# eigenvalue (blogs) — RSS Amplifier

Recent posts from the 2 feeds in the RSS Amplifier directory that cover eigenvalue.

Page: <https://rssamplifier.com/topics/eigenvalue/blogs>  
Feed: <https://rssamplifier.com/topics/eigenvalue/blogs.md>

---

## [When Eigenvalues Collide](https://alexshtf.github.io/2026/07/01/Spectrum-Moreau.html)

_2026-07-01 · Alex Shtoff · Alex Shtoff_

Accelerating convergence of eigenvalue models by Moreau regularization of the mid-eigenvalue.

## [Cheaper eigenvalue training and inference](https://alexshtf.github.io/2026/03/15/Spectrum-Banded.html)

_2026-03-15 · Alex Shtoff · Alex Shtoff_

Cheaper eigenvalue training and inference with symmetric tridiagonal matrices: preserve useful expressiveness, use fast SciPy-backed PyTorch autograd, and avoid dense eigensolvers.

## [Interpreting eigenvalue models](https://alexshtf.github.io/2026/02/03/Spectrum-Interpretation.html)

_2026-02-03 · Alex Shtoff · Alex Shtoff_

Interpreting eigenvalue-based ML models: read the k-th eigenvalue as a two-player game using Courant principle, a sequential orthogonality process via Courant-Fischer theorem, and a difference-of-convex function using Ky Fan variational principle.

## [I feel the need for Eigen-Speed](https://alexshtf.github.io/2026/01/20/Spectrum-Speed.html)

_2026-01-20 · Alex Shtoff · Alex Shtoff_

PyTorch eigenvalues on CUDA can be unexpectedly slow due to device synchronization. This post shows how to call CuPy via DLPack for fast GPU eigvalsh/eigh while keeping gradients for training.

## [Robustness, interpretability, and scaling of eigenvalue models](https://alexshtf.github.io/2026/01/01/Spectrum-Props.html)

_2026-01-01 · Alex Shtoff · Alex Shtoff_

Robustness, interpretability, and scaling of eigenvalue models: stability bounds from Weyl's inequality, operator-norm feature importance, and regularization experiments for tabular data.

## [Behold the power of the spectrum!](https://alexshtf.github.io/2025/12/16/Spectrum.html)

_2025-12-16 · Alex Shtoff · Alex Shtoff_

Eigenvalues as neurons: represent nonlinear models as the k-th eigenvalue of a learned symmetric matrix pencil. Explore monotonicity/convexity properties and train simple spectral models.

## [Paying attention to feature distribution alignment](https://alexshtf.github.io/2025/08/19/Orthogonality.html)

_2025-08-19 · Alex Shtoff · Alex Shtoff_

Orthogonal polynomial features are only uncorrelated when the feature distribution matches the basis weight. Use CDF/quantile transforms to align distributions and get more informative Legendre features.

## [Off with the polynomial’s tail!](https://alexshtf.github.io/2025/04/17/Polynomial-Pruning.html)

_2025-04-17 · Alex Shtoff · Alex Shtoff_

Legendre polynomial feature regression on California Housing shows double descent; a simple tail-pruning of high-degree coefficients yields smaller, competitive models. Implemented in scikit-learn.

## [Let the polynomial monster free](https://alexshtf.github.io/2025/03/27/Free-Poly.html)

_2025-03-27 · Alex Shtoff · Alex Shtoff_

Overparameterized polynomial regression can exhibit double descent: past the interpolation threshold, some bases memorize and still generalize. Experiments compare power, Legendre, Chebyshev, and Fourier features.

## [Home of the Agentic Pull Request (Sponsored)](https://crawlproof.com/a/Ln6r7CcJr4yT)

_2025-03-26 · **Sponsored**_

CLI-first, TUI-ready toolkits and a moshscript installer — join the waitlist

## [Shape restricted function models via polyhedral cones](https://alexshtf.github.io/2024/11/09/Shape-Restricted-Models-Polyhedral.html)

_2024-11-09 · Alex Shtoff · Alex Shtoff_

Enforce richer shape constraints (convexity/concavity and combinations with monotonicity) by constraining coefficient vectors to polyhedral cones. Implement a PyTorch cone layer and fit concave functions.