Install torchdiffeq
from https://github.com/rtqichen/torchdiffeq.
conda env create -f attentive_ode.yml
By default, the dataset are downloadeded and processed when script is run for the first time.
Raw datasets:
- Attentive ODE
sh test.sh
or
python3 run_models.py --niters 20 -n 8000 -l 20 --dataset physionet --attentive-ode --rec-dims 40 --rec-layers 3 --gen-layers 3 --units 50 --gru-units 50 --quantization 0.016 --classif
- DATA : Physionet
Time Series Latent(Seed) | AUC Score |
---|---|
1991 (91194) | 0.8470 |
2022 (93669) | 0.8430 |
2021 (53879) | 0.8496 |
2020 (6152) | 0.8454 |
2019 (51159) | 0.8470 |
Test MSE on PhysioNet. Encoder-decoder models.
Model | Interp(x 10^-3) | Extrap(x 10^-3) |
---|---|---|
RNN-VAE | 5.930±0.249 | 3.055±0.145 |
LatentODE (RNN enc.) | 3.907±0.252 | 3.162±0.052 |
LatentODE (ODE enc.) | 2.118±0.271 | 2.231±0.029 |
LatentODE + Poisson | 2.789±0.771 | 2.208±0.050 |
ACE-NODE(ODE enc.) | -- | 2.045±0.039 |