Fig. 3: Illustration of GP reconstruction on selected 2D data slices.
From: Imaging mechanism for hyperspectral scanning probe microscopy via Gaussian process modelling
Selected 2D representations corresponding to the application of GP regression to original (full) data (a) and data corrupted by removing 70% (b), 90% (c), and 99% (d) of observations. In each panel, the top row shows the input data (2D slices of hyperspectral data average over selected frequency range and spectroscopic curves extracted at A, B, and C points) and the bottom row shows the model output (the same averaged slices and the reconstructed curves from the same locations as in the input data). The vertical blue span in the plots indicate a slice used for 2D image plots (the same in all four panels). The spatial resolution in all the images is ~9.76 nm/px. The full 3D data sets are available from the notebook at https://colab.research.google.com/github/ziatdinovmax/GP/blob/master/notebooks/GP_BEPFM.ipynb.
