The extreme learning machine (ELM)-based visualization method (ELMVIS) is proposed Their exact position is weakly relevant to data and can be chosen arbitrarily as a grid or Gaussian distributed points. The prototypes are then randomly assigned to data points, and an ELM is used to estimate the reconstruction error. To train the visualizer, several points are chosen, their assignment permuted, and the error re-estimated. Although the exact solution requires a factorial number of trials experiments show acceptable convergence rates with up to several hundred points due to the ELM's extremely fast reconstruction error estimation. ELMVIS starts by initializing N visualization space points, taken either from a Gaussian distribution or from a regular grid. Then an ELM is initialized, and the ordering matrix is set to an identity matrix. An initial reconstruction MSE is calculated, after which an iteration starts by choosing a random number of samples.
|Number of pages||6|
|Journal||IEEE INTELLIGENT SYSTEMS|
|Publication status||Published - 2013|
|MoE publication type||A1 Journal article-refereed|