Search

Researchers are defining new types of interactions between humans and machine learning algorithms generically called human-in-the-loop machine learning. Depending on who is in control of the learning process, we can identify: active learning, in which the system remains in control; interactive machine learning, in which there is a closer interaction between users and learning systems; and machine teaching, where human domain experts have control over the learning process. Aside from control, humans can also be involved in the learning process in other ways. In curriculum learning human domain experts try to impose some structure on the examples presented to improve the learning; in explainable AI the focus is on the ability of the model to explain to humans why a given solution was chosen. This collaboration between AI models and humans should not be limited only to the learning process; if we go further, we can see other terms that arise such as Usable and Useful AI. In this paper we review the state of the art of the techniques involved in the new forms of relationship between humans and ML algorithms. Our contribution is not merely listing the different approaches, but to provide definitions clarifying confusing, varied and sometimes contradictory terms; to elucidate and determine the boundaries between the different methods; and to correlate all the techniques searching for the connections and influences between them.

Financiado para publicación en acceso aberto: Universidade da Coruña/CISUG [Abstract] In this study, we analyze the capability of several state of the art machine learning methods to predict whether patients diagnosed with CoVid-19 (CoronaVirus disease 2019) will need different levels of hospital care assistance (regular hospital admission or intensive care unit admission), during the course of their illness, using only demographic and clinical data. For this research, a data set of 10,454 patients from 14 hospitals in Galicia (Spain) was used. Each patient is characterized by 833 variables, two of which are age and gender and the other are records of diseases or conditions in their medical history. In addition, for each patient, his/her history of hospital or intensive care unit (ICU) admissions due to CoVid-19 is available. This clinical history will serve to label each patient and thus being able to assess the predictions of the model. Our aim is to identify which model delivers the best accuracies for both hospital and ICU admissions only using demographic variables and some structured clinical data, as well as identifying which of those are more relevant in both cases. The results obtained in the experimental study show that the best models are those based on oversampling as a preprocessing phase to balance the distribution of classes. Using these models and all the available features, we achieved an area under the curve (AUC) of 76.1% and 80.4% for predicting the need of hospital and ICU admissions, respectively. Furthermore, feature selection and oversampling techniques were applied and it has been experimentally verified that the relevant variables for the classification are age and gender, since only using these two features the performance of the models is not degraded for the two mentioned prediction problems. This research has been supported by GAIN (Galician Innovation Agency) and the Regional Ministry of Economy, Employment and Industry, Xunta de Galicia grant COV20/00604 through the ERDF Funds. Also, it has been possible thanks to the support of the Xunta de Galicia (Dirección Xeral de Saúde Pública) by providing the anonymous patient data. Also, it has been supported by the Xunta de Galicia (Grant ED431C 2018/34 and IN845D 2020/26 of the Axencia Galega de Innovación) with European Union ERDF funds. CITIC, as Research Center accredited by Galician University System, is funded by Consellería de Cultura, Educación e Universidades from Xunta de Galicia, supported in an 80% through ERDF Funds, ERDF Operational Programme Galicia 2014-2020, and the remaining 20% by Secretaría Xeral de Universidades (Grant ED431G 2019/01). Finally, we would also like to thank Prof. Ricardo Cao, as Chairman of the Committee of Experts for Mathematical Action against Coronavirus, for his kind request to collaborate in this project Xunta de Galicia; COV20/00604 Xunta de Galicia; ED431C 2018/34 Xunta de Galicia; IN845D 2020/26 Xunta de Galicia; ED431G 2019/01

