159 research outputs found

    Hash Kernels for Structured Data

    No full text
    We propose hashing to facilitate efficient kernels. This generalizes previous work using sampling and we show a principled way to compute the kernel matrix for data streams and sparse feature spaces. Moreover, we give deviation bounds from the exact kernel matrix. This has applications to estimation on strings and graphsQinfeng Shi, James Petterson, Gideon Dror, John Langford, Alex Smola and S.V.N. Vishwanatha

    Hash kernels

    No full text
    We propose hashing to facilitate efficient kernels. This generalizes previous work using sampling and we show a principled way to compute the kernel matrix for data streams and sparse feature spaces. Moreover, we give deviation bounds from the exact kernel matrix. This has applications to estimation on strings and graphs.Qinfeng Shi, James Petterson, Gideon Dror, John Langford, Alex Smola, Alex Strehl and Vishy Vishwanatha

    Delta-type senior civil service for the 21st century

    No full text
    O artigo estuda um aspecto da reforma do Estado que, segundo o autor, tem sido sistematicamente negligenciado pelas atuais propostas que focalizam o modelo da administração gerencial. Trata-se das funções vitais do governo de tomar decisões críticas e adotar políticas diante das mudanças provocadas pela revolução global. Segundo Dror, as tarefas de alto comando (high-order tasks) de definir trajetórias e as novas formas de governança exigem um ajuste significativo do governo central. Este ajuste refere-se, principalmente, à concepção e ao desenvolvimento de um novo padrão de funcionários do primeiro escalão público, o qual contribuiria com conhecimento e perspectivas para enfrentar as tarefas de alto comando. O autor estabelece uma tipologia para caraterizar a evolução do perfil do serviço público, marcando suas fases históricas: a) tipo alpha (status atribuído, fusão de papéis políticos e administrativos); b) tipo beta (compra de cargos governamentais) e c) tipo gamma (quase profissionalismo). O novo funcionário sênior, do tipo delta, se concentraria nas questões de ordem estratégica, deixando as funções gerenciais para servidores do tipo gamma e para os serviços técnicos. Após uma breve análise, Dror conclui que o funcionalismo público de primeiro escalão, na maioria dos países (com exceção de alguns países do Sudeste Asiático), encontra-se obsoleto, com base profissional inadequada e capacidade insuficiente para lidar com escolhas críticas.El artículo estudia un aspecto de la reforma del Estado que, según el autor, ha sido sistematicamente negligenciado por las actuales propuestas que enfocan el modelo de la administración gerencial. Se trata de las funciones vitales del gobierno de tomar decisiones críticas y adoptar políticas delante de los cambios provocados por la revolución global. Según Dror, las tareas de alto comando (high-order tasks) de definir trayectorias y las nuevas formas de governanza exigen un ajuste significativo del gobierno central. Este ajuste se refiere, principalmente, a la concepción y al desarollo de un nuevo padrón de funcionarios de primera jerarquía, lo que contribuiria con conocimiento y perspectivas para afrontar las tareas de alto comando. El autor establece una tipologia para caracterizar la evolución del perfil del servicio público, marcando sus fases historicas: a) tipo alpha: (status atribuido, fusión de papeles políticos y administrativos); b) tipo beta: (compra de cargos gubernamentales) y c) tipo gamma: (casi profesionalismo). El nuevo funcionario senior, del tipo delta, se concentraria en las cuestiones de orden estratégica, quedándose las funciones gerenciales para los servidores del tipo gamma y para los servicios técnicos. Tras un breve análisis, Dror concluye que el funcionarismo público de primera jerarquía, en la mayoria de los países (excepto algunos países del Sudeste Asiático), se encuentra obsoleto, con bases professionales inadecuadas y capacidad insuficiente para lidar con opciones críticas.The article alludes to one aspect of the State reform that, according to the author, has been systematically neglected by present proposals focused on the management administration model. Government’s vital functions are to make critical decisions and to adopt policies when facing changes brought about by the global revolution. According to Dror, the high-order tasks of defining trajectories and new forms of governance demand a significant adjustment of central governments. This adjustment refers mainly to the creation and development of a new profile of senior civil servants, who would present their contribution to high-order tasks with more knowledge and perspective. The author establishes a categorization to illustrate the evolution of the civil servant profile, highlighting its historical phases: a) alpha type (attributed status, fusion of political and administrative roles); b) beta type (purchase of governmental positions); and c) gamma type (quasi-professionalism). The new delta-type senior civil servant should concentrate on strategical issues, leaving managerial functions to gamma-type civil servants and to technical services. After a brief analysis, Dror declares that the senior civil service in most countries (exceptions to be made to some East-Asian countries) is obsolete, lacking adequate professional basis and capacity to deal with critical decisions.Número padronizado: v. 48, n. 2 (1997) Revista do Serviço Público - RSP, ano 48, n. 2, p. 5-26Administração PúblicaISSN impresso: 0034-9240ISSN eletrônico: 2357-801

