C. Guo, G. Pleiss, Y. Sun, and K. Weinberger. roceedings of the 34th International Conference on Machine Learning, 70, page 1321-1330. PMLR, (2017)
Abstract
Confidence calibration – the problem of predicting probability estimates representative of the true correctness likelihood – is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important factors influencing calibration. We evaluate the performance of various post-processing calibration methods on state-of-the-art architectures with image and document classification datasets. Our analysis and experiments not only offer insights into neural network learning, but also provide a simple and straightforward recipe for practical settings: on most datasets, temperature scaling – a single-parameter variant of Platt Scaling – is surprisingly effective at calibrating predictions.
%0 Conference Paper
%1 guo2017calibration
%A Guo, Chuan
%A Pleiss, Geoff
%A Sun, Yu
%A Weinberger, Kilian Q.
%B roceedings of the 34th International Conference on Machine Learning
%D 2017
%I PMLR
%K calibration learning machine ml network neural
%P 1321-1330
%T On Calibration of Modern Neural Networks
%U http://proceedings.mlr.press/v70/guo17a.html
%V 70
%X Confidence calibration – the problem of predicting probability estimates representative of the true correctness likelihood – is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important factors influencing calibration. We evaluate the performance of various post-processing calibration methods on state-of-the-art architectures with image and document classification datasets. Our analysis and experiments not only offer insights into neural network learning, but also provide a simple and straightforward recipe for practical settings: on most datasets, temperature scaling – a single-parameter variant of Platt Scaling – is surprisingly effective at calibrating predictions.
@inproceedings{guo2017calibration,
abstract = {Confidence calibration – the problem of predicting probability estimates representative of the true correctness likelihood – is important for classification models in many applications. We discover that modern neural networks, unlike those from a decade ago, are poorly calibrated. Through extensive experiments, we observe that depth, width, weight decay, and Batch Normalization are important factors influencing calibration. We evaluate the performance of various post-processing calibration methods on state-of-the-art architectures with image and document classification datasets. Our analysis and experiments not only offer insights into neural network learning, but also provide a simple and straightforward recipe for practical settings: on most datasets, temperature scaling – a single-parameter variant of Platt Scaling – is surprisingly effective at calibrating predictions. },
added-at = {2023-11-03T13:23:31.000+0100},
author = {Guo, Chuan and Pleiss, Geoff and Sun, Yu and Weinberger, Kilian Q.},
biburl = {https://www.bibsonomy.org/bibtex/24f83b229e504dc3903ed15e15f7acd6c/jaeschke},
booktitle = {roceedings of the 34th International Conference on Machine Learning},
description = {On Calibration of Modern Neural Networks},
interhash = {315c62001aeac75c9ddacc6a8afa7f8e},
intrahash = {4f83b229e504dc3903ed15e15f7acd6c},
issn = {2640-3498},
keywords = {calibration learning machine ml network neural},
pages = {1321-1330},
publisher = {PMLR},
timestamp = {2023-11-03T13:23:44.000+0100},
title = {On Calibration of Modern Neural Networks},
url = {http://proceedings.mlr.press/v70/guo17a.html},
volume = 70,
year = 2017
}