Machine learning has changed how we predict weather and decode proteins, but scientists who study the ocean still face a problem: most models act like black boxes that give answers without explaining why.  

A new idea tackles this issue by using a technique called symbolic regression, which searches for clear math formulas that match real data.  

Instead of starting with a fixed physics law, the method creates many candidate equations and lets the data decide which ones work best.  

This approach was tested on predicting how coastlines shift, a key question as seas rise and human activity alters shorelines.  

Traditional models often assume the same rules everywhere, but those assumptions can fail in different places along a coast.  

By building models directly from global observations, the symbolic regression method finds equations that are both accurate and simple.  

The resulting formulas highlight which physical factors—like wave energy or sediment supply—drive changes in each region.  

Because the equations are readable, scientists can see how their data supports or contradicts known physics, leading to new insights.  

The technique shows that it is possible to let data discover laws while still keeping the models understandable and grounded in real science.

Finding Simple Shoreline Rules with Machine Learning

Actions