Age Prediction through hyperbolic radius extraction from hyperbolic embeddings of MEG brain networks using FHNN

Age Prediction using Fully Hyperbolic Neural Networks

@article{chen2021fully,
  title={Fully Hyperbolic Neural Networks},
  author={Chen, Weize and Han, Xu and Lin, Yankai and Zhao, Hexu and Liu, Zhiyuan and Li, Peng and Sun, Maosong and Zhou, Jie},
  journal={arXiv preprint arXiv:2105.14686},
  year={2021}
}
{"mode":"full","isActive":false}

Codes for Network Embedding

Source code based on HGCN and FHNN repositories. File structure for FHNN source code:

📦gcn
 ┣ 📂data
 ┣ 📂layers
 ┃ ┣ 📜__init__.py
 ┃ ┣ 📜att_layers.py
 ┃ ┣ 📜hyp_layers.py    # Defines Lorentz Graph Convolutional Layer
 ┃ ┗ 📜layers.py
 ┣ 📂manifolds
 ┃ ┣ 📜__init__.py
 ┃ ┣ 📜base.py
 ┃ ┣ 📜euclidean.py
 ┃ ┣ 📜hyperboloid.py
 ┃ ┣ 📜lmath.py         # Math related to our manifold
 ┃ ┣ 📜lorentz.py       # Our manifold
 ┃ ┣ 📜poincare.py
 ┃ ┗ 📜utils.py
 ┣ 📂models
 ┃ ┣ 📜__init__.py
 ┃ ┣ 📜base_models.py
 ┃ ┣ 📜decoders.py      # Include FHNN decoder
 ┃ ┗ 📜encoders.py      # Include FHNN encoder
 ┣ 📂optim
 ┣ 📂utils

1. Training FHNN model using Graph Iteration

Arguments passed to program:

--task Specifies task. Can be [lp], lp denotes link prediction.

--dataset Specifies dataset.

--lr Specifies learning rate.

--dim Specifies dimension of embeddings.

--num-layers Specifies number of layers.

--bias To enable the bias, set it to 1.

--dropout Specifies dropout rate.

--weight-decay Specifies weight decay value.

--log-freq Interval for logging.

For other arguments, see config.py

In a Jupyter notebook, you can run an an example-run as follows:

! python train_graph_iteration.py
--task lp
--act None
--dataset cam_can_multiple
--model HyboNet
--lr 0.05
--dim 3
--num-layers 2
--bias 1
--dropout 0.25
--weight-decay 1e-3
--manifold Lorentz
--log-freq 5
--cuda -1
--patience 500
--grad-clip 0.1
--seed 1234
--save 1