Yahoo India Web Search

Search results

    • Healthcare practitioner

      • Dr Du Plessis, Christoffel Francois (Chris), is a healthcare practitioner, specialising as a Emergency Medicine specialist, in Somerset West, Cape Town, Western Cape, South Africa
      www.medpages.info/sf/index.php?page=person&personcode=19252
  1. People also ask

  2. Oct 16, 2023 · Dr Du Plessis, Christoffel Francois, is a healthcare practitioner, specialising as a Psychiatrist, in Hospitaalpark, Bloemfontein, Free State, South Africa.

    • 5 Service Road, Hospitaalpark, Free State
    • 051 502 1835
  3. Wentzel Christoffel du Plessis held many important positions: South African Ambassador to the USA from 7 August 1956- 13 September 1960. Administrator of South West Africa from 1 December 1963 to 1 November 1968. South Africa's Permanent Representative to the United Nations - June 20, 1954.

    • Male
    • December 27, 1953
    • Anna Elisabeth (Macdonald) du Plessis
    • Abstract
    • RN(g) := EN[l(−g(x))], RU,N(g) := EU[l(−g(x))],
    • 2.2. PN Classification
    • RC-PU(g) := θPRL P(g) + RU,N(g),
    • 2.4. NU Classification
    • 3. Semi-Supervised Classification Based on PN, PU, and NU Classification
    • 4. Theoretical Analyses
    • 5.1. Experimental Analyses
    • 6. Conclusions

    Most of the semi-supervised classification meth-ods developed so far use unlabeled data for reg-ularization purposes under particular distribu-tional assumptions such as the cluster assump-tion. In contrast, recently developed methods of classification from positive and unlabeled data (PU classification) use unlabeled data for risk evaluation, i.e....

    where EP, EN, and EU denote the expectations over pP(x), pN(x), and p(x), respectively. Since we do not have any samples from p(x, y), the true risk R(g) = Ep(x,y)[l(yg(x))], which we want to minimize, should be recovered without using p(x, y) as shown below.

    In standard supervised classification (PN classification), we have both positive and negative data, i.e., fully labeled data. The goal of PN classification is to train a classifier using labeled data. The risk in PN classification (the PN risk) is defined as

    where RL P(g) := EP[−g(x)] is the risk with the linear loss lLin(m) := −m. This formulation yields the convex opti-mization problem that can be solved efficiently.

    As a mirror of PU classification, we can consider NU clas-sification. The risk in NU classification (the NU risk) is given by

    In this section, we propose semi-supervised classification methods based on PN, PU, and NU classification.

    In this section, we theoretically analyze the behavior of the empirical versions of the proposed semi-supervised classification methods. We first derive generalization er-ror bounds and then discuss variance reduction. Finally, we discuss whether PUNU or PNU classification is more promising. All proofs can be found in Appendix A.

    Here, we numerically analyze the behavior of our proposed approach. Due to limited space, we show results on two out of six data sets and move the rest to Appendix C. Common Setup: As a classifier, we use the Gaussian kernel model: g(x) = Pn i=1 wi where exp(−kx

    In this paper, we proposed a novel semi-supervised clas-sification approach based on classification from positive and unlabeled data. Unlike most of the conventional meth-ods, our approach does not require strong assumptions on the data distribution such as the cluster assumption. We theoretically analyzed the variance of risk estimators and showed...

    • 207KB
    • Tomoya Sakai, Marthinus Christoffel du Plessis, Gang Niu, Masashi Sugiyama
    • 9
    • 2017
  4. When Francois Christoffel Du Plessis was born on 19 August 1902, in Chimanimani, Chimanimani, Manicaland, Zimbabwe, his father, Hendrik Francois Conradie Du Plessis, was 51 and his mother, Johanna Maria Robberts, was 36. He died on 24 January 1975, in Makoni, Makoni, Manicaland, Zimbabwe, at the age of 72.

  5. May 23, 2016 · Tomoya Sakai, Marthinus Christoffel du Plessis, Gang Niu, Masashi Sugiyama. Most of the semi-supervised classification methods developed so far use unlabeled data for regularization purposes under particular distributional assumptions such as the cluster assumption.

  6. Semi-Supervised Classification Based on Classification from Positive and Unlabeled Data. Tomoya Sakai, Marthinus Christoffel Plessis, Gang Niu, Masashi Sugiyama. Proceedings of the 34th International Conference on Machine Learning , PMLR 70:2998-3006, 2017.

  7. Marthinus Christoffel (Christo) du Plessis Assistant Professor, University of Tokyo. Education. Bachelor, Bachelor (Hons), cum laude. Computer Engineering; University of Pretoria; Best software engineering project; Master of Engineering, cum laude. Computer Engineering; University of Pretoria