On Calibration of Modern Neural Networks
This addresses the problem of unreliable confidence estimates in neural networks for applications requiring accurate probability predictions, offering a practical solution.
The paper discovered that modern neural networks are poorly calibrated, unlike older models, and identified key factors like depth and Batch Normalization affecting calibration. It found that temperature scaling is highly effective for calibrating predictions on most datasets.
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.