This repo is for the UAI submission number 344: ICE-BeeM: Identifiable Conditional Energy-Based Deep Models
This is the code to run the simulations presented in the manuscript as well as the transfer learning experiments (it also allows for additional experiments not discussed in the manuscript)
This project was tested with the following versions:
- python 3.6 and 3.7
- pytorch 1.4
- tensorflow 1.14
- PyYAML 5.3.1
- seaborn 0.10
- scikit-learn 0.22.2
- scipy 1.4.1
To reproduce the simulations, run for e.g.:
python simulations.py --dataset IMCA --method icebeem
Type python simulations.py --help to learn about the arguments:
usage: simulations.py [-h] [--dataset DATASET] [--method METHOD]
[--config CONFIG] [--run RUN] [--nSims NSIMS] [--test]
[--plot]
optional arguments:
-h, --help show this help message and exit
--dataset DATASET Dataset to run experiments. Should be TCL or IMCA
--method METHOD Method to employ. Should be TCL, iVAE or ICE-BeeM
--config CONFIG Path to the config file
--run RUN Path for saving running related data.
--nSims NSIMS Number of simulations to run
--test Whether to evaluate the models from checkpoints
--plot Plot comparison of performances
The results of each simulation is saved in the value of the flag --run (defulat is run/ folder).
These experiments are run through the script main.py. To learn about its arguments type python main.py --help:
usage: main.py [-h] [--dataset DATASET] [--config CONFIG] [--run RUN] [--doc DOC] [--nSims NSIMS] [--SubsetSize SUBSETSIZE] [--seed SEED] [--all] [--baseline] [--semisupervised] [--transfer] [--plot]
optional arguments:
-h, --help show this help message and exit
--dataset DATASET Dataset to run experiments. Should be MNIST or CIFAR10, or FMNIST
--config CONFIG Path to the config file
--run RUN Path for saving running related data.
--doc DOC A string for documentation purpose
--nSims NSIMS Number of simulations to run
--SubsetSize SUBSETSIZE
Number of data points per class to consider -- only relevant for transfer learning
--seed SEED Random seed
--all Run transfer learning experiment for many seeds and subset sizes -- only relevant for transfer learning
--baseline Run the script for the baseline
--semisupervised Run semi-supervised experiments
--transfer Run the transfer learning experiments after pretraining
--plot Plot transfer learning experiment for the selected dataset
To run the experiment on MNIST:
# pretrain an ICE-BeeM on labels 0-7
python main.py --dataset MNIST --config mnist.yaml --doc mnist
# fix f and only learn g on labels 8-9 for many different dataset sizes and different seeds
python main.py --dataset MNIST --config mnist.yaml --doc mnist --transfer --all
# train an ICE-BeeM on labels 8-9 for many different dataset sizes and different seeds
python main.py --dataset MNIST --config mnist.yaml --doc mnist --transfer --baseline --all
The results are saved in the value of the flag --run (defulat is run/ folder). To plot the comparison for MNIST after running the steps above:
python main.py --dataset MNIST --plot
The same can be done on CIFAR-10 by changing the value of the flag --dataset to CIFAR10 and of the flag --config to cifar.yaml. Also make sure to change the value of --doc not to overwrite the mnist checkpoints.
To run on MNIST:
# pretrain an ICE-BeeM on labels 0-7
python main.py --dataset MNIST --config mnist.yaml --doc mnist
# pretrain an unconditional EBM on labels 0-7
python main.py --dataset MNIST --config mnist.yaml --doc mnist --baseline
# classify labels 8-9 using the pretrained ICE-BeeM
python main.py --dataset MNIST --config mnist.yaml --doc mnist --semisupervised
# classify labels 8-9 using the pretrained unconditional EBM
python main.py --dataset MNIST --config mnist.yaml --doc mnist --semisupervised --baseline
The same can be done on CIFAR-10 by changing the value of the flag --dataset to CIFAR10 and of the flag --config to cifar.yaml. Also make sure to change the value of --doc not to overwrite the mnist checkpoints.