Learning weighted metrics to minimize nearest-neighbor classification error

In order to optimize the accuracy of the nearest-neighbor classification rule, a weighted distance is proposed, along with algorithms to automatically learn the corresponding weights. These weights may be specific for each class and feature, for each individual prototype, or for both. The learning algorithms are derived by (approximately) minimizing the leaving-one-out classification error of the given training set. The proposed approach is assessed through a series of experiments with UCI/STATLOG corpora, as well as with a more specific task of text classification which entails very sparse data representation and huge dimensionality. In all these experiments, the proposed approach shows a uniformly good behavior, with results comparable to or better than state-of-the-art results published with the same data so far.     

Parallel Partial Stabilizing Algorithms for Large Linear Control Systems

In this paper we present parallel algorithms for stabilizing large linear control systems on multicomputers. Our algorithms first separate the stable part of the linear control system and then compute a stabilizing feedback for the unstable part. Both stages are solved by means of the matrix sign function which presents a high degree of parallelism and scalability.The experimental results on an IBM SP2 platform show the performance of our approach.     

IdentityRank: Named entity disambiguation in the news domain

News companies produce news items that describe events that happen in the world. These news items usually contain mentions to persons, organizations, locations and other types of named entities that are involved in the events being described. These named entities may have an ambiguous meaning, which impacts the performance of free-text information retrieval systems. In this paper the IdentityRank algorithm, designed to address the problem of named entity disambiguation in news items, is described. It has been developed as part of the EU-funded project News Engine Web Services (NEWS) and is specifically designed to operate within the editorial environment of a news company. The algorithm was implemented and evaluated using several corpora of actual news items, achieving an average accuracy of around 96%.     

Feedback-reversibility and reachability of linear impulsive systems

For a linear impulsive system, the set of states that are reachable from the origin when the initial time, impulse times, and final time are fixed is contained in an invariant subspace determined by the system data. It is known that reversibility of the system is sufficient to yield, for a specified initial time, the existence of some impulse time set and final time for which the reachable set equals the invariant subspace. In this paper, we relax the reversibility requirement and present a condition that is necessary as well as sufficient under which this property holds. This new condition involves the property of achieving reversibility via feedback and admits an explicit geometric characterization. Moreover, this feedback-reversibility property only needs to hold for the subsystem defined as the full system restricted to the invariant subspace. We further show that feedback-reversibility of the restricted system ensures that the reachable set equals the invariant subspace for almost any impulse time set and final time for which the number of impulse times contained in the underlying time interval exceeds a lower bound.     

A hybrid content-based and item-based collaborative filtering approach to recommend TV programs enhanced with singular value decomposition

With the advent of new cable and satellite services, and the next generation of digital TV systems, people are faced with an unprecedented level of program choice. This often means that viewers receive much more information than they can actually manage, which may lead them to believe that they are missing programs that could likely interest them. In this context, TV program recommendation systems allow us to cope with this problem by automatically matching user’s likes to TV programs and recommending the ones with higher user preference.This paper describes the design, development, and startup of queveo.tv: a Web 2.0 TV program recommendation system. The proposed hybrid approach (which combines content-filtering techniques with those based on collaborative filtering) also provides all typical advantages of any social network, such as supporting communication among users as well as allowing users to add and tag contents, rate and comment the items, etc. To eliminate the most serious limitations of collaborative filtering, we have resorted to a well-known matrix factorization technique in the implementation of the item-based collaborative filtering algorithm, which has shown a good behavior in the TV domain. Every step in the development of this application was taken keeping always in mind the main goal: to simplify as much as possible the user task of selecting what program to watch on TV.     

WatchBot: A building maintenance and surveillance system based on autonomous robots

There is a growing interest in the use of intelligent technologies in new buildings. An intelligent system should be designed in a manner that allows minimum human intervention during daily operation. However, large buildings need substantial manpower for maintenance, management as well as surveillance to ensure a quality environment for the occupants. This paper describes a multi-robot system for building maintenance and surveillance applications over the Internet. Each robot can handle autonomously some daily maintenance and surveillance routine tasks, although remote control of the robots via the Internet or intranet is also possible. Apart from the user-started and scheduled tasks, the robots can also execute tasks to handle alarms triggered by the building automation system (BAS). Robots are connected to the central management office via a local area network. This paper shows the advantages of using mobile robots for building maintenance and surveillance tasks by improving efficiency and reducing manpower. The system was developed with the Robotics Integrated Development Environment (RIDE) and was tested intensively in different environments.     

