Setup Instructions

1. Extract the dataset
   Unzip the dataset archive into the data/ directory:

   unzip data.zip -d data

2. Install dependencies
   Install all required Python packages:

   pip install -r requirements.txt

Make sure you're using Python 3.9 or higher.

Best configurations

YAGO6K103

transe: python train.py -d yago_new -m transe -e 1000 -ed 1024 -nen 128 -b 1024 -lo nssa_loss -lm 9 -lr 0.0001

rotate: python train.py -d yago_new -m rotate -e 1000 -ed 256 -nen 256 -b 1024 -lo nssa_loss -lm 9 -lr 0.001

cst: python train.py -d yago_new -m cs-transe -e 1000 -ed 512 -ced 128 -nen 32 -nenT 32 -b 1024 -lo nssa_loss -lm 9 -i_per 0.8 -lr 0.02 -lr_kappa 0.02 -lr_beta 0.3 -lr_eta 0.002

csr: python train.py -d yago_new -m cs-rotate -e 1000 -ed 256 -ced 64 -nen 512 -nenT 32 -b 1024 -lo nssa_loss -lm 9 -i_per 0.8 -lr 0.02 -lr_kappa 0.02 -lr_beta 0.2 -lr_eta 0.01

NELL995

transe: python train.py -d NELL-995_new -m transe -e 1000 -ed 256 -nen 512 -b 1024 -lo nssa_loss -lm 9 -lr 0.0001

rotate: python train.py -d NELL-995_new -m rotate -e 1000 -ed 512 -nen 512 -b 1024 -lo nssa_loss -lm 9 -lr 0.001

cst: python train.py -d NELL-995_new -m cs-transe -e 1000 -ed 512 -ced 64 -nen 512 -nenT 32 -b 1024 -lo nssa_loss -lm 9 -i_per 0.8 -lr 0.001 -lr_kappa 0.001 -lr_beta 0.01 -lr_eta 0.01

csr: python train.py -d NELL-995_new -m cs-rotate -e 1000 -ed 512 -ced 128 -nen 512 -nenT 32 -b 1024 -lo nssa_loss -lm 9 -i_per 0.8 -lr 0.005 -lr_kappa 0.005 -lr_beta 0.05 -lr_eta 0.01

FB15K237

transe: python train.py -d FB_new -m transe -e 1000 -ed 1024 -nen 512 -b 1024 -lo nssa_loss -lm 9 -lr 0.001

rotate: python train.py -d FB_new -m rotate -e 1000 -ed 512 -nen 512 -b 1024 -lo nssa_loss -lm 9 -lr 0.001 -at 2.0

cst: python train.py -d FB_new -m cs-transe -e 1000 -ed 512 -ced 512 -nen 512 -nenT 64 -b 1024 -lo nssa_loss -lm 9 -i_per 0.8 -lr 0.005 -lr_kappa 0.01 -lr_beta 0.5 -lr_eta 0.005 -at 2.0

csr: python train.py -d FB_new -m cs-rotate -e 1000 -ed 1024 -ced 256 -nen 512 -nenT 512 -b 1024 -lo nssa_loss -lm 9 -i_per 0.8 -lr 0.005 -lr_kappa 0.001 -lr_beta 0.05 -lr_eta 0.01 -at 1.5

DB241

transe: python train.py -d DB_new -m transe -e 1000 -ed 256 -nen 512 -b 1024 -lo nssa_loss -lm 18 -lr 0.001

rotate: python train.py -d DB_new -m rotate -e 1000 -ed 512 -nen 512 -b 1024 -lo nssa_loss -lm 15 -lr 0.002

cst: python train.py -d DB_new -m cs-transe -e 1000 -ed 512 -ced 64 -nen 512 -nenT 32 -b 1024 -lo nssa_loss -lm 15 -i_per 0.8 -lr 0.001 -lr_kappa 0.001 -lr_beta 0.1 -lr_eta 0.01 -at 2.0

csr: python train.py -d DB_new -m cs-rotate -e 1000 -ed 1024 -ced 128 -nen 512 -nenT 16 -b 1024 -lo nssa_loss -lm 18 -i_per 0.8 -lr 0.005 -lr_kappa 0.001 -lr_beta 0.05 -lr_eta 0.01