Features/1458 add incremental SVD/PCA

[mrfh92]

Due Diligence

• General:
  • title of the PR is suitable to appear in the Release Notes
• Implementation:
  • unit tests: all split configurations tested
  • unit tests: multiple dtypes tested
  • documentation updated where needed

Description

Issue/s resolved: #1458
(Should be merged after #1561 as it already contains these changes)

Changes proposed:

Adds incremental SVD (see M. Brand, Fast low-rank modifications of the thin singular value decomposition, Linear Algebra and its Applications 415 (2006)) and corresponding interface for PCA

Type of change

new feature

Does this change modify the behaviour of other functions? If so, which?

no

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[JuanPedroGHM]

Looks good to me, but we have to make sure to merge randomized SVD first, and then solve all the merge conflicts that will appear from this. I'm guessing this was branched from that, but some of the functions have been renamed.

Also, it would be nice to have benchmarks for randomized SVD and incremental SVD.

[github-actions]

Thank you for the PR!

[mrfh92]

@JuanPedroGHM @ClaudiaComito I have merged the rSVD and resolved the conflicts. Now it should be possible to merge after merging rSVD without conflicts.

[mrfh92]

@JuanPedroGHM You're right regarding the benchmarks. I have added this as todo to the corresponding Issue #1615

[github-actions]

Thank you for the PR!

[JuanPedroGHM]

Benchmarks results - Sponsored by perun

function	mpi_ranks	device	metric	value	ref_value	std	% change	type	alert	lower_quantile	upper_quantile
apply_inplace_normalizer	4	CPU	RUNTIME	0.00301788	0.00101893	0.00588907	196.181	jump-detection	True	nan	nan
qr_split_0	4	GPU	RUNTIME	0.0536457	0.0701262	0.0036273	-23.5012	jump-detection	True	nan	nan
qr_split_1	4	GPU	RUNTIME	0.0522529	0.0664146	0.0012327	-21.3232	jump-detection	True	nan	nan
apply_inplace_standard_scaler_and_inverse	4	GPU	RUNTIME	0.0111402	0.009996	0.00162195	11.4462	jump-detection	True	nan	nan
qr_split_1	4	CPU	RUNTIME	0.169213	0.185829	0.00172974	-8.94135	trend-deviation	True	0.179067	0.196003
hierachical_svd_rank	4	CPU	RUNTIME	0.0515223	0.0472803	0.00999751	8.97223	trend-deviation	True	0.0464289	0.0487015
hierachical_svd_tol	4	CPU	RUNTIME	0.0564265	0.052285	0.0129836	7.92092	trend-deviation	True	0.051572	0.0534139
kmeans	4	CPU	RUNTIME	0.306328	0.317888	0.00495235	-3.63626	trend-deviation	True	0.307904	0.329682
qr_split_1	4	GPU	RUNTIME	0.0522529	0.0671226	0.0012327	-22.1529	trend-deviation	True	0.0622825	0.0740967
hierachical_svd_tol	4	GPU	RUNTIME	0.122298	0.119138	0.00582516	2.65215	trend-deviation	True	0.116942	0.121381
reshape	4	GPU	RUNTIME	0.260241	0.248041	0.0216572	4.91848	trend-deviation	True	0.238506	0.259244

Grafana Dashboard
Last updated: 2024-10-28T08:47:41Z

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!