Bayesian Wavelet and Fourier Transform Kernel Regression and Classification in RKHS

  • Author / Creator
    Zhang, Xueying
  • Kernel methods are often used for nonlinear regression and classification in machine learning because they are computationally cheap and accurate. Fourier basis and wavelet basis are the bases that can efficiently approximate the kernel functions. In previous research, Bayesian approximate kernel regression with Fourier transform has been proposed. With the proposed method, we use the analytic properties of the reproducing kernel Hilbert space (RKHS) to define a linear vector space that captures nonlinear structures. We map the data into a low-dimensional randomized feature space using Fourier transform and convert kernel function into operations of a
    linear machine. A Bayesian approximate kernel regression model is then formulated with the application of a generalized kernel model and the Bayesian method. We replace Fourier transform with wavelet transform in randomized feature space to approximate kernel functions. We formulate a new Bayesian approximate kernel model with wavelet transform and use the Gibbs sampler to compute the parameters of the model. We then make a comparison between the performance of Fourier based
    and wavelet-based Bayesian approximate kernels solving both regression and classification problems.

  • Subjects / Keywords
  • Graduation date
    Fall 2021
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.