Mollifier Layers for Efficient High-Order Inverse PDE Learning

About this episode

This episode discusses the introduction of Mollifier Layers, a new module aimed at improving the efficiency of high-order inverse PDE learning in machine learning.

This paper introduces Mollifier Layers, a novel, lightweight module designed to enhance Physics-Informed Machine Learning (PhiML) by replacing recursive automatic differentiation with convolutional operations. While traditional methods like Physics-Informed Neural Networks (PINNs) struggle with computational costs, memory blow-up, and noise instability when calculating high-order derivatives, this new approach uses analytically defined smooth kernels to transform differentiation into stable i...

People in this episode

Host: Mike Breault

Topics covered

• Physics-Informed Machine Learning
• Mollifier Layers
• high-order derivatives
• computational efficiency
• machine learning

Keywords

• Mollifier Layers
• Physics-Informed Machine Learning
• high-order derivatives
• computational costs
• convolutional operations

Mentioned in this episode

Organizations: Physics-Informed Machine Learning, Physics-Informed Neural Networks

Books & works: Mollifier Layers

More episodes of Intellectually Curious

• The Giant Space Umbrella · June 5, 2026 · 5 min
• Natural Language Autoencoders for Unsupervised LLM Interpretability · May 8, 2026 · 6 min
• The Rise of Point Absorbers · May 6, 2026 · 6 min
• Autocompleting Reality: The Rise of Large Event Models · May 5, 2026 · 6 min
• Agentic Commerce 2026: AI Shoppers Do the Shopping · May 4, 2026 · 6 min
• Autodata Unleashed: How AI Learns to Learn · May 3, 2026 · 6 min