Neural network and crop residue index multiband models for estimating crop residue cover from Landsat TM and ETM+ images

Crop residues on the soil surface provide not only a barrier against water and wind erosion, but they also contribute to improving soil organic matter content, infiltration, evaporation, temperature, and soil structure, among others. In Argentina, soybean (Glycine max (L.) Merill) and corn (Zea mays L.) are the most important crops. The objective of this work was to develop and evaluate two different types of model for estimating soybean and corn residue cover: neural networks (NN) and crop residue index multiband (CRIM) index, from Landsat images. Data of crop residue were acquired throughout the summer growing season in the central plains of Córdoba (Argentina) and used for training and validating the models. The CRIM, a linear mixing model of composite soil and residue, and the NN design, included reflectance and digital numbers from a combination of different TM bands to estimate the fractional residue cover. The results show that both methodologies are appropriate for estimating the residue cover from Landsat data. The best developed NN model yielded R² = 0.95 when estimating soybean and corn residue cover fraction, whereas the best fit using CRIM yielded R² = 0.87; in addition, this index is dependent on the soil and residue lines considered.     

Empirical models for estimating daily surface water vapour pressure,air temperature,and humidity using MODIS and spatiotemporal variables. Applications to peninsular Spain

In this article, we present empirical models for estimating daily surface water vapour pressure (e₀), air temperature (T_a), and relative humidity (RH) over cloud-free land areas in peninsular Spain using Moderate Resolution Imaging Spectroradiometer (MODIS) and spatiotemporal variables. The models were obtained and validated using daily mean, maximum, and minimum e₀, T_a, and RH data (year 2010) from 331 ground-level meteorological stations and the diurnal Terra-MODIS data in peninsular Spain, but the methodology can easily be extrapolated and used to obtain algorithms for other regions around the world. The best e₀ models are based on total precipitable water (W) estimations obtained by MOD05 or IMAPP WVNIR products and the spatiotemporal variables of longitude (λ), distance to the coast (d_coast), and Julian day (JD). Other models based on Sobrino’s W algorithm or on Recondo’s e₀ algorithm for Asturias (in northern Spain) were also tested. The best T_a models are based on land surface temperature (LST) obtained by the MOD11 LST or IMAPP LST products and on other remote-sensing variables, such as W and the normalized difference vegetation index (NDVI), and the spatiotemporal variables λ, JD, and height (h). Models based on Sobrino’s LST algorithms were also tested. RH can be derived directly from e₀ and T_a or from models similar to those used to obtain e₀ and T_a. Models based on the NASA standard products MOD05 and MOD11 LST are slightly better than those based on IMAPP products, but the advantage of IMAPP products for our purposes is that they can be generated in almost real time from the data obtained by the MODIS antenna at the University of Oviedo. IMAPP models obtain the following: R² = 0.83-0.79-0.70 and RSE = 1.62-1.59-1.76 hPa for e_0mean, e_0max, and e_0min; and R² = 0.91-0.91-0.80 and RSE = 1.96-2.25-3.00 K for T_mean, T_max, and T_min. Worse results are obtained for RH: R² = 0.49-0.39 and RSE = 7.21-9.75% for RH_min and RH_mean, with no correlation found for RH_max. Model validations yield R² and RSE values similar to those obtained in the models, with an RMSD = 1.86-1.99-2.21 hPa for e_0mean, e_0max, and e_0min; an RMSD = 2.05-2.40-2.95 K for T_mean, T_max, and T_min; and RMSD = 8-11% for RH_min and RH_mean. The bias is small in all cases: <0.2 hPa for e₀, ≤0.1 K for T_a, and ≤ |1|% for RH. From the results of this article, we propose substituting the traditionally used RH variable with the e₀ variable to be used as meteorological variable in environmental risk models such as, for example, fire risk models.