Online Symbolic Regression with Informative Query

Symbolic regression, the task of extracting mathematical expressions from observed data, plays a crucial role in scientific discovery. Despite promising performance existing methods, most them conduct symbolic regression an offline setting. That is, they treat data points as given ones that are simply sampled uniform distributions without exploring expressive potential data. However, for real-world problems, used usually actively obtained by doing experiments, which is online Thus, how to obtain informative can facilitate process important problem remains challenging. In this paper, we propose QUOSR, query-based framework automatically iterative manner. Specifically, at each step, QUOSR receives historical points, generates new x, and then queries expression get corresponding y, where (x, y) serves points. This repeats until maximum number query steps reached. To make generated informative, implement with neural network train it maximizing mutual information between target expression. Through comprehensive show modern methods generating