In Multi-Marginal Schrödinger Bridge Matching (MSBM), we extend IMF algorithm into multi-marginal case and efficient (temporally parallel) learning algorithm.

This code is developed with Python3 and Pytorch. To set up an environment with the required packages,

1. Create a virtual environment, for example:

conda create -n MSBM python=3.10
conda activate MSBM

2. Install Pytorch according to the official instructions.
3. Install the requirements:

pip install -r requirements.txt

• To train and evaluate an MSBM, use the command below

python main.py --problem_name <PROBLEM_NAME>

• with PROBLEM_NAME is the tasks such as semicircle, Petal, hesc, RNA5dim.
• For Leave-one-out experiment including cite5, multi5, cite100, multi100, RNAsc, use the command below

python main.py --problem_name <PROBLEM_NAME> --LOO 1

python main.py --problem_name <PROBLEM_NAME> --LOO 2

You can find more details about other configurations in options.py, and default settings for each task are available in configs.

Our code builds upon an outstanding open source projects and papers:

• Deep Momentum Multi-Marginal Schrödinger Bridge.