New PINN framework enhances DDFT equation solving with modified activation

Researchers have developed a novel physics-informed neural network (PINN) framework designed to tackle complex nonlocal partial differential equations found in dynamic density functional theory (DDFT). This new approach incorporates a modified Lorentzian activation function, which improves convergence speed compared to standard activation functions like tanh. Additionally, the framework utilizes a precomputed discrete operator to efficiently handle nonlocal convolution terms, enhancing the training process. The effectiveness of this PINN framework was demonstrated through tests on four examples, showing good agreement with reference solutions and preserving the expected gradient-flow behavior. AI

IMPACT This research introduces a novel neural network approach that could improve the accuracy and efficiency of simulations in fields like dynamic density functional theory.

RANK_REASON Academic paper detailing a new methodology for solving specific types of equations. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• CatalyzeX
• DagsHub
• Dynamic density functional theory versus kinetic theory of simple fluids
• GALERKIN FINITE ELEMENT METHOD AND FINITE DIFFERENCE METHOD FOR SOLVING CONVECTIVE NON-LINEAR EQUATION
• Gotit.pub
• Gradient-Flow Equations
• Hugging Face
• hyperbolic tangent
• Lorentzian Activation
• Physics-Informed Neural Network
• ScienceCast
• Soumaya Elkantassi

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →