Evaluating retraining rules for semi-supervised learning in neural network based cursive word recognition

Training a system to recognize handwritten words is a task that requires a large amount of data with their correct transcription. However, the creation of such a training set, including the generation of the ground truth, is tedious and costly. One way of reducing the high cost of labeled training data acquisition is to exploit unlabeled data, which can be gathered easily. Making use of both labeled and unlabeled data is known as semi-supervised learning. One of the most general versions of semi-supervised learning is self-training, where a recognizer iteratively retrains itself on its own output on new, unlabeled data. In this paper we propose to apply semi-supervised learning, and in particular self-training, to the problem of cursive, handwritten word recognition. The special focus of the paper is on retraining rules that define what data are actually being used in the retraining phase. In a series of experiments it is shown that the performance of a neural network based recognizer can be significantly improved through the use of unlabeled data and self-training if appropriate retraining rules are applied.

Evaluation of Three State-of-the-Art Classifiers for Recognition of Activities of Daily Living from Smart Home Ambient Data

Smart homes for the aging population have recently started attracting the attention of the research community. The "health state" of smart homes is comprised of many different levels; starting with the physical health of citizens, it also includes longer-term health norms and outcomes, as well as the arena of positive behavior changes. One of the problems of interest is to monitor the activities of daily living (ADL) of the elderly, aiming at their protection and well-being. For this purpose, we installed passive infrared (PIR) sensors to detect motion in a specific area inside a smart apartment and used them to collect a set of ADL. In a novel approach, we describe a technology that allows the ground truth collected in one smart home to train activity recognition systems for other smart homes. We asked the users to label all instances of all ADL only once and subsequently applied data mining techniques to cluster in-home sensor firings. Each cluster would therefore represent the instances of the same activity. Once the clusters were associated to their corresponding activities, our system was able to recognize future activities. To improve the activity recognition accuracy, our system preprocessed raw sensor data by identifying overlapping activities. To evaluate the recognition performance from a 200-day dataset, we implemented three different active learning classification algorithms and compared their performance: naive Bayesian (NB), support vector machine (SVM) and random forest (RF). Based on our results, the RF classifier recognized activities with an average specificity of 96.53%, a sensitivity of 68.49%, a precision of 74.41% and an F-measure of 71.33%, outperforming both the NB and SVM classifiers. Further clustering markedly improved the results of the RF classifier. An activity recognition system based on PIR sensors in conjunction with a clustering classification approach was able to detect ADL from datasets collected from different homes. Thus, our PIR-based smart home technology could improve care and provide valuable information to better understand the functioning of our societies, as well as to inform both individual and collective action in a smart city scenario.

A systematized workflow for assessing land resources using readily available spatial datasets

Land degradation as well as land conservation maps at a (sub-) national scale are critical for pro-ject planning for sustainable land management. It has long been recognized that online accessible and low-cost raster data sets (e.g. Landsat imagery, SRTM-DEM’s) provide a readily available basis for land resource assessments for developing countries. However, choice of spatial, tempo-ral and spectral resolution of such data is often limited. Furthermore, while local expert knowl-edge on land degradation processes is abundant, difficulties are often encountered when linking existing knowledge with modern approaches including GIS and RS. The aim of this study was to develop an easily applicable, standardized workflow for preliminary spatial assessments of land degradation and conservation, which also allows the integration of existing expert knowledge. The core of the developed method consists of a workflow for rule-based land resource assess-ment. In a systematic way, this workflow leads from predefined land degradation and conserva-tion classes to field indicators, to suitable spatial proxy data, and finally to a set of rules for clas-sification of spatial datasets. Pre-conditions are used to narrow the area of interest. Decision tree models are used for integrating the different rules. It can be concluded that the workflow presented assists experts from different disciplines in col-laboration GIS/RS specialists in establishing a preliminary model for assessing land degradation and conservation in a spatially explicit manner. The workflow provides support when linking field indicators and spatial datasets, and when determining field indicators for groundtruthing.

Prevalence of depressive symptoms in urban middle and high school Hispanic and African American students

This cross-sectional study examined the prevalence of depressive symptoms in urban Hispanic and African American middle and high school students (N=1,292) using data collected from a multi-component, multi-wave violence and substance use intervention program targeted at a large urban school district in Texas. Chi-square analysis was used to examine differences in race/ethnicity, gender, grade level and whether or not a student had been held back/repeated a grade in school. Univariate and multivariate logistic regression were used to analyze the association between depressive symptoms and demographic variables. Being female and being held back/repeating a grade was significantly associated with depressive symptoms in both univariate and multivariate analyses. Overall 16% of the students reported depressive symptoms; Hispanic youth had a higher prevalence of depressive symptoms (16.8%) than the African American youth (14.8%). Minority females and those who had been held back/repeated a grade reported a prevalence of 19.4% and 21.2%, respectively. Further research is needed to understand why Hispanic youth continue to report a higher prevalence of depressive symptoms than other minorities. Additionally research is required to further explore the association between academic performance and depressive symptoms in urban minorities, specifically the effect of being held back/repeating a grade.^

Multi-year high-frequency current and water quality in-situ observations at the Guadiana Estuary

High-frequency data collected continuously over a multiyear time frame are required for investigating the various agents that drive ecological and hydrodynamic processes in estuaries. Here, we present water quality and current in-situ observations from a fixed monitoring station operating from 2008 to 2014 in the lower Guadiana Estuary, southern Portugal (37°11.30' N, 7°24.67' W). The data were recorded by a multi-parametric probe providing hourly records (temperature, salinity, chlorophyll, dissolved oxygen, turbidity, and pH) at a water depth of ~1 m, and by a bottom-mounted acoustic Doppler current profiler measuring the pressure, near-bottom temperature, and flow velocity through the water column every 15 min. The time-series data, in particular the probe ones, present substantial gaps arising from equipment failure and maintenance, which are ineluctable with this type of observations in harsh environments. However, prolonged (months-long) periods of multi-parametric observations during contrasted external forcing conditions are available. The raw data are reported together with flags indicating the quality status of each record. River discharge data from two hydrographic stations located near the estuary head are also provided to support data analysis and interpretation.

Secchi disk data collection for the North Sea and Baltic Sea

This paper presents the results of a Secchi depth data mining study for the North Sea - Baltic Sea region. 40,829 measurements of Secchi depth were compiled from the area as a result of this study. 4.3% of the observations were found in the international data centers [ICES Oceanographic Data Center in Denmark and the World Ocean Data Center A (WDC-A) in the USA], while 95.7% of the data was provided by individuals and ocean research institutions from the surrounding North Sea and Baltic Sea countries. Inquiries made at the World Ocean Data Center B (WDC-B) in Russia suggested that there could be significant additional holdings in that archive but, unfortunately, no data could be made available. The earliest Secchi depth measurement retrieved in this study dates back to 1902 for the Baltic Sea, while the bulk of the measurements were gathered after 1970. The spatial distribution of Secchi depth measurements in the North Sea is very uneven with surprisingly large sampling gaps in the Western North Sea. Quarterly and annual Secchi depth maps with a 0.5° x 0.5° spatial resolution are provided for the transition area between the North Sea and the Baltic Sea (4°E-16°E, 53°N-60°N).

Proposal for the Management of temporal and semantic components of Geographic Information

This poster raises the issue of a research work oriented to the storage, retrieval, representation and analysis of dynamic GI, taking into account The ultimate objective is the modelling and representation of the dynamic nature of geographic features, establishing mechanisms to store geometries enriched with a temporal structure (regardless of space) and a set of semantic descriptors detailing and clarifying the nature of the represented features and their temporality. the semantic, the temporal and the spatiotemporal components. We intend to define a set of methods, rules and restrictions for the adequate integration of these components into the primary elements of the GI: theme, location, time [1]. We intend to establish and incorporate three new structures (layers) into the core of data storage by using mark-up languages: a semantictemporal structure, a geosemantic structure, and an incremental spatiotemporal structure. Thus, data would be provided with the capability of pinpointing and expressing their own basic and temporal characteristics, enabling them to interact each other according to their context, and their time and meaning relationships that could be eventually established

