Geodesically convex $M$-estimation in metric spaces

Victor-Emmanuel Brunel
Proceedings of Thirty Sixth Conference on Learning Theory, PMLR 195:2188-2210, 2023.

Abstract

We study the asymptotic properties of geodesically convex $M$-estimation on non-linear spaces. Namely, we prove that under very minimal assumptions besides geodesic convexity of the cost function, one can obtain consistency and asymptotic normality, which are fundamental properties in statistical inference. Our results extend the Euclidean theory of convex $M$-estimation; They also generalize limit theorems on non-linear spaces which, essentially, were only known for barycenters, allowing to consider robust alternatives that are defined through non-smooth $M$-estimation procedures.

Cite this Paper

BibTeX
@InProceedings{pmlr-v195-brunel23a, title = {Geodesically convex $M$-estimation in metric spaces}, author = {Brunel, Victor-Emmanuel}, booktitle = {Proceedings of Thirty Sixth Conference on Learning Theory}, pages = {2188--2210}, year = {2023}, editor = {Neu, Gergely and Rosasco, Lorenzo}, volume = {195}, series = {Proceedings of Machine Learning Research}, month = {12--15 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v195/brunel23a/brunel23a.pdf}, url = {https://proceedings.mlr.press/v195/brunel23a.html}, abstract = {We study the asymptotic properties of geodesically convex $M$-estimation on non-linear spaces. Namely, we prove that under very minimal assumptions besides geodesic convexity of the cost function, one can obtain consistency and asymptotic normality, which are fundamental properties in statistical inference. Our results extend the Euclidean theory of convex $M$-estimation; They also generalize limit theorems on non-linear spaces which, essentially, were only known for barycenters, allowing to consider robust alternatives that are defined through non-smooth $M$-estimation procedures.} }
Endnote
%0 Conference Paper %T Geodesically convex $M$-estimation in metric spaces %A Victor-Emmanuel Brunel %B Proceedings of Thirty Sixth Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2023 %E Gergely Neu %E Lorenzo Rosasco %F pmlr-v195-brunel23a %I PMLR %P 2188--2210 %U https://proceedings.mlr.press/v195/brunel23a.html %V 195 %X We study the asymptotic properties of geodesically convex $M$-estimation on non-linear spaces. Namely, we prove that under very minimal assumptions besides geodesic convexity of the cost function, one can obtain consistency and asymptotic normality, which are fundamental properties in statistical inference. Our results extend the Euclidean theory of convex $M$-estimation; They also generalize limit theorems on non-linear spaces which, essentially, were only known for barycenters, allowing to consider robust alternatives that are defined through non-smooth $M$-estimation procedures.
APA
Brunel, V.. (2023). Geodesically convex $M$-estimation in metric spaces. Proceedings of Thirty Sixth Conference on Learning Theory, in Proceedings of Machine Learning Research 195:2188-2210 Available from https://proceedings.mlr.press/v195/brunel23a.html.

Related Material