Benchmark Dataset of Complex Ginzburg Landau Simulations

This dataset contains a collection of simulations of the Complex Ginzburg Landau equation for benchmarking purposes. Simulations were run in COMSOL. The dataset contains a sweep over the parameters α ($\alpha \in \{-2.0, -1.8, -1.6, \ldots, 3.0\}$; 26 values), β ($\beta \in \{0.0, 0.1, 0.2 , \ldots, 2.5\}$; 26 values), and a random seed (15 values), resulting in 10,140 total samples. Each sample contains a snapshot of the real and imaginary component of the field after 300 time units of simulation time, on a 61x61 pixel grid (simulation domain size: 150 x 150 length units). To facilitate processing with DNNs, this data was pre-processed using ResNet-50. For small architectures, the first 32 principal components of this embedding are provided in a separate dataset.