Browsing by Author "Marwala, Tshilidzi"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Datos abiertos en un mundo de grandes datos. Un acuerdo internacional ICSU-IAP-ISSC-TWAS
    (Ibero-American magazine of science, technology and society, 2017) Boulton, Geoffrey; Hodson, Simon; Babini, Dominique; Li, Jianhui; Marwala, Tshilidzi; Musoke, Maria G. N.; Uhlir, Paul F.; Wyatt, Sally
    La revolución digital de décadas recientes es un evento histórico mundial tan profundo y más penetrante que la introducción de la imprenta. Ha creado una explosión sin precedentes en la capacidad de adquirir, almacenar, manipular y transmitir instantáneamente grandes y complejos volúmenes de datos, con profundas implicaciones para la ciencia.1 La velocidad del cambio es formidable. En 2003 los científicos declararon que el mapeo del genoma humano estaba completo. Llevó más de diez años y costó un billón de dólares; hoy se tarda apenas unos días y cuesta una pequeña fracción de dicho monto (mil dólares). Los grandes volúmenes de datos (big data), de donde emanan flujos sin precedentes de datos desde y hacia los sistemas computacionales, y los datos amplios (broad data), en los que numerosos conjuntos de datos pueden ser semánticamente vinculados para crear significados más profundos, son los motores de esta revolución, ofreciendo nuevas oportunidades a las ciencias naturales, sociales y humanas.
  • Thumbnail Image
    Item
    Image Classification Using SVMs: One-against-One Vs One-against-All
    (arXiv preprint arXiv, 2007) Gidudu, Anthony,; Hulley, Gregg; Marwala, Tshilidzi
    Support Vector Machines (SVMs) are a relatively new supervised classification technique to the land cover mapping community. They have their roots in Statistical Learning Theory and have gained prominence because they are robust, accurate and are effective even when using a small training sample. By their nature SVMs are essentially binary classifiers, however, they can be adopted to handle the multiple classification tasks common in remote sensing studies. The two approaches commonly used are the One-Against-One (1A1) and One-Against-All (1AA) techniques. In this paper, these approaches are evaluated in as far as their impact and implication for land cover mapping. The main finding from this research is that whereas the 1AA technique is more predisposed to yielding unclassified and mixed pixels, the resulting classification accuracy is not significantly different from 1A1 approach. It is the authors conclusion therefore that ultimately the choice of technique adopted boils down to personal preference and the uniqueness of the dataset at hand.
  • Thumbnail Image
    Item
    Random ensemble feature selection for land cover mapping
    (IEEE, 2009) Gidudu, Anthony; Bolanle, Abe T.; Marwala, Tshilidzi
    Random ensemble feature selection is a means through which diversity in ensemble systems is imposed by randomly selecting the features (bands) that constitute the base classifiers. This paper provides insight and discusses the interplay between the size of the resulting ensembles and the consequent classification accuracy. From the results, it was observed that classification accuracy increased more as the number of features per base classifier increases than as the number of base classifiers increases. That said however, classification accuracy was seen to increase with additional features up to a given limit beyond which increasing the number of features per base classifier did not significantly increase classification accuracy, a peaking effect probably due to Hughes phenomenon.