Data processing wishlist

We have
- data normalization in inputs
- truncated PCA
- individual dimension Standardization in outputs
- ParameterDistributions that can be GaussianRandomFields (on cube-type fixed discretization domains)

I find more and more a key issue is dealing with other data deficiencies (scalings, dimensionality etc) in general problems. Maybe it would be good to catalogue a wishlist of some other tools that might be worth looking at for messier data in future
- Correlation-based vs Covariance-based processing (maybe an indicator/automatic selection)
- Diffusion maps? (e.g. https://docs.juliahub.com/ManifoldLearning/Uw8bd/0.9.0/)
- Nonlinear PCA (e.g. kernel PCA, or random feature PCA https://arxiv.org/pdf/1706.06296.pdf)
- Graph-Manifold projections with say, UMAP.jl
- Summary of methods, in particular CCA https://www.cs.cmu.edu/~tom/10701_sp11/recitations/Recitation_11.pdf
- derivative free likelihood informed subspaces https://www.arxiv.org/abs/2410.19990