La sociosemiótica de la transformación universitaria es un constructo que facilita el entendimiento de la problemática de asignación de sentidos atribuidos por la sociedad venezolana al proceso de transformación universitaria en el período 2010 – 2012. Sus hipótesis emergen del procesamiento de 20 discursos utilizando el método de análisis interdiscursivos y las técnicas: codificación axial, selectiva, comparación teórica constante y distante de la teoría fundamentada con apoyo del software Atla.ti. Ello plantea variadas formas de interpretar la discursividad sobre la transformación que originan los interrogantes de la presente investigación entre ellas: ¿por qué en discursos con sentidos semejantes la transformación se comprendería como Acuerdos Escalares Progresivos?, ¿por qué con sentidos opuestos, la misma se comprendería como una Acción colectiva Focalizadora?, ¿por qué con sentidos contradictorios o antagónicos se comprendería como Diálogos Pactuales? Esta investigación tiene como objetivo fundamentar desde los aportes de la complejidad dichos supuestos discursivos, para ello se aplicó el método analítico en tres fases: i) Identificación de los aportes de la Complejidad equivalentes con la Sociosemiótica de la transformación venezolana; ii) Comparación de conceptos claves entre las formas de interpretar la discursiva de la Sociosemiótica y la Complejidad; iii) Diseño desde la complejidad de los argumentos que fundamentan las formas de interpretar la discursividad sobre la transformación universitaria venezolana. De forma que los principios y procesos complejos tales como: auto-eco-organización, auto-referente, principio de identidad, bucle tetralógico, la auto-producción y auto-eco-organización; la dialógica retroactiva positiva regulativa y las inter-retro-acciones culturales, fundamentan dichos supuestos. Además la complejidad se recomienda para comprender la co-interdependencia entre los sentidos de: modernización, reforma y cambio radical de la transformación universitaria venezolana.

El presente ensayo aborda cuáles son los desafíos de la educación a distancia en tiempo de confinamiento por la enfermedad Coronavirus, tanto para el facilitador como para el aprendiz, en un contexto de pandemia mundial resultado de un virus letal como el SARS-CoV-2 al cual aún no se ha detectado tratamiento ni vacuna que lo elimine. El desarrollo metodológico para dar respuesta a la interrogante planteada, se centró en una investigación documental, en la Web y revistas arbitradas, así como una exploración de opinión a docentes de varias universidades nacionales y a estudiantes del 6to semestre de turismo de la UNEFA. Esta indagación generó como principal desafío la frustración generada por poco el tiempo disponible, el miedo a la tecnología, entre otros motivos, mientras que para los especialistas, resultó ser la desmotivación de docentes y estudiantes. Interpretando los resultados a la luz de diversas teorías se demuestra que el desafío real es desprenderse del paradigma de la presencialidad y de la marginalidad a la virtualidad. Sugiriéndose al respecto el uso de estrategias que posean diferentes lógicas “rizomáticas” tales como: divergente, crítica creativa e interactiva colaborativa.

This chapter follows the steps undertaken by other researchers in the field of knowledge engineering in Medicine. The material here presented is concerned with four key issues: The nature of the medical knowledge, the characteristics of the reasoning processes in medicine, the automatic acquisition of medical knowledge, and the effective handling of uncertain medical knowledge. The chapter reviews first categorical logic models and Bayesian methods. From a first conclusion on inherent uncertainty of medical decision making and reasoning, a vector representation of medical knowledge is proposed. This vector representation facilitates automation of the processes involved in the development of medical decision support systems. The use of contingency tables for automatic and objective knowledge acquisition is also proposed. To facilitate the understanding of the presented material, we have tried to illustrate each statement, idea, proposal or approach, with examples taken from the literature. Finally, the chapter concludes with an analysis of the possibilities of the overall method. Major potential contributions are justified, explained and discussed. The mentioned analysis focus on the application of the proposed approach on a simplified clinical case derived from the experience of the authors in the domain of the Sleep Medicine. Finally, the chapter concludes with a discussion, and with the establishment of the required conclusions.

Background Clinical diagnosis of sleep disorders relies on the polysomnographic test to examine the neurophysiological markers of the sleep process. In this test, the recording of the electroencephalographic activity and the submental electromyogram is the source of the analysis for the detection of electroencephalographic arousals. The identification of these events is important for the evaluation of the sleep continuity because they cause the fragmentation of the normal sleep process. This work proposes a new technique for the automatic detection of arousals in polysomnographic recordings, presenting a non-computationally complex method with the idea of providing an easy integration with other algorithms. Methods The proposed algorithm combines different well-known signal analysis solutions to identify relevant arousal patterns with special emphasis on robustness and artifacts tolerance. It is a multistage method that after obtaining an initial set of events, improves the detection finding common EEG arousal patterns. Finally, false positives are discarded after examining each candidate within the context of clinical definitions. Results 22 polysomnographic recordings from real patients were used to validate the method. The results obtained were encouraging, achieving a precision value of 0.86 and a F 1 score value of 0.79. When compared with the gold standard, the method achieves a substantial agreement (Kappa coefficient of 0.78), with an almost perfect agreement with ten recordings. Conclusions The algorithm designed achieved encouraging results and shows robust behavior in presence of signal artifacts. Its low-coupled design allows its implementation on different development platforms, and an easy combination with other methods.

