Penn Engineers Pioneer Mollifier Layers to Solve High-Order Inverse Partial Differential Equations Using AI

Discover how Penn Engineers use Mollifier Layers to solve inverse PDEs, uncovering hidden dynamics in DNA and materials science through advanced AI math.

By: AXL Media

Published: May 2, 2026, 7:37 AM EDT

Source: Information for this report was sourced from EurekAlert!

Bridging Ancient Calculus and Modern Machine Learning

A research team at Penn Engineering has introduced a transformative AI framework designed to tackle inverse partial differential equations, which are among the most taxing challenges in modern mathematics. These equations are essential for scientists who must work backward from visible effects, such as ripples in a pond, to identify the specific hidden causes that triggered them. By developing what they term Mollifier Layers, the engineers have provided a way for artificial intelligence to decipher the underlying rules of complex systems that were previously obscured by mathematical instability and high computational costs.

The Failure of Recursive Automatic Differentiation

For decades, the standard approach for AI to handle these problems involved recursive automatic differentiation, a method that repeatedly calculates how variables change within a neural network. However, according to co-first author Vinayak Vinayak, simply increasing computing power is insufficient for certain scientific hurdles that demand superior mathematical structures. The traditional recursive method often fails when dealing with higher-order systems or noisy data, effectively magnifying errors like a lens focusing on a jagged line, which leads to unreliable results and excessive power consumption during the training of AI models.

Reviving Mid-Century Mathematics for Signal Stability

To resolve this bottleneck, the researchers adapted a concept introduced in the 1940s by mathematician Kurt Otto Friedrichs known as mollifiers. These mathematical tools are designed to smooth out jagged or noisy functions, creating a more stable signal for the AI to process. By implementing a Mollifier Layer, the Penn team, including co-first author Ananyae Kumar Bhartari, found they could eliminate the noise before attempting to measure change. This adjustment allows the neural network to calculate higher-order derivatives with significantly improved reliability and a reduced computational burden.