Ontology-based access to sensor data streams

Sensor networks are increasingly becoming one of the main sources of Big Data on the Web. However, the observations that they produce are made available with heterogeneous schemas, vocabularies and data formats, making it difficult to share and reuse these data for other purposes than those for which they were originally set up. In this thesis we address these challenges, considering how we can transform streaming raw data to rich ontology-based information that is accessible through continuous queries for streaming data. Our main contribution is an ontology-based approach for providing data access and query capabilities to streaming data sources, allowing users to express their needs at a conceptual level, independent of implementation and language-specific details. We introduce novel query rewriting and data translation techniques that rely on mapping definitions relating streaming data models to ontological concepts. Specific contributions include: • The syntax and semantics of the SPARQLStream query language for ontologybased data access, and a query rewriting approach for transforming SPARQLStream queries into streaming algebra expressions. • The design of an ontology-based streaming data access engine that can internally reuse an existing data stream engine, complex event processor or sensor middleware, using R2RML mappings for defining relationships between streaming data models and ontology concepts. Concerning the sensor metadata of such streaming data sources, we have investigated how we can use raw measurements to characterize streaming data, producing enriched data descriptions in terms of ontological models. Our specific contributions are: • A representation of sensor data time series that captures gradient information that is useful to characterize types of sensor data. • A method for classifying sensor data time series and determining the type of data, using data mining techniques, and a method for extracting semantic sensor metadata features from the time series.

Algoritmos de agrupación para flujos de datos en entornos centralizados y distribuidos

Los avances en el hardware permiten disponer de grandes volúmenes de datos, surgiendo aplicaciones que deben suministrar información en tiempo cuasi-real, la monitorización de pacientes, ej., el seguimiento sanitario de las conducciones de agua, etc. Las necesidades de estas aplicaciones hacen emerger el modelo de flujo de datos (data streaming) frente al modelo almacenar-para-despuésprocesar (store-then-process). Mientras que en el modelo store-then-process, los datos son almacenados para ser posteriormente consultados; en los sistemas de streaming, los datos son procesados a su llegada al sistema, produciendo respuestas continuas sin llegar a almacenarse. Esta nueva visión impone desafíos para el procesamiento de datos al vuelo: 1) las respuestas deben producirse de manera continua cada vez que nuevos datos llegan al sistema; 2) los datos son accedidos solo una vez y, generalmente, no son almacenados en su totalidad; y 3) el tiempo de procesamiento por dato para producir una respuesta debe ser bajo. Aunque existen dos modelos para el cómputo de respuestas continuas, el modelo evolutivo y el de ventana deslizante; éste segundo se ajusta mejor en ciertas aplicaciones al considerar únicamente los datos recibidos más recientemente, en lugar de todo el histórico de datos. En los últimos años, la minería de datos en streaming se ha centrado en el modelo evolutivo. Mientras que, en el modelo de ventana deslizante, el trabajo presentado es más reducido ya que estos algoritmos no sólo deben de ser incrementales si no que deben borrar la información que caduca por el deslizamiento de la ventana manteniendo los anteriores tres desafíos. Una de las tareas fundamentales en minería de datos es la búsqueda de agrupaciones donde, dado un conjunto de datos, el objetivo es encontrar grupos representativos, de manera que se tenga una descripción sintética del conjunto. Estas agrupaciones son fundamentales en aplicaciones como la detección de intrusos en la red o la segmentación de clientes en el marketing y la publicidad. Debido a las cantidades masivas de datos que deben procesarse en este tipo de aplicaciones (millones de eventos por segundo), las soluciones centralizadas puede ser incapaz de hacer frente a las restricciones de tiempo de procesamiento, por lo que deben recurrir a descartar datos durante los picos de carga. Para evitar esta perdida de datos, se impone el procesamiento distribuido de streams, en concreto, los algoritmos de agrupamiento deben ser adaptados para este tipo de entornos, en los que los datos están distribuidos. En streaming, la investigación no solo se centra en el