IC-1400When steerable feature dimension is held constant, increasing the type-l of steerable features does not improve performance of ESCN or EquiformerV2 on IS2RE and S2EF molecular property prediction

Shih-Hsin Wang, Yung-Chang Hsu, Justin Baker, Andrea L. Bertozzi, Jack Xin, Bao Wang

SourceRethinking the Benefits of Steerable Features in 3D Equivariant Graph Neural Networks

An ablation study on the OC20 IS2RE dataset (10k training molecules) shows that ESCN with l=2, c=206 (feat dim 1854) achieves 842±13 meV energy MAE, while l=4, c=98 (feat dim 1862) gives 929±15 meV and l=6, c=64 (feat dim 1856) gives 872±6 meV. EquiformerV2 shows a similar pattern: l=2, c=34 (feat dim 306) gives 841±21 meV versus l=4, c=16 (feat dim 304) at 832±11 meV. On S2EF, ESCN l=2 c=824 (feat dim 7416) achieves 246 meV MAE, outperforming ESCN l=6 c=256 (feat dim 7424) at 294 meV. The results support the theoretical claim that information-carrying ability of steerable features is characterized by dimension, not by the highest type utilized.

Evidence
correlational
Key metric
IS2RE: ESCN l=2 c=206: 842±13 meV MAE; l=4 c=98: 929±15; l=6 c=64: 872±6. EquiformerV2 l=2 c=34: 841±21 meV; l=4 c=16: 832±11 meV. S2EF: ESCN l=2 c=824: 246 meV MAE; ESCN l=6 c=256: 294 meV MAE
Caveat
The S2EF comparison uses results from [28] for l=2 c=256 and l=6 c=256 (trained 12 epochs on 16 GPUs) versus this paper's l=2 c=824 (trained 8 epochs on 2 GPUs), so training budget differs. The authors note these tasks do not rule out confounding factors like over-squashing. IS2RE uses only 10k training molecules due to computational constraints.
Model
ESCN, EquiformerV2
Concepts
Scale-dependent behaviour
Datasets
OC20 IS2RE [eval], OC20 S2EF [eval]
Related findings
IC-1399
Extraction
automatic-extraction