compatibility between tutorial data and generated data
juliaggggggg55555 opened this issue · comments
When carrying out the ensemble tutorial, following exactly the same steps provided, the calculated PCA values were different from the tutorial. how to solve?
Can you give more detail of what you mean so we can try to reproduce it please?
In [23]: abs(svd.getEigvals()[:20] - pca.getEigvals()).max()
Out[23]: 6.661338147750939e-15
In [24]: abs(svd.getEigvals()[:20] - pca.getEigvals()).max()
Out[24]: 6.661338147750939e-15
In [25]: abs(calcOverlap(pca, svd).diagonal()[:20].min())
Out[25]: 1.0000000000000009
In [26]: writeNMD('ubi_pca.nmd',pca[:3], ubi)
Out[26]: 'ubi_pca.nmd'
In [27]: for mode in pca[:4]:
...: print(calcFractVariance(mode).round(3))
...:
0.134
0.094
0.083
0.065
these are the results I got.
These are the ones on the website and in the tutorial
In [19]: abs(svd.getEigvals()[:20] - pca.getEigvals()).max()
Out[19]: 7.105427357601002e-15
In [20]: abs(calcOverlap(pca, svd).diagonal()[:20]).min()
Out[20]: 0.9999999999999993
These are essentially the same numbers. Maximum difference in eigvals is 0 and absolute minimum overlap is 1.