diseño para tareas generales, como la agrupación, sino también en la búsqueda de nuevos enfoques que se adapten mejor a escenarios particulares. Como ejemplo, un mecanismo de agrupación ad-hoc resulta ser más adecuado para la defensa contra la denegación de servicio distribuida (Distributed Denial of Services, DDoS) que el problema tradicional de k-medias. En esta tesis se pretende contribuir en el problema agrupamiento en streaming tanto en entornos centralizados y distribuidos. Hemos diseñado un algoritmo centralizado de clustering mostrando las capacidades para descubrir agrupaciones de alta calidad en bajo tiempo frente a otras soluciones del estado del arte, en una amplia evaluación. Además, se ha trabajado sobre una estructura que reduce notablemente el espacio de memoria necesario, controlando, en todo momento, el error de los cómputos. Nuestro trabajo también proporciona dos protocolos de distribución del cómputo de agrupaciones. Se han analizado dos características fundamentales: el impacto sobre la calidad del clustering al realizar el cómputo distribuido y las condiciones necesarias para la reducción del tiempo de procesamiento frente a la solución centralizada. Finalmente, hemos desarrollado un entorno para la detección de ataques DDoS basado en agrupaciones. En este último caso, se ha caracterizado el tipo de ataques detectados y se ha desarrollado una evaluación sobre la eficiencia y eficacia de la mitigación del impacto del ataque. ABSTRACT Advances in hardware allow to collect huge volumes of data emerging applications that must provide information in near-real time, e.g., patient monitoring, health monitoring of water pipes, etc. The data streaming model emerges to comply with these applications overcoming the traditional store-then-process model. With the store-then-process model, data is stored before being consulted; while, in streaming, data are processed on the fly producing continuous responses. The challenges of streaming for processing data on the fly are the following: 1) responses must be produced continuously whenever new data arrives in the system; 2) data is accessed only once and is generally not maintained in its entirety, and 3) data processing time to produce a response should be low. Two models exist to compute continuous responses: the evolving model and the sliding window model; the latter fits best with applications must be computed over the most recently data rather than all the previous data. In recent years, research in the context of data stream mining has focused mainly on the evolving model. In the sliding window model, the work presented is smaller since these algorithms must be incremental and they must delete the information which expires when the window slides. Clustering is one of the fundamental techniques of data mining and is used to analyze data sets in order to find representative groups that provide a concise description of the data being processed. Clustering is critical in applications such as network intrusion detection or customer segmentation in marketing and advertising. Due to the huge amount of data that must be processed by such applications (up to millions of events per second), centralized solutions are usually unable to cope with timing restrictions and recur to shedding techniques where data is discarded during load peaks. To avoid discarding of data, processing of streams (such as clustering) must be distributed and adapted to environments where information is distributed. In streaming, research does not only focus on designing for general tasks, such as clustering, but also in finding new approaches that fit bests with particular scenarios. As an example, an ad-hoc grouping mechanism turns out to be more adequate than k-means for defense against Distributed Denial of Service (DDoS). This thesis contributes to the data stream mining clustering technique both for centralized and distributed environments. We present a centralized clustering algorithm showing capabilities to discover clusters of high quality in low time and we provide a comparison with existing state of the art solutions. We have worked on a data structure that significantly reduces memory requirements while controlling the error of the clusters statistics. We also provide two distributed clustering protocols. We focus on the analysis of two key features: the impact on the clustering quality when computation is distributed and the requirements for reducing the processing time compared to the centralized solution. Finally, with respect to ad-hoc grouping techniques, we have developed a DDoS detection framework based on clustering.We have characterized the attacks detected and we have evaluated the efficiency and effectiveness of mitigating the attack impact.

