Depth separation for reduced deep networks in nonlinear model reduction: Distilling shock waves in nonlinear hyperbolic problems
This work addresses the challenge of efficient model reduction for nonlinear hyperbolic problems, offering a significant improvement over classical methods for applications in computational physics and engineering.
The authors tackled the problem of approximating solutions to parametrized hyperbolic PDEs with shock waves using reduced deep networks, achieving an approximation error ε with O(|log(ε)|) degrees of freedom, while proving that classical reduced models have exponentially worse rates.
Classical reduced models are low-rank approximations using a fixed basis designed to achieve dimensionality reduction of large-scale systems. In this work, we introduce reduced deep networks, a generalization of classical reduced models formulated as deep neural networks. We prove depth separation results showing that reduced deep networks approximate solutions of parametrized hyperbolic partial differential equations with approximation error $ε$ with $\mathcal{O}(|\log(ε)|)$ degrees of freedom, even in the nonlinear setting where solutions exhibit shock waves. We also show that classical reduced models achieve exponentially worse approximation rates by establishing lower bounds on the relevant Kolmogorov $N$-widths.