Gradient-Free Topology Adaptation for Power Flow Surrogates via In-Context Whitening
arXiv:2607.12241
2026
Architecture
1 ideas extracted · analyzed Aug 30, 2026
What the math gives to ML
The paper provides a concrete operator-shift adaptation mechanism: represent outputs in a ZCA-whitened coordinate system and adapt to a shifted domain by estimating only the target output mean and covariance from calibration cases, without changing network weights. The transferable asset is a weight-free affine transport map between source and target output distributions, with a measurable prediction that much of topology-induced error should disappear after matching first and second moments. This is suitable for neural surrogates whose outputs are physical fields, graph-node quantities, or multi-task regression vectors.
Ideas from this paper
✗ Failed on benchmark
2026
Train a neural surrogate to predict outputs in a source-domain ZCA-whitened space, then adapt to a shifted domain using only the shifted domain's output mean and covariance. At inference, transport the network prediction through the target covariance square root, yielding a weight-free correction that preserves output-coordinate semantics and can be applied to MLP, CNN, graph-NN, or transformer regressors.
Useful7/10
Difficulty4/10
Novelty5/10