Fernando Alvarez

Increasing attention has been paid to the prediction of earthquakes with data mining techniques during the last decade. Several works have already proposed the use of certain features serving as inputs for supervised classifiers. However, they have been successfully used without any further transformation so far. In this work, the use of principal component analysis to reduce data dimensionality and generate new datasets is proposed. In particular, this step is inserted in a successfully already used methodology to predict earthquakes. Tokyo, one of the cities mostly threatened by large earthquakes occurrence in Japan, is studied. Several well-known classifiers combined with PCA have been used. Noticeable improvement in the results is reported.

En el presente artículo pretendo fijarme en el giro o transformación que se ha producido del concepto de ser y, con él, del concepto mismo de metafísica. La famosa “muerte de Dios”, el giro de la metafísica hacia lo existente y la vida, no han de ser interpretados, como proponía un cierto pensamiento positivistaanalítico y un cierto posmodernismo decadente, como “muerte de la metafísica”, sino como su metamorfosis. El concepto tradicional de “ser”, acuñado por Aristóteles y llevado a su perfección en el medioevo, ha sufrido un cambio radical de sentido: no es ya el ser eterno, inmutable; ni tan sólo una mera physis o naturaleza eterna. Todos los pensadores del siglo XX apelan a una metafísica donde el ser es histórico, frágil, cambiante, y cuya presencia no está asegurada: un ser limítrofe.In the present article I pretend to focus on the turn or transformation that has taken place in the concept of being, and with it, the concept of metaphysics as well. The famous “death of God”, the turn of metaphysics to the existence and life, must not be interpreted, as certain analytical and positivistic thought and a certain decadent postmodernism proposes as “the death of metaphysics” but as its metamorphosis. The traditional concept of ‘being’ coined by Aristotle and brought to its perfection in the Middle Age, has suffered a radical change of meaning: it is not now the eternal and immutable being; not only a sheer physisor eternal nature. All the thinkers of the XX century call for a metaphysics in which being is understood as historical, fragile, changing, and whose presence is not secured: a bordering being

Feature selection is a widely studied technique whose goal is to reduce the dimensionality of the problem by removing irrelevant features. It has multiple benefits, such as improved efficacy, efficiency and interpretability of almost any type of machine learning model. Feature selection techniques may be divided into three main categories, depending on the process used to remove the features known as Filter, Wrapper and Embedded. Embedded methods are usually the preferred feature selection method that efficiently obtains a selection of the most relevant features of the model. However, not all models support an embedded feature selection that forces the use of a different method, reducing the efficiency and reliability of the selection. Neural networks are an example of a model that does not support embedded feature selection. As neural networks have shown to provide remarkable results in multiple scenarios such as classification and regression, sometimes in an ensemble with a model that includes an embedded feature selection, we attempt to embed a feature selection process with a general-purpose methodology. In this work, we propose a novel general-purpose layer for neural networks that removes the influence of irrelevant features. The Feature-Aware Drop Layer is included at the top of the neural network and trained during the backpropagation process without any additional parameters. Our methodology is tested with 17 datasets for classification and regression tasks, including data from different fields such as Health, Economic and Environment, among others. The results show remarkable improvements compared to three different feature selection approaches, with reliable, efficient and effective results.

To accomplish a simple technique to study precognition in the black planarian Girardia dorotocephala, and taking advantage of its strong photophobic response, an experiment was performed where 25 subjects were allowed to crawl in a 10 cm horizontal trough with one LED at each end. Two kinds of trials were carried out: during the experimental trials, and after a 5-min accustoming period, one of the two LEDs would be selected by a true random event generator REG and the experimenter would light it up for other 5 minutes. After one week the same subjects were submitted to the control trials, when exactly the same procedure was followed, although in this case the two LEDs were disconnected from the electric source. Subjects were filmed and their distance towards both LEDs was recorded at the very moment of stimulation or sham stimulation, and in the 1-min period preceding stimulation or sham stimulation, the number of 5-sec periods spent locomoting away from the LED to be lit or sham lit was also recorded. The results showed that during the experimental triaIs, at the very moment of stimulation, the planarians were significantly more distant to the lit LED than to the unlit one. Unexpectedly, during the control trials the subjects were more significantly distant to the sham lit LED than to the other LED. The number of 5-sec periods spent moving away from the lit LED was also higher than that toward the unlit one, the effect being even stronger than during the experimental trials. In both kinds of trials the number of such periods significantly increased as time passed. In conclusion, distance measures provide a reliable technique for studying precognition in planarians, which appear not only able to predict impending stimulation, being also aware of it without real stimulation.

The ability for short-term alarm precognition was explored in Bengalese finches. During the experimental trials, subjects were put individually for 20 minutes in a testing cage and in the last 5 minutes a 15-second video clip of a slowly crawling snake was shown to them in a TFT screen. The video clip was presented at random starting out from 20 possible randomly predetermined options. During the control trials, no snake video clip was shown to the birds. Subjects were filmed, and, in a double-blind fashion, the frequency of their display of alarm was registered 0-3, 3-6, and 6-9 seconds immediately before stimulus presentation and before the same moment of the control trial for each bird. As a second control, behavior frequency was also registered immediately before the 10-minute point after the initiation of the experimental trial. The results showed that the birds reacted to the snake video clip at least 9 seconds before presentation, the frequency of the alarm display during that period being higher than that during both controls. Females and males did not differ significantly in any of the measures. Keywords: alarm display--Bengalese finches--birds--precognition.