krishnaswamylab / harmonic-alignment Goto Github PK

The minimal working python version will:

1. Compute the anisotropic diffusion map of X1 and X2
2. Compute the itersine wavelet banks for X1 and X2
3. correlate the spectral domain wavelet coefficients of X1 and X2
4. Transform X1 and X2 into the new space
5. Return the transformed matrix and a Laplacian for the new graph.

A couple nice things I want to integrate into my Python implementation:

1. Actually use diffusion maps (degree normalize the eigenvectors)
2. Use some intelligent downsampling or abstraction strategy to speed things up.
3. Diffusion wavelets (this may be a component of #2)
4. Many batch support

/home/travis/virtualenv/python3.6.7/lib/python3.6/site-packages/graphtools/base.py:168: RuntimeWarning: Cannot perform PCA to 100 dimensions on data with min(n_samples, n_features) = 62