MMSR: Symbolic Regression is a multi-modal information fusion task

MMSR is a new symbolic regression algorithm. It regards symbolic regression as a task of multi-modal information fusion. Moreover, the experimental results show that it achieves the effect of SOTA on multiple symbolic regression datasets.

Model download

You can download the pre-trained model from the following links.

1, https://huggingface.co/Yanjie-CN/MMSR/tree/main

2, https://pan.baidu.com/s/1VzBbL48u0F7mt20mL-FPNA?pwd=p2ir

Getting started

Begin by setting up a virtual environment with Python 3.9. Once it's active, proceed by executing the following command at the root level of the repository:

pip install -r requirements.txt

Data acquisition

To produce a one-dimensional dataset for training the univariate model, execute the commands below:

python -m MMSR generate-dataset \
    --output-dir general/train \
    --dataset-size 130000000 \
    --n-processes 128 \
    --seed 1234
python -m MMSR generate-dataset \
    --output-dir general/valid \
    --dataset-size 10000 \
    --n-processes 128 \
    --seed 5678

To generate a two-dimensional dataset (used to train the bivariate model) run the following commands:

python -m MMSR generate-dataset \
    --output-dir general/train \
    --dataset-size 100000000 \
    --n-processes 128 \
    --seed 1234 \
    --num-variables 2
python -m MMSR generate-dataset \
    --output-dir general/valid \
    --dataset-size 10000 \
    --n-processes 128 \
    --seed 5678 \
    --num-variables 2

For further hyperparameters see python -m MMSR generate-dataset --help.

Running the inference

You can specify your own path to trained models using --model, including 'MMSR-univariate' or 'MMSR-bivariate'.

Single equation

To run a single equation:

python -m MMSR predict --model /path/to/MMSR-model 'x**2 + x'

Output:

Function: x^2 + x
R2: 1.0
Relative error: 4.21475839274614e-17

Additionally, you have the option to switch to a model of your choice.

Benchmark dataset

To run the benchmark use command bellow:

python -m MMSR evaluate-benchmark --univariate-model /path/to/MMSR-univariate --bivariate-model /path/to/MMSR-bivariate

Evaluation on dataset

To run the evaluation on dataset run the following:

python -m MMSR evaluate --model /path/to/MMSR-univariate --test-dataset-path /path/to/datast

Training a model from scratch

To train a model run the following:

python -m MMSR train \
    --config configs/{config name}.json \
    --dataset-path /path/to/train/dataset/ \
    --dataset-valid-path /path/to/valid/dataset/

where {config name} is one of the files contained in the configs directory.

Citation