aspcol.kernelinterpolation.learn_kernel_params_directional_l2

aspcol.kernelinterpolation.learn_kernel_params_directional_l2(pos, wave_num, dir_candidates, beta_candidates, reg_param, data_vec, num_iter=100)

Computes the directional kernel parameters for the directional kernel according to the L2 method in [1].

Parameters:

  • pos (ndarray of shape (num_pos, 3)) – The positions where the kernel is evaluated

  • wave_num (ndarray of shape (num_freqs,)) – The wave number for the kernel

  • dir_candidates (ndarray of shape (num_dir, 3) or int) – If int, the candidates will be generated by the function

  • beta_candidates (ndarray of shape (num_beta,) or int) – If int, the candidates will be generated by the function.

  • reg_param (float) – The regularization parameter for the kernel learning

  • data_vec (ndarray of shape (num_freqs, num_pos)) – The data vector for the kernel

  • num_iter (int) – The number of maximum iterations for the algorithm

Returns:

  • kernel_weights (ndarray of shape (num_dir, num_beta)) – The positive factor determining the relative weight of the particular kernel represented by gamma in [1]

  • dir_candidates (ndarray of shape (num_dir, 3)) – The directions for which the kernel weights are calculated

  • beta_candidates (ndarray of shape (num_beta,)) – The beta values for which the kernel weights are calculated

References

[1] R. Horiuchi, S. Koyama, J. G. C. Ribeiro, N. Ueno, and H. Saruwatari, “Kernel learning for sound field estimation with l1 and l2 regularizations,” in 2021 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), Oct. 2021, pp. 261–265. doi: 10.1109/WASPAA52581.2021.9632731.