This implements the code for Globally Optimal Training of Neural Networks with Threshold Activation Functions paper in [1].
Run the main.ipynb file to rerun the experiment in the top row of Figure 2.
[1] Ergen, Tolga, et al. ‘Globally Optimal Training of Neural Networks with Threshold Activation Functions’. International Conference on Learning Representations, 2023, https://openreview.net/forum?id=_9k5kTgyHT.