A nice survey that addresses the computational bottleneck of sampling from the conditional distribution of X (given Y=y). Is mainly about using measure transport to construct maps with triangular structure (canonically the so-called Knothe-Rosenblatt [+] arxiv.org/abs/2611.01436
Shubhendu Trivedi
@_onionesque
Cultivated Abandon.
Twitter interests: Machine learning research, applied mathematics, mathematical miscellany, ML for physics/chemistry, books.
Twitter interests: Machine learning research, applied mathematics, mathematical miscellany, ML for physics/chemistry, books.
907 Following 11K Followers 25K Posts
Latest posts 5
Really enjoyed reading this. One of my favourite topics.
arxiv.org/abs/2610.37581
arxiv.org/abs/2610.37581
Looks cool
arxiv.org/abs/2610.40618
arxiv.org/abs/2610.40618
Nice work looking at how to combine dependent tests using shared permutations, without the excessive conservatism of corrections such as Bonferroni. Finite sample error control is ensured by exchangeability, while calibration adapts to dependence.
arxiv.org/abs/2607.15823
arxiv.org/abs/2607.15823
This has an interesting idea, but not sure of what all it means. It calls a risk measure coherent if it satisfies four axioms (translation invariance, subadditivity, positive homogeneity, and monotonicity), which leads to several desirable properties. people.math.ethz.ch/~delbaen/ftp/p…