Making risk minimization tolerant to label noise

In many applications, the training data, from which one needs to learn a classifier, is corrupted with label noise. Many standard algorithms such as SVM perform poorly in the presence of label noise. In this paper we investigate the robustness of risk minimization to label noise. We prove a sufficient condition on a loss function for the risk minimization under that loss to be tolerant to uniform label noise. We show that the 0-1 loss, sigmoid loss, ramp loss and probit loss satisfy this condition though none of the standard convex loss functions satisfy it. We also prove that, by choosing a sufficiently large value of a parameter in the loss function, the sigmoid loss, ramp loss and probit loss can be made tolerant to nonuniform label noise also if we can assume the classes to be separable under noise-free data distribution. Through extensive empirical studies, we show that risk minimization under the 0-1 loss, the sigmoid loss and the ramp loss has much better robustness to label noise when compared to the SVM algorithm. (C) 2015 Elsevier B.V. All rights reserved.

Wild Ungulate Decision-Making and the Role of Tiny Refuges in Human-Dominated Landscapes

Wildlife conservation in human-dominated landscapes requires that we understand how animals, when making habitat-use decisions, obtain diverse and dynamically occurring resources while avoiding risks, induced by both natural predators and anthropogenic threats. Little is known about the underlying processes that enable wild animals to persist in densely populated human-dominated landscapes, particularly in developing countries. In a complex, semi-arid, fragmented, human-dominated agricultural landscape, we analyzed the habitat-use of blackbuck, a large herbivore endemic to the Indian sub-continent. We hypothesized that blackbuck would show flexible habitat-use behaviour and be risk averse when resource quality in the landscape is high, and less sensitive to risk otherwise. Overall, blackbuck appeared to be strongly influenced by human activity and they offset risks by using small protected patches (similar to 3 km(2)) when they could afford to do so. Blackbuck habitat use varied dynamically corresponding with seasonally-changing levels of resources and risks, with protected habitats registering maximum use. The findings show that human activities can strongly influence and perhaps limit ungulate habitat-use and behaviour, but spatial heterogeneity in risk, particularly the presence of refuges, can allow ungulates to persist in landscapes with high human and livestock densities.

Making consistent contacts to graphene: effect of architecture and growth induced defects

The effects of contact architecture, graphene defect density and metal-semiconductor work function difference on the resistivity of metal-graphene contacts have been investigated. An architecture with metal on the bottom of graphene is found to yield resistivities that are lower, by a factor of four, and most consistent as compared to metal on top of graphene. Growth defects in graphene film were found to further reduce resistivity by a factor of two. Using a combination of method and metal used, the contact resistivity of graphene has been decreased by a factor of 10 to 1200. +/-. 250 Omega mu m using palladium as the contact metal. While the improved consistency is due to the metal being able to contact uncontaminated graphene in the metal on the bottom architecture, lower contact resistivities observed on defective graphene with the same metal are attributed to the increased number of modes of quantum transport in the channel.

Impossibilità di pensare Dio non esistente : il Proslogion di Anselmo d'Aosta come applicazione implicita dei moderni procedimenti di logica aletica

Resumen: La vera intenzione teologica di Anselmo d’Aosta nello scrivere il Proslogion e il vero significato del suo celebre unum argumentum vanno visti nella funzione che la ragione svolge necessariamente all’interno della vita di fede del cristiano. Anselmo, come Tommaso d’Aquino, non sostiene che l’esistenza di Dio sia un “articulus fidei” ma piuttosto uno dei “praeambula fidei”. La ragione naturale ha la certezza che Dio esiste, ancora prima della dimostrazione metafisica, e questo non fa che confermare l’assurdità di pensare che non esista il fondamento reale di tutte le cose esistenti.

Opciones conceptuales, metodológicas y prácticas en la medición de la pobreza : el Informe sobre Pobreza de Irak 2010

Resumen: En el presente trabajo, la autora sostiene que las opciones conceptuales, metodológicas y prácticas involucradas en todo ejercicio de medición de la pobreza tienen consecuencias sobre la estimación de su dimensión, profundidad y estructura. Sin embargo, dichas opciones no están siempre explícitas. El informe de pobreza de Irak de 2010, publicado por el Banco Mundial, representa una excepción en este sentido. La autora propone en este caso, explorar en detalle las opciones subyacentes a la construcción del indicador de bienestar, la línea de pobreza y las medidas de pobreza, señalando la importancia de la explicitación de dichas elecciones.

La moral y los modelos de optimización económica : una reflexión acerca de la antropología en la teoría económica neoclásica

Resumen: Luego de la crisis financiera global de 2008, el blindaje conceptual del paradigma neoclásico, edificado sobre un atractivo formalismo matemático, ha dejado entrever sus falencias. En ese contexto, este artículo plantea la necesidad de revisar sus fundamentos, en especial la concepción antropológica y la metodología que subyace detrás del modelo. El autor analiza al mainstream como un modelo reduccionista de la naturaleza humana, y postula la necesidad de modificar el rumbo de la ciencia económica hacia una dirección personalista, enfocada en el bienestar social de las personas, que considere la importancia de la moral en la toma de decisiones. Asimismo, el economista debería desarrollar su trabajo desde una perspectiva reflexiva y de diálogo con el resto de las ciencias sociales.

Sequential decision making in general state space models

Many problems in control and signal processing can be formulated as sequential decision problems for general state space models. However, except for some simple models one cannot obtain analytical solutions and has to resort to approximation. In this thesis, we have investigated problems where Sequential Monte Carlo (SMC) methods can be combined with a gradient based search to provide solutions to online optimisation problems. We summarise the main contributions of the thesis as follows. Chapter 4 focuses on solving the sensor scheduling problem when cast as a controlled Hidden Markov Model. We consider the case in which the state, observation and action spaces are continuous. This general case is important as it is the natural framework for many applications. In sensor scheduling, our aim is to minimise the variance of the estimation error of the hidden state with respect to the action sequence. We present a novel SMC method that uses a stochastic gradient algorithm to find optimal actions. This is in contrast to existing works in the literature that only solve approximations to the original problem. In Chapter 5 we presented how an SMC can be used to solve a risk sensitive control problem. We adopt the use of the Feynman-Kac representation of a controlled Markov chain flow and exploit the properties of the logarithmic Lyapunov exponent, which lead to a policy gradient solution for the parameterised problem. The resulting SMC algorithm follows a similar structure with the Recursive Maximum Likelihood(RML) algorithm for online parameter estimation. In Chapters 6, 7 and 8, dynamic Graphical models were combined with with state space models for the purpose of online decentralised inference. We have concentrated more on the distributed parameter estimation problem using two Maximum Likelihood techniques, namely Recursive Maximum Likelihood (RML) and Expectation Maximization (EM). The resulting algorithms can be interpreted as an extension of the Belief Propagation (BP) algorithm to compute likelihood gradients. In order to design an SMC algorithm, in Chapter 8 uses a nonparametric approximations for Belief Propagation. The algorithms were successfully applied to solve the sensor localisation problem for sensor networks of small and medium size.