Supervised pattern recognition to discriminate the geographical origin of rice bran oils: a first study

Supervised pattern recognition appears to be a useful tool to authenticate foodstuffs according to their geographical or varietal origin, when a set of samples whose classification is known a priori are available. In this work, linear discriminant analysis and artificial neural networks trained by the back-propagation algorithm have been used to discriminate rice bran oils manufactured in three different countries (Italy, Thailand and Switzerland) according to their geographical origin. The variables to be included in the mathematical models have been chosen by means of Fisher F-ratio value among the chemical indices routinely determined on vegetable oils (particularly fatty acids, triglycerides and sterol composition). The prediction ability of all the classifiers was 100% as evaluated by cross-validation.     

Approssimazione «tipo Monte Carlo» per le soluzioni del 1o problema al contorno per equazioni paraboliche «forward-backward»

Si studia il primo problema di valori al contorno in aperti ω diR ^m+1 per l’equazione parabolica: i coefficientik(x), a _ik(t, x) e il termine notof(t, x) sono funzioni reali, misurabili e limitate(k(x) può cambiare segno in ω. Il risultato principale ottenuto consiste in ciò: se una soluzione, appartenente ad una determinata classe funzionale, esiste, questa è approssimata uniformemente dalle soluzioni di certe equazioni integrali (le quali esistono e sono uniche). Queste equazioni integrali simulano, in un certo senso, il problema differenziale. Questo metodo di approssimazione assomiglia (nel senso precisato nel paragrafo 4) al noto metodo Monte-Carlo.     

Matrix-assisted pulsed laser evaporation of chemoselective polymers

The numerous properties of liposomes, i.e., nontoxicity, biodegradability, and their ability to encapsulate different biological active substances in aqueous and lipid phase, make them perfect models of biomembranes. Liposomes made up of phospholipids may be used to study new applications such as cell targeting or, under specific experimental conditions, may be applied in micro and nano-sized biosensors.     

Absence of Critical Mass Densities for a Vibrating Membrane

We study the dependence of the eigenvalues of a N-dimensional vibrating membrane upon variation of the mass density. We prove that the elementary symmetric functions of the eigenvalues depend real-analytically on the mass density and that such functions have no critical points with constant mass constraint. In particular, the elementary symmetric functions of the eigenvalues, hence all simple eigenvalues, have no local maxima or minima on the set of those mass densities with a prescribed total mass.     

A reduction approach for determining generalized simple waves

A reduction approach is developed in order to construct generalized simple wave solutions to quasilinear nonhomogenous hyperbolic systems of first order PDEs. The solutions sought must possess a special ansatz which permits time-evolution of the profile of a simple wave due to a source-like term. These solutions involve a free function which can be used to fit classes of initial or boundary value problems. By means of the proposed approach two governing models of interest in a variety of applications are investigated. Model constitutive laws consistent with the full reduction process are obtained and the occurence of singularities at a finite time for the resulting solutions is analysed. Furthermore a comparison is made between the results obtained within the present theoretical framework and the standard simple wave solutions of the corresponding homogeneous (source free) governing models.     

On projective spaces PG(r, q) with r ≥ 4

In this paper a characterization of PG(r, q), r ≥ 4, in terms of planar spaces is given. Domenico Olanda: This research was carried out within the activity of INdAM-GNSAGA and supported by the Italian Ministry M.I.U.R.