search

max-andr / relu_networks_overconfident

Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem [CVPR 2019, oral]
https://arxiv.org/abs/1812.05720
181 stars 21 forks source link

Figure 1 generate #1

Closed wahrheit-git closed 5 years ago

wahrheit-git commented 5 years ago

Can you tell me which scripts did you run to get the specific input partitioning as shown in Figure 1 in the refered paper's code. The reference's code does not seem to work. If you can share those scripts, it would be great.

max-andr commented 5 years ago

Hi,

I attach the code used to create Figure 1. You should run Experiment_2.ipynb. geocert_linear_regions_code.zip

Let me know if you have some questions.