Jorge Magallanes

The mean angular distance among objects and its relationships with Kohonen artificial neural networks.

This job refers to classification of multidimensional objects and Kohonen artificial neural networks. A new concept is introduced, called the mean angular distance among objects (MADO). Its value can be calculated as the cosine of the mean centered vectors between objects. It can be expressed in matrix form for any number of objects. The MADO allows us to interpret the final organization of the objects in a Kohonen map. Simulated examples demonstrate the relationship between MADO and Kohonen maps and show a way to take advantage of the information present in both of them. Finally, a real analytical chemistry case is analyzed as an application on a big data set of an air quality monitoring campaign. It is possible to discover in it a subgroup of objects with different characteristics than those of the general trend. This subgroup is linked to the existence of an unidentified SO(2) source that, a priori, has not been taken into account.

2D mapping by Kohonen networks of the air quality data from a large city.

The 15-variable environmental data (7 concentrations: CO, SO2, O3, NOx, NO, NO2, particulate matter smaller than 10 micron (PM10), and 8 weather data: cloudiness, rainfall, insolation factor (Isfi), temperature, pressure at two locations, and wind intensity with direction) in a period of 45 days with 1-h intervals were extracted from a larger database of concentrations recorded in minute intervals for the same time period. The monitoring site was located in the City of Buenos Aires in a relatively heavy traffic crossroad of two avenues. The data required special pretreatment where the hourly content of rain, wind intensity, wind velocity, and cloudiness were concerned. The new variable named insolation factor (relative UV radiation) calculated on the basis of the general meteorological data, the geographic position of the monitoring site, cloudiness, date, and the time of the recording was composed. The relative intensity of UV radiation was modeled by a Gaussian function, multiplied by a cloudiness factor. Based on the 14-variable input and the 1-variable output (ozone) data, first, the clustering of all 980 data records was made. The top map clustering showing the ozone concentration was related to the maps of all 14 variables. The link between O3 clusters, NO2, and Isfi weight levels is shown and discussed. As a preliminary result of this study some of the most interesting correlations between the maps and remaining variables are given.

Uncovering interactions in Plackett-Burman screening designs applied to analytical systems. A Monte Carlo ant colony optimization approach.

Screening of relevant factors using Plackett-Burman designs is usual in analytical chemistry. It relies on the assumption that factor interactions are negligible; however, failure of recognizing such interactions may lead to incorrect results. Factor associations can be revealed by feature selection techniques such as ant colony optimization. This method has been combined with a Monte Carlo approach, developing a new algorithm for assessing both main and interaction terms when analyzing the influence of experimental factors through a Plackett-Burman design of experiments. The results for both simulated and analytically relevant experimental systems show excellent agreement with previous approaches, highlighting the importance of considering potential interactions when conducting a screening search.

Kohonen classification applying ‘missing variables’ criterion to evaluate the p-boronophenylalanine human-body-concentration decreasing profile of boron neutron capture therapy patients

The irradiation dose in tumor and healthy tissue of a boron neutron capture therapy (BNCT) patient depends on the boron concentration in blood. In most treatments, this concentration is experimentally determined before and after irradiation but not while irradiation is being carried out because it is troublesome to take the blood samples when the patient remains isolated in the irradiation room. A few models are used to predict the boron profile during that period, which until now involves a biexponential decay. For the prediction of decay concentration profiles of the p‐boronophenylalanine (BPA) in the human body during BNCT treatment, a Kohonen‐based neural network method is suggested. The results of various (20 × 20 × 40 Kohonen network) models based on different trainings on the data set of 67 concentration sets (profiles) are described and discussed. The prediction ability and robustness of the modeling method were tested by the leave‐one‐out procedure. The results show that the method is very robust and mostly independent of small variations. It can yield predictions, root mean squared prediction error (RMSPE), with a maximum of 3.30 µg g−1 for the present cases. In order to show the abilities and limitations of the method, the best and the few worst results are discussed in detail. It should be emphasized that one of the main advantages of this method is the automatic improvement in the prediction ability and robustness of the model by feeding it with an increasing number of data. Copyright

Artificial neural networks to evaluate the boron concentration decreasing profile in blood-BPA samples of BNCT patients.

For the prediction of decay concentration profiles of the p-boronophenylalanine (BPA) in blood during BNCT treatment, a method is suggested based on Kohonen neural networks. The results of a model trained with the concentration profiles from the literature are described. The prediction of the model was validated by the leave-one-out method. Its robustness shows that it is mostly independent on small variations. The ability to fit retrospective experimental data shows an uncertainty lower than the two compartment model used previously.

Strategy for the Prediction, Control, and Optimization of the Functional Properties of Food Proteins: Using Statistical and Chemometric Tools

Abstract Proteins have functional properties that govern their behavior in foods during processing, storage, and consumption. Proteins can have high nutritional quality and not have functional properties suitable for incorporation into determined food systems. Furthermore, a desirable functional attribute for an additive may be undesirable for everyone else. This chapter describes the design of a new strategy to predict, control, and optimize the functional parameters of food proteins hydrolyzed or not, using chemometrics tools. The starting material consists of proteins whose functional properties are desired to modify. Functional properties (e.g., emulsifying and foaming properties) can be simultaneously evaluated by an experimental statistical design, response surface graphics, and multiple linear regressions. This strategy expands the applications of food proteins and allows the following facilities: assess interactions between variables of multivariate systems, evaluate the dependence between functional parameters, and optimize the additive production with tailor-made functional properties for different food systems.

Analysis of the environmental liabilities generated by past activities in uranium mining exploitation in the Province of Córdoba, Argentina

Worldwide, old disposal sites of uranium mine and mill wastes are the objects of environmental restoration programs or have already been remediated. This is the case of Los Gigantes in Córdoba, Argentina, where uranium was extracted and processed in the 1980s; a local source of pollution to watercourses was generated as a consequence of disposal of solid and liquid wastes. The present study aims at describing the physicochemical characteristics of the surface watercourses that run across the complex, finding a grouping structure of the sampling sites according to their degree of pollution, and defining the variables that are significant to that grouping. The problem is addressed with both traditional and robust statistics techniques; additional chemometrical tools are also applied. It was found that streams close to the tailings and the pond exhibit lower pH and higher concentration of anions and metals compared to upstream watercourses. As the distance downstream to these pollution sources increases, all physicochemical parameters recover gradually, reaching levels close to background and complying with provincial and national regulations, proving that pollution is locally constrained.