A multi-fidelity SBI method using feature matching and knowledge distillation outperforms weight-initialization transfer learning at small high-fidelity simulation budgets.
Optuna: A next-generation hyperparameter optimization framework
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
astro-ph.CO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI
A multi-fidelity SBI method using feature matching and knowledge distillation outperforms weight-initialization transfer learning at small high-fidelity simulation budgets.