take-koshizuka/NLSB

Neural Lagrangian Schrödinger bridge

Train

1. OU-sde data (artificial)

   python3 train.py -c config/ou-sde/[config path] -d [save dir]
   

2. sc-RNA data (real)

   python3 train.py -c config/rna/[config path] -d [save dir]
   

3. uniform (artificial)

   python3 train.py -c config/uniform/[config path] -d [save dir]
   

Evaluate

OU-sde

1. MDD

   python3 eval_mdd.py -c config/eval_mdd.json -p [checkpoint path]
   

2. CDD

   python3 eval_cdd.py -c config/eval_cdd.json -p [checkpoint path]
   

3. plot

   python3 plot_ou-sde.py -c config/plot.json -p [checkpoint path]
   

sc-RNA

1. MDD

   python3 eval_mdd.py -c config/eval_mdd.json -p [checkpoint path]
   

2. CDD

   python3 eval_lmt_traj.py -c config/eval_lmt_traj.json -p [ckpt] -pl [ckpt2]
   

3. plot on 2D-PCA space

   python3 plot_rna.py -c config/plot.json -p [ckpt] -o [out_dir]