Extending the expressiveness of fuzzy logic languages

En el trabajo que aquí presentamos se incluye la base teórica (sintaxis y semántica) y una implementación de un framework para codificar el razonamiento de la representación difusa o borrosa del mundo (tal y como nosotros, seres humanos, entendemos éste). El interés en la realización de éste trabajo parte de dos fuentes: eliminar la complejidad existente cuando se realiza una implementación con un lenguaje de programación de los llamados de propósito general y proporcionar una herramienta lo suficientemente inteligente para dar respuestas de forma constructiva a consultas difusas o borrosas. El framework, RFuzzy, permite codificar reglas y consultas en una sintaxis muy cercana al lenguaje natural usado por los seres humanos para expresar sus pensamientos, pero es bastante más que eso. Permite representar conceptos muy interesantes, como fuzzificaciones (funciones usadas para convertir conceptos no difusos en difusos), valores por defecto (que se usan para devolver resultados un poco menos válidos que los que devolveríamos si tuviésemos la información necesaria para calcular los más válidos), similaridad entre atributos (característica que utilizamos para buscar aquellos individuos en la base de datos con una característica similar a la buscada), sinónimos o antónimos y, además, nos permite extender el numero de conectivas y modificadores (incluyendo modificadores de negación) que podemos usar en las reglas y consultas. La personalización de la definición de conceptos difusos (muy útil para lidiar con el carácter subjetivo de los conceptos borrosos, donde nos encontramos con que cualificar a alguien de “alto” depende de la altura de la persona que cualifica) es otra de las facilidades incluida. Además, RFuzzy implementa la semántica multi-adjunta. El interés en esta reside en que introduce la posibilidad de obtener la credibilidad de una regla a partir de un conjunto de datos y una regla dada y no solo el grado de satisfacción de una regla a partir de el universo modelado en nuestro programa. De esa forma podemos obtener automáticamente la credibilidad de una regla para una determinada situación. Aún cuando la contribución teórica de la tesis es interesante en si misma, especialmente la inclusión del modificador de negacion, sus multiples usos practicos lo son también. Entre los diferentes usos que se han dado al framework destacamos el reconocimiento de emociones, el control de robots, el control granular en computacion paralela/distribuída y las busquedas difusas o borrosas en bases de datos. ABSTRACT In this work we provide a theoretical basis (syntax and semantics) and a practical implementation of a framework for encoding the reasoning and the fuzzy representation of the world (as human beings understand it). The interest for this work comes from two sources: removing the existing complexity when doing it with a general purpose programming language (one developed without focusing in providing special constructions for representing fuzzy information) and providing a tool intelligent enough to answer, in a constructive way, expressive queries over conventional data. The framework, RFuzzy, allows to encode rules and queries in a syntax very close to the natural language used by human beings to express their thoughts, but it is more than that. It allows to encode very interesting concepts, as fuzzifications (functions to easily fuzzify crisp concepts), default values (used for providing results less adequate but still valid when the information needed to provide results is missing), similarity between attributes (used to search for individuals with a characteristic similar to the one we are looking for), synonyms or antonyms and it allows to extend the number of connectives and modifiers (even negation) we can use in the rules. The personalization of the definition of fuzzy concepts (very useful for dealing with the subjective character of fuzziness, in which a concept like tall depends on the height of the person performing the query) is another of the facilities included. Besides, RFuzzy implements the multi-adjoint semantics. The interest in them is that in addition to obtaining the grade of satisfaction of a consequent from a rule, its credibility and the grade of satisfaction of the antecedents we can determine from a set of data how much credibility we must assign to a rule to model the behaviour of the set of data. So, we can determine automatically the credibility of a rule for a particular situation. Although the theoretical contribution is interesting by itself, specially the inclusion of the negation modifier, the practical usage of it is equally important. Between the different uses given to the framework we highlight emotion recognition, robocup control, granularity control in parallel/distributed computing and flexible searches in databases.

Mining mobile apps reviews to support release planning

The mobile apps market is a tremendous success, with millions of apps downloaded and used every day by users spread all around the world. For apps’ developers, having their apps published on one of the major app stores (e.g. Google Play market) is just the beginning of the apps lifecycle. Indeed, in order to successfully compete with the other apps in the market, an app has to be updated frequently by adding new attractive features and by fixing existing bugs. Clearly, any developer interested in increasing the success of her app should try to implement features desired by the app’s users and to fix bugs affecting the user experience of many of them. A precious source of information to decide how to collect users’ opinions and wishes is represented by the reviews left by users on the store from which they downloaded the app. However, to exploit such information the app’s developer should manually read each user review and verify if it contains useful information (e.g. suggestions for new features). This is something not doable if the app receives hundreds of reviews per day, as happens for the very popular apps on the market. In this work, our aim is to provide support to mobile apps developers by proposing a novel approach exploiting data mining, natural language processing, machine learning, and clustering techniques in order to classify the user reviews on the basis of the information they contain (e.g. useless, suggestion for new features, bugs reporting). Such an approach has been empirically evaluated and made available in a web-­‐based tool publicly available to all apps’ developers. The achieved results showed that the developed tool: (i) is able to correctly categorise user reviews on the basis of their content (e.g. isolating those reporting bugs) with 78% of accuracy, (ii) produces clusters of reviews (e.g. groups together reviews indicating exactly the same bug to be fixed) that are meaningful from a developer’s point-­‐of-­‐view, and (iii) is considered useful by a software company working in the mobile apps’ development market.