A Survey of Active Learning Algorithms for Supervised Remote Sensing Image Classification

Defining an efficient training set is one of the most delicate phases for the success of remote sensing image classification routines. The complexity of the problem, the limited temporal and financial resources, as well as the high intraclass variance can make an algorithm fail if it is trained with a suboptimal dataset. Active learning aims at building efficient training sets by iteratively improving the model performance through sampling. A user-defined heuristic ranks the unlabeled pixels according to a function of the uncertainty of their class membership and then the user is asked to provide labels for the most uncertain pixels. This paper reviews and tests the main families of active learning algorithms: committee, large margin, and posterior probability-based. For each of them, the most recent advances in the remote sensing community are discussed and some heuristics are detailed and tested. Several challenging remote sensing scenarios are considered, including very high spatial resolution and hyperspectral image classification. Finally, guidelines for choosing the good architecture are provided for new and/or unexperienced user.

Dinàmica de sistemes mesoscòpics amb correlacions quàntiques mitjançant la formulació de de Broglie-Bohm

Per a determinar la dinàmica espai-temporal completa d’un sistema quàntic tridimensional de N partícules cal integrar l’equació d’Schrödinger en 3N dimensions. La capacitat dels ordinadors actuals permet fer-ho com a molt en 3 dimensions. Amb l’objectiu de disminuir el temps de càlcul necessari per a integrar l’equació d’Schrödinger multidimensional, es realitzen usualment una sèrie d’aproximacions, com l’aproximació de Born–Oppenheimer o la de camp mig. En general, el preu que es paga en realitzar aquestes aproximacions és la pèrdua de les correlacions quàntiques (o entrellaçament). Per tant, és necessari desenvolupar mètodes numèrics que permetin integrar i estudiar la dinàmica de sistemes mesoscòpics (sistemes d’entre tres i unes deu partícules) i en els que es tinguin en compte, encara que sigui de forma aproximada, les correlacions quàntiques entre partícules. Recentment, en el context de la propagació d’electrons per efecte túnel en materials semiconductors, X. Oriols ha desenvolupat un nou mètode [Phys. Rev. Lett. 98, 066803 (2007)] per al tractament de les correlacions quàntiques en sistemes mesoscòpics. Aquesta nova proposta es fonamenta en la formulació de la mecànica quàntica de de Broglie– Bohm. Així, volem fer notar que l’enfoc del problema que realitza X. Oriols i que pretenem aquí seguir no es realitza a fi de comptar amb una eina interpretativa, sinó per a obtenir una eina de càlcul numèric amb la que integrar de manera més eficient l’equació d’Schrödinger corresponent a sistemes quàntics de poques partícules. En el marc del present projecte de tesi doctoral es pretén estendre els algorismes desenvolupats per X. Oriols a sistemes quàntics constituïts tant per fermions com per bosons, i aplicar aquests algorismes a diferents sistemes quàntics mesoscòpics on les correlacions quàntiques juguen un paper important. De forma específica, els problemes a estudiar són els següents: (i) Fotoionització de l’àtom d’heli i de l’àtom de liti mitjançant un làser intens. (ii) Estudi de la relació entre la formulació de X. Oriols amb la aproximació de Born–Oppenheimer. (iii) Estudi de les correlacions quàntiques en sistemes bi- i tripartits en l’espai de configuració de les partícules mitjançant la formulació de de Broglie–Bohm.

A linear optimization technique for graph pebbling

Graph pebbling is a network model for studying whether or not a given supply of discrete pebbles can satisfy a given demand via pebbling moves. A pebbling move across an edge of a graph takes two pebbles from one endpoint and places one pebble at the other endpoint; the other pebble is lost in transit as a toll. It has been shown that deciding whether a supply can meet a demand on a graph is NP-complete. The pebbling number of a graph is the smallest t such that every supply of t pebbles can satisfy every demand of one pebble. Deciding if the pebbling number is at most k is NP 2 -complete. In this paper we develop a tool, called theWeight Function Lemma, for computing upper bounds and sometimes exact values for pebbling numbers with the assistance of linear optimization. With this tool we are able to calculate the pebbling numbers of much larger graphs than in previous algorithms, and much more quickly as well. We also obtain results for many families of graphs, in many cases by hand, with much simpler and remarkably shorter proofs than given in previously existing arguments (certificates typically of size at most the number of vertices times the maximum degree), especially for highly symmetric graphs. Here we apply theWeight Function Lemma to several specific graphs, including the Petersen, Lemke, 4th weak Bruhat, Lemke squared, and two random graphs, as well as to a number of infinite families of graphs, such as trees, cycles, graph powers of cycles, cubes, and some generalized Petersen and Coxeter graphs. This partly answers a question of Pachter, et al., by computing the pebbling exponent of cycles to within an asymptotically small range. It is conceivable that this method yields an approximation algorithm for graph pebbling.