The final publication is available at ScienceDirect via http://dx.doi.org/10.1016/j.ins.2015.08.022 [Abstract] The main objective of this work is to obtain a method that achieves the best accuracy results with a low false positive rate in the classification of K-complexes, a kind of transient waveform found in the Electroencephalogram. With this in mind, the capabilities of several machine learning techniques were tried. The inputs for the models were a set of features based on amplitude and duration measurements obtained from waveforms to be classified. Among all the classifiers tested, the Support Vector Machine obtained the best results with an accuracy of 88.69%. Finally, to enhance the generalization capabilities of the classifiers, while at the same time discarding the existing irrelevant features, feature selection methods were employed. After this process, the classification performance was significantly improved. The best result was obtained applying a correlation-based filter, achieving a 91.40% of accuracy using only 36% of the total input features. Xunta de Galicia; 09SIN003CT Ministerio de Economía y Competitividad; TIN2013-40686P Ministerio de Economía y Competitividad; TIN2012-37954 Xunta de Galicia; GRC2014/35

Sleep spindles are transient waveforms and one of the key features that contributes to sleep stages assessment. Due to the large number of sleep spindles appearing on an overnight sleep, automating the detection of this waveforms is desirable. This paper presents a comparative study over the sleep spindle classification task involving the discrete wavelet decomposition of the EEG signal, and seven different classification algorithms. The main goal was to find a classifier that achieves the best performance. The results reported that Random Forest stands out over the rest of models, achieving an accuracy value of \(94.08 \pm 2.8\) and \(94.08 \pm 2.4\,\%\) with the symlet and biorthogonal wavelet families.

The success of any Intrusion Detection System (IDS) lies in the selection of a set of significant features, that can be quantitative or qualitative, taken out from a network traffic data stream. The machine learning methods provide potential solutions for the IDS problem. However, most of these methods used for classification are not able to handle symbolic attributes directly. In this paper, three methods for symbolic features conversion - indicator variables, conditional probabilities and the Separability Split Value method - are contrasted with the arbitrary conversion method, all of them applied to an intrusion detection problem, the KDD Cup 99 data set. In particular, three classification methods were subsequently applied to the dataset: a one-layer feedforward neural network, a support vector machine and a multilayer feedforward neural network. The results obtained demonstrate that the three conversion methods improve the prediction ability of the classifiers utilized, with respect to the arbitrary and commonly used assignment of numerical values.

This paper presents a comparative study over the respiratory pattern classification task involving three missing data imputation techniques, and four different machine learning algorithms. The main goal was to find a classifier that achieves the best accuracy results using a scalable imputation method in comparison to the method used in a previous work of the authors. The results obtained show that the Self-organization maps imputation method allows any classifier to achieve improvements over the rest of the imputation methods, and that the Feedforward neural network classifier offers the best performance regardless the imputation method used.

An effective diagnosis of the sleep apnea syndrome (SAS) is based on a contextual analysis of the patient's polysomnograph, consisting of simultaneously recording electrophysiological and pneumological signals during a night's sleep. Currently, the prevalence of this disorder has caused an increase in the demand for specialist clinical assistance and sleep units. As in other areas of medicine, the volume of clinical data that has to be processed is enormous, which justifies the construction of computerized decisionmaking tools that partially automate these routine tasks. Our system, SAMOA, belongs to this category of help tools, being an automatic SAS diagnostic system that incorporates both conventional programming and artificial intelligence techniques. This article describes the most important aspects of the temporal data management in the different analysis processes and the final correlation of all the symbolic information generated by the different cooperative modules.

The sleep apnea syndrome (SAS) is a respiratory disorder, which is characterised by the occurrence of five or more apneic events (apnea or hypopnea) per hour of sleep. Diagnosis of the SAS is a process that is markedly heuristic by nature, in that doctors handle information that is both numerical and symbolic, and employ qualitative descriptive terminology. An expert draws up a contextualised clinical interpretation that relates a patient's sleep process and respiratory physiology, involving a detailed analysis of the polysomnograph corresponding to a night's sleep. This task, implying a great deal of work on the part of clinical staff and a high economic cost, can in fact be partially automated. Our paper describes a modular system based on artificial intelligence techniques that provides an individual SAS diagnosis on the basis of a patient's polysomnograph. The main tasks of our system are the identification and classification of respiratory events, the construction of the patient's hypnogram and the correlation of all the information obtained so as to arrive at a final diagnosis with respect to the existence of the syndrome. Finally our article presents and discusses the results obtained following a preliminary validation of the developed system.

