Statistical Foundations of Prior-Data Fitted Networks

Authors: Thomas Nagler

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The main contribution of this work is establishing the theoretical foundation for PFNs and identifying statistical mechanisms explaining their empirical behavior. ... 6.5. Numerical Validation
Researcher Affiliation Academia 1Department of Statistics, LMU Munich, Munich, Germany 2Munich Center for Machine Learning, Munich, Germany.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes An R script to reproduce the results can be found at https://gist.github.com/tnagler/62f6ce1f996333c799c81f1aef147e72.
Open Datasets No The paper uses simulated data sets and does not provide concrete access information for a publicly available or open dataset.
Dataset Splits No The paper describes data generation for evaluation but does not specify traditional training/validation/test dataset splits.
Hardware Specification No The paper does not provide specific hardware details used for running its experiments.
Software Dependencies Yes We run the pre-trained Tab PFN of Hollmann et al. (2022, pip version 0.1.8).
Experiment Setup Yes We simulate 500 data sets Dn from the model p0(1 | X) = 1/2 + sin(1 X)/2 with Y {0, 1}, X N(0, I5), and run the pre-trained Tab PFN of Hollmann et al. (2022, pip version 0.1.8). ... Figure 1 also shows the results of a localized version of Tab PFN (as in Section 6.4 with kn = min{500, n4/(d+4) }).