices of twenty asymmetrical phosphono bisazo derivatives of chromotropic acid have been calculated. The structure-property relationships between colour reagents and their colour reactions with ytterbium have been studied by A(x1)-A(x3) indices and molecular connectivity indices with satisfactory results. Multiple regression analysis and neural networks were employed simultaneously in this study.

QUANTITATIVE STRUCTURE-ACTIVITY-RELATIONSHIPS FOR TOXICITY OF PHENOLS USING REGRESSION-ANALYSIS AND COMPUTATIONAL NEURAL NETWORKS

Quantitative structure-toxicity models were developed that directly link the molecular structures of a et of 50 alkYlated and/or halogenated phenols with their polar narcosis toxicity, expressed as the negative logarithm of the IGC50 (50% growth inhibitor

EQUATION OF STATE FOR POLYMER SOLIDS

A new equation of state for polymer solids is given by P = B0/4 98[(V0/V)7.14 - (V0/V)2.16 + T/T0] comparison of the equation of state with experimental data is made for six kinds of polymers at different temperatures and pressures. The results obtained shown that the equation is suitable to describe the compression behavior of solid polymers in the region without transition.

A NEW ISOTHERMAL EQUATION OF STATE FOR POLYMERS

A new isothermal equation of state for polymers in the solid and the liquid is given by P = B(T, 0)/(n - m){[V(T, 0)/V(T, P)]n + 1 - [V(T, 0)/V(T, P)]m + 1} where n = 6.14 and m = 1.16 are general constant's for polymer systems. Comparison of the equation with experimental data is made for six polymers at different temperatures and pressures. The results predict that the equation of state describes the isothermal compression behaviour of polymers in the glass and the melt states, except at the transition temperature.

DETERMINATION OF THE MARK-HOUWINK EQUATION FOR CHITOSANS WITH DIFFERENT DEGREES OF DEACETYLATION

The values of k and alpha in the Mark-Houwink equation have been determined for chitosans with different degrees of deacetylation (DD) (69, 84, 91 and 100% respectively), in 0.2 M CH3COOH/0.1 M CH3COONa aqueous solution at 30-degrees-C by the light scattering method. It was shown that the values of alpha-decreased from 1.12 to 0.81 and the values of k increased from 0.104 x 10(-3) to 16.80 x 10(-3) ml/g, when the DD varied from 69 to 100%. This is due to a reduction of rigidity of the molecular chain and an increase of the electrostatic repulsion force of the ionic groups along the polyelectrolyte chain in chitosan solution, when the DD of chitosan increases gradually.