EF-F3 = (Throughput_adaptive / Throughput_non-adaptive) × (1 - Latency_overhead / Latency_baseline)
It looks like you’re referencing a — specifically L2‑norm error estimates for adaptive refinement based on hierarchical error indicators, using basis functions or spaces labeled f1, f3, f5 (possibly edge, face, or bubble functions in a hp‑FEM context). l2hforadaptivity ef f1 f3 f5
Unlike F1 (accuracy of mapping), F3 focuses on . It measures: f5 (possibly edge
– Flux jump across interior faces (H¹‑sensitive): f3 = h_e * || [∇u_h · n] ||_L²(e) Detects discontinuities in the numerical gradient, indicating need for refinement. l2hforadaptivity ef f1 f3 f5
Our results show that:
Are you trying to or just curious about what's under the hood of your network settings?