### Current browse context:

math.NA

### Change to browse by:

### References & Citations

# Mathematics > Numerical Analysis

# Title: The Deep Ritz Method for Parametric $p$-Dirichlet Problems

(Submitted on 5 Jul 2022)

Abstract: We establish error estimates for the approximation of parametric $p$-Dirichlet problems deploying the Deep Ritz Method. Parametric dependencies include, e.g., varying geometries and exponents $p\in (1,\infty)$. Combining the derived error estimates with quantitative approximation theorems yields error decay rates and establishes that the Deep Ritz Method retains the favorable approximation capabilities of neural networks in the approximation of high dimensional functions which makes the method attractive for parametric problems. Finally, we present numerical examples to illustrate potential applications.

Link back to: arXiv, form interface, contact.