[fMapTree] = explore_map_space(S1, S2, para)
% Input:
%   S1: The LB basis of the source shape S1
%   S2: The LB basis of the target shape S2
%   para: a structure stores the following parameters
%     num_samples_on_shape: intermediate functional maps are computed on a subset of vertices
%     num_eigs: num eigenfunctions to compute (= stop_dim)
%     thres_lapcomm: threshold for the Laplacian Commutativity
%     thres_ortho: threshold for the orthogonality
%     thres_fmap_dist: if the normalized distance between two fmaps is smaller than this threshold, we assume these two maps are equivalent after applying zoomout
%     stop_dim: the maximum depth of the tree
%     max_width: the largest width of the tree at each depth (to speed up with expanding the tree too wide)
%     num_maps_keep: the number of maps that are selected from the map tree
%     thres_repeating_eigs: threshold for repeating eigenvalues detection
%
% Output:
%   fMapTree: a tree structure that contains multiple (organized) maps