Practice / Training and validation

Validation

Compare surrogate and solver on designs the network never saw, inside and outside the training range.

Validation compares surrogate and solver on unseen designs, field by field, and checks how error grows outside the training range.

Related

• Training: Fit the weights once on the training simulations and watch the validation curve.
• Limits and trust: A surrogate is only as good as its data: know the domain, quantify error, verify the winner.
• Aerospace: Aerodynamic loads, performance and thermal design across the flight envelope, explored before committing to tunnel and flight tests.