Code for the paper Expectation consistency for calibration of neural networks.

• File main.py is used to train models for datasets SVHN and CIFAR10

• The code to compute Expectation Consistency and Temperature Scaling are respectively find_optimal_temperature_expectation_consistency and find_optimal_temperature_expectation_consistency, in the file comparison.py.

The following repositories were partly used :

• https://github.com/akamaster/pytorch_resnet_cifar10
• https://github.com/chenyaofo/pytorch-cifar-models