    Momentum reconstruction of particles in the forward muon trigger system of the ATLAS detector

    No full text
    We devise a feed forward neural network which identifies the charge and momentum of muons in the forward trigger system of the ATLAS detector. We use second order learning methods to train the network on a set comprising a few thousand simulated events over a wide range of energies and transverse momenta. The network produces an unbiased estimate of the particles' momentum and charge. On a test set, the charge is correctly identified in 94% of the events, and the mean relative error in transverse momentum is about 1:9%. [email protected] y [email protected] z [email protected] 1 Introduction The Large Hadron Collider (LHC) is an accelerator which will bring proton beams into head-on collisions at energies of 14 TeV and luminosity L = 10 34 cm \Gamma2 s \Gamma1 , higher than ever before. The high luminosity produces an event flux of about 1 GHz. In such an environment there is a unique importance to an efficient triggering system operating within the short bun..

    Result Analysis of the NIPS 2003 Feature Selection Challenge

    No full text
    The NIPS 2003 workshops included a feature selection competition organized by the authors. We provided participants with five datasets from different application domains and called for classification results using a minimal number of features. The competition took place over a period of 13 weeks and attracted 78 research groups. Participants were asked to make on-line submissions on the validation and test sets, with performance on the validation set being presented immediately to the participant and performance on the test set presented to the participants at the workshop. In total 1863 entries were made on the validation sets during the development period and 135 entries on all test sets for the final competition. The winners used a combination of Bayesian neural networks with ARD priors and Dirichlet diffusion trees. Other top entries used a variety of methods for feature selection, which combined filters and/or wrapper or embedded methods using Random Forests, kernel methods, or neural networks as a classification engine. The results of the benchmark (including the predictions made by the participants and the features they selected) and the scoring software are publicly available. The benchmark is available at www.nipsfsc.ecs.soton.ac.uk for post-challenge submissions to stimulate further research.

    Learning Facial Attractiveness

    No full text
    In this work we study of the notion of "attractiveness" of faces in a machine-learning context. To this end, we collected human beauty ratings for datasets of facial images and used various techniques for learning the average attractiveness of a face. The results clearly show that beauty is a universal concept, which can be learned by a machine. Due to the limited size of the dataset, most of the information about the target is extracted from features that are simply correlated with facial beauty

    Open Access to Research: Changing Researcher Behavior Through University and Funder Mandates

    No full text
    The primary target of the worldwide Open Access initiative is the 2.5 million articles published every year in the planet's 25,000 peer-reviewed research journals across all scholarly and scientific fields. Without exception, every one of these articles is an author give-away, written, not for royalty income, but solely to be used, applied and built upon by other researchers. The optimal and inevitable solution for this give-away research is that it should be made freely accessible to all its would-be users online and not only to those whose institutions can afford subscription access to the journal in which it happens to be published. Yet this optimal and inevitable solution, already fully within the reach of the global research community for at least two decades now, has been taking a remarkably long time to be grasped. The problem is not particularly an instance of "eDemocracy" one way or the other; it is an instance of inaction because of widespread misconceptions (reminiscent of Zeno's Paradox). The solution is for the world's research institutions and funders to (1) extend their existing "publish or perish" mandates so as to (2) require their employees and fundees to maximize the usage and impact of the research they are employed and funded to conduct and publish by (3) depositing their final drafts in their Open Access (OA) Institutional Repositories immediately upon acceptance for publication in order to (4) make their findings freely accessible to all their potential users webwide. OA metrics can then be used to measure and reward research progress and impact; and multiple layers of links, tags, commentary and discussion can be built upon and integrated with the primary research

    Patterns in knot cohomology I

    No full text
    Cohomology theory of links, introduced by the author, is combinatorial. Dror Bar-Natan recently wrote a program that found ranks of cohomology groups of all prime knots with up to 11 crossings. His surprising experimental data is discussed in this note
    corecore