Figure 6:

Results on example 2: the benchmark (Mechanical MNIST [60]) dataset. (a) The visualization of different tasks, their underlying microstructure field $b^{\eta }$, and the corresponding ground-truth solution. (b) Prediction results based on few samples $(N^{\text{test }}=2$ and $N^{\text{test }}=8)$ on a test task. (c) Comparison of MetaNO and five baseline methods.