Examining Scaling Laws of Parametric Matrix Models

Parametric Matrix Models (PMMs) are a powerful new tool in machine learning, which allow for a multitude of ways to approach complex equations. In this work, we examine the scaling laws of PMMs and how their accuracy and efficiency changes with several different factors. Factors include but are not limited to trainable parameter count, amount of data, and initial formulation of the PMM. Through various plots of accuracy and inference time, we can get an idea of how PMMs scale compared to other, more common, machine learning models.