Use pointer network to compute the convex hull.
This is an unofficial implementation of the research paper: https://arxiv.org/abs/1506.03134
- Create a dataset:
python3 data.py
- Training:
python3 run_training.py
- Run:
python3 predict_convex_hull.py -c ckpts/hk_state_0099999.ckpt -d 256 -l 200 -s 42
Attention hidden subspace
-
Checkout
01-pca-attention.ipynb
for the details. -
Encoder attention hidden states are in a 2d plane which is a 1-to-1 mapping from the point space.
- Decoder attention hidden states go around a circle in the 2d hidden plane.