%0 Conference Paper %1 pmlr-v139-biggio21a %A Biggio, Luca %A Bendinelli, Tommaso %A Neitz, Alexander %A Lucchi, Aurelien %A Parascandolo, Giambattista %B Proceedings of the 38th International Conference on Machine Learning %D 2021 %E Meila, Marina %E Zhang, Tong %I PMLR %K ak-symbolic-numeric equations mathematicalrelation maths neural-network reading-done regression symbolic %P 936--945 %T Neural Symbolic Regression that scales %U https://proceedings.mlr.press/v139/biggio21a.html %V 139 %X Symbolic equations are at the core of scientific discovery. The task of discovering the underlying equation from a set of input-output pairs is called symbolic regression. Traditionally, symbolic regression methods use hand-designed strategies that do not improve with experience. In this paper, we introduce the first symbolic regression method that leverages large scale pre-training. We procedurally generate an unbounded set of equations, and simultaneously pre-train a Transformer to predict the symbolic equation from a corresponding set of input-output-pairs. At test time, we query the model on a new set of points and use its output to guide the search for the equation. We show empirically that this approach can re-discover a set of well-known physical equations, and that it improves over time with more data and compute.

@inproceedings{pmlr-v139-biggio21a, abstract = {Symbolic equations are at the core of scientific discovery. The task of discovering the underlying equation from a set of input-output pairs is called symbolic regression. Traditionally, symbolic regression methods use hand-designed strategies that do not improve with experience. In this paper, we introduce the first symbolic regression method that leverages large scale pre-training. We procedurally generate an unbounded set of equations, and simultaneously pre-train a Transformer to predict the symbolic equation from a corresponding set of input-output-pairs. At test time, we query the model on a new set of points and use its output to guide the search for the equation. We show empirically that this approach can re-discover a set of well-known physical equations, and that it improves over time with more data and compute.}, added-at = {2023-04-26T15:07:41.000+0200}, author = {Biggio, Luca and Bendinelli, Tommaso and Neitz, Alexander and Lucchi, Aurelien and Parascandolo, Giambattista}, biburl = {https://www.bibsonomy.org/bibtex/2b4764963c5a1dd1f3470c84ae47bcfa5/adulny}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, description = {Neural Symbolic Regression that scales}, editor = {Meila, Marina and Zhang, Tong}, interhash = {f3e2718fefb52ab3582c90f341464516}, intrahash = {b4764963c5a1dd1f3470c84ae47bcfa5}, keywords = {ak-symbolic-numeric equations mathematicalrelation maths neural-network reading-done regression symbolic}, month = {18--24 Jul}, pages = {936--945}, pdf = {http://proceedings.mlr.press/v139/biggio21a/biggio21a.pdf}, publisher = {PMLR}, series = {Proceedings of Machine Learning Research}, timestamp = {2023-10-23T13:07:19.000+0200}, title = {Neural Symbolic Regression that scales}, url = {https://proceedings.mlr.press/v139/biggio21a.html}, volume = 139, year = 2021 }