Hard instances of algorithms and proof systems

"Vegeu el resum a l'inici del document del fitxer adjunt."

An approximation algorithm for #k-SAT

"Vegeu el resum a l'inici del document del fitxer adjunt"

Credit risk contributions under the Vasicek one-factor model: a fast wavelet expansion approximation

To measure the contribution of individual transactions inside the total risk of a credit portfolio is a major issue in financial institutions. VaR Contributions (VaRC) and Expected Shortfall Contributions (ESC) have become two popular ways of quantifying the risks. However, the usual Monte Carlo (MC) approach is known to be a very time consuming method for computing these risk contributions. In this paper we consider the Wavelet Approximation (WA) method for Value at Risk (VaR) computation presented in [Mas10] in order to calculate the Expected Shortfall (ES) and the risk contributions under the Vasicek one-factor model framework. We decompose the VaR and the ES as a sum of sensitivities representing the marginal impact on the total portfolio risk. Moreover, we present technical improvements in the Wavelet Approximation (WA) that considerably reduce the computational effort in the approximation while, at the same time, the accuracy increases.

Extreme Precipitation Modelling Using Geostatistics and Machine Learning Algorithms

The paper presents an approach for mapping of precipitation data. The main goal is to perform spatial predictions and simulations of precipitation fields using geostatistical methods (ordinary kriging, kriging with external drift) as well as machine learning algorithms (neural networks). More practically, the objective is to reproduce simultaneously both the spatial patterns and the extreme values. This objective is best reached by models integrating geostatistics and machine learning algorithms. To demonstrate how such models work, two case studies have been considered: first, a 2-day accumulation of heavy precipitation and second, a 6-day accumulation of extreme orographic precipitation. The first example is used to compare the performance of two optimization algorithms (conjugate gradients and Levenberg-Marquardt) of a neural network for the reproduction of extreme values. Hybrid models, which combine geostatistical and machine learning algorithms, are also treated in this context. The second dataset is used to analyze the contribution of radar Doppler imagery when used as external drift or as input in the models (kriging with external drift and neural networks). Model assessment is carried out by comparing independent validation errors as well as analyzing data patterns.

Machine learning algorithms for spatial data. Case studies: Environmental pollution, natural hazards, renewable resources

This paper presents general problems and approaches for the spatial data analysis using machine learning algorithms. Machine learning is a very powerful approach to adaptive data analysis, modelling and visualisation. The key feature of the machine learning algorithms is that they learn from empirical data and can be used in cases when the modelled environmental phenomena are hidden, nonlinear, noisy and highly variable in space and in time. Most of the machines learning algorithms are universal and adaptive modelling tools developed to solve basic problems of learning from data: classification/pattern recognition, regression/mapping and probability density modelling. In the present report some of the widely used machine learning algorithms, namely artificial neural networks (ANN) of different architectures and Support Vector Machines (SVM), are adapted to the problems of the analysis and modelling of geo-spatial data. Machine learning algorithms have an important advantage over traditional models of spatial statistics when problems are considered in a high dimensional geo-feature spaces, when the dimension of space exceeds 5. Such features are usually generated, for example, from digital elevation models, remote sensing images, etc. An important extension of models concerns considering of real space constrains like geomorphology, networks, and other natural structures. Recent developments in semi-supervised learning can improve modelling of environmental phenomena taking into account on geo-manifolds. An important part of the study deals with the analysis of relevant variables and models' inputs. This problem is approached by using different feature selection/feature extraction nonlinear tools. To demonstrate the application of machine learning algorithms several interesting case studies are considered: digital soil mapping using SVM, automatic mapping of soil and water system pollution using ANN; natural hazards risk analysis (avalanches, landslides), assessments of renewable resources (wind fields) with SVM and ANN models, etc. The dimensionality of spaces considered varies from 2 to more than 30. Figures 1, 2, 3 demonstrate some results of the studies and their outputs. Finally, the results of environmental mapping are discussed and compared with traditional models of geostatistics.

UOC Algorithms

Aplicació per a iPad a mode de repositori de continguts relacionats amb l'ensenyament d'assignatures d'informàtica.