The implementation of our ICDM-21 paper "Disentangled Deep Multivariate Hawkes Process".
python == 3.6.2
numpy == 1.17.4
scipy == 1.3.1
torch == 1.6.0
Two commercial datasets are available here:
- ML-1M
- Taobao
The format of data is given as:
entity_id type_id timestamp
Two social datasets provided by previous works can be downloaded here:
- Retweets
- StackOverflow
For running ML-1M:
CUDA_VISIBLE_DEVICES=0 python -u train.py --model d2mhp --undebug --data_dir ml-1m --dis_k 4 --mi_loss 0.5
For running Taobao:
CUDA_VISIBLE_DEVICES=0 python -u train.py --model d2mhp --undebug --data_dir taobao_item --dis_k 6 --mi_loss 0.5
For running Retweets:
CUDA_VISIBLE_DEVICES=0 python -u train.py --model d2mhp --undebug --data_dir retweet --dis_k 2 --mi_loss 1
For running StackOverflow:
CUDA_VISIBLE_DEVICES=0 python -u train.py --model d2mhp --undebug --data_dir so --dis_k 4 --mi_loss 1