Pytorch implementation of paper: Continuous Indeterminate Probability Neural Network,
- Our environment is: Python 3.10.11 torch==2.0.1, torchvision==0.15.2
pip install -r requirements.txt
-
Run CIPNN on Datasets to check classfication performance.
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --num_latent_vars 3 --learning_rate 1e-3 --forget_num 3000
python3 demo_fashionmnist.py --num_epoch 10 --gamma 0.9 --num_latent_vars 3 --learning_rate 1e-3 --forget_num 3000
python3 demo_cifar10.py --num_epoch 10 --gamma 0.85 --num_latent_vars 3 --learning_rate 1e-4 --forget_num 3000
python3 demo_stl10.py --num_epoch 10 --gamma 0.9 --num_latent_vars 3 --learning_rate 1e-4 --forget_num 3000 -
Run CIPNN on large latent space.
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 1000 --scaler_factor 4.05
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 500
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 200
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 100
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 50
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 20
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 10
python3 demo_mnist.py --num_epoch 10 --gamma 0.9 --learning_rate 1e-3 --forget_num 500 --num_latent_vars 5 -
Run CIPNN on MNIST to check the auto-encoder task performance with different latent spaces.
python3 demo_autoencoder_mnist.py --mnist_data_path ../ --num_epoch 10 --gamma 0.99 --num_latent_vars 1
python3 demo_autoencoder_mnist.py --mnist_data_path ../ --num_epoch 10 --gamma 0.98 --num_latent_vars 2
python3 demo_autoencoder_mnist.py --mnist_data_path ../ --num_epoch 10 --gamma 0.93 --num_latent_vars 10
The logs of above commands are stored into log folder, you can easily access them if you do want.