doi: 10.2174/1874431101408010001 This work deals with the development of an intelligent approach for clinical decision making in the diagnosis of the Sleep Apnea/Hypopnea Syndrome, SAHS, from the analysis of respiratory signals and oxygen saturation in arterial blood, SaO2. In order to accomplish the task the proposed approach makes use of different artificial intelligence techniques and reasoning processes being able to deal with imprecise data. These reasoning processes are based on fuzzy logic and on temporal analysis of the information. The developed approach also takes into account the possibility of artifacts in the monitored signals. Detection and characterization of signal artifacts allows detection of false positives. Identification of relevant diagnostic patterns and temporal correlation of events is performed through the implementation of temporal constraints.

Objective: The involuntary periodic repetition of respiratory pauses or apneas constitutes the sleep apnea-hypopnea syndrome (SAHS). This paper presents two novel approaches for sleep apnea classification in one of their three basic types: obstructive, central and mixed. The goal is to improve the classification accuracy obtained in previous works. Materials and methods: Both models are based on a combination of classifiers whose inputs are the coefficients obtained by a discrete wavelet decomposition applied to the raw samples of the apnea in the thoracic effort signal. The first model builds adaptive data-dependent committees, subsets of classifiers that are specifically selected for each input pattern. The second one uses a new classification approach based on the characteristics each type of apnea presents in different segments of the apnea. This model is based on the Error Correcting Output Code and its input coefficients were determined by a feature selection method (SVM Recursive Feature Elimination). In order to train and test the systems, 120 events from six different patients were used. The true error rate was estimated using 10 different simulations of a 10-fold cross validation. Results: The mean test accuracy, obtained over the test set was 85.20% ± 1.25 for the first model and 90.27% ± 0.79 for the second one. Conclusions: The proposed classifiers surpass, up to the author's knowledge, other previous results. Moreover, the results achieved are correctly enough to obtain a reliable diagnosis of SAHS, taking into account the average duration of a sleep test and the number of apneas presented for a patient who suffers SAHS.

Over the last two decades in southern Europe, more than 10 million hectares of forest have been damaged by fire. Due to the costs and complications of fire-fighting a number of technical developments in the field have been appeared in recent years. This paper describes a system developed for the region of Galicia in NW Spain, one of the regions of Europe most affected by fires. This system fulfills three main aims: it acts as a preventive tool by predicting forest fire risks, it backs up the forest fire monitoring and extinction phase, and it assists in planning the recuperation of the burned areas. The forest fire prediction model is based on a neural network whose output is classified into four symbolic risk categories, obtaining an accuracy of 0.789. The other two main tasks are carried out by a knowledge-based system developed following the CommonKADS methodology. Currently we are working on the trail of the system in a controlled real environment. This will provide results on real behaviour that can be used to fine-tune the system to the point where it is considered suitable for installation in a real application environment., This research has been funded by the European Regional Development Fund (ERDF) project 1FD97-1122-C06-01 and by the Spanish Comisión Interministerial de Ciencia y Tecnologı́a (CICYT) under project REN-2001-3216-CO4-01.

Automation of the medical diagnosis of the Sleep Apnea Syndrome(SAS) requires an intelligent analysis of the pneumological and neurophysiological signals of the patient that combines both conventional and Artificial Intelligence techniques in order to detect respiratory abnormalities and construct a hypnogram for the patient, and a process of temporal fusion and correlation between the signals for both a correct classification of the apneic events within a sleep stage framework, and to explain the occurrence of abnormal sleep patterns as a consequence of these events. In this article, the most important aspects of the analysis and information integration processes are described and the preliminary validation results obtained are discussed.

In this paper, the authors present a work-in-process neural model trained to symbolically characterize the numeric parameters of critical patients in Intensive Care Units (ICUs). The nets were designed to assign the state {very high, high, normal, low, very low} of the main haemodinamic and respiratory parameters based on the particular context of the patient. This symbolic processing allows the use of an heuristic, general module, which prescribes ventilation and oxygenation therapies and shows the diagnoses corresponding to the symbolic labels of those variables.