Extension beyond Gaussian-mixture class-conditional distributions

Extend the proportional-asymptotic inferential theory for the L1-penalized support vector machine from Gaussian-mixture class-conditional distributions to broader distributional classes.

Background

The class-conditional distributions in the current theory are assumed to form a symmetric two-component Gaussian mixture. The authors identify generalization to broader distributional classes as an unresolved direction needed to increase the applicability of the inferential framework.

References

Several directions remain open. First, the present theory is restricted to balanced, linearly separable data; extending it to class-imbalanced designs \citep{Loffredo2024RestoringBalance} or to the nonseparable perceptron regime \citep{Franz2019CriticalJammed} is a natural next step. Second, the theoretical framework is based on a replica-symmetric characterization and does not provide a rigorous proof of the RS ansatz or its stability; establishing such results, or identifying conditions under which the RS solution can be justified rigorously, is an important theoretical problem. Third, the framework assumes a Gaussian-mixture model for the class-conditional distributions, and extending the theory to broader distributional classes would increase its applicability.

— High-Dimensional Statistical Inference for Sparse Support Vector Machines  (2610.08345 - Zeng et al., 6 Oct 2026) in Conclusion, Section 5