How does it handle performance?

It promotes best practices like using jax.lax.scan for time loops and vmap for parallelizing across grid points or boundary conditions to maximize GPU utilization.

Does it support traditional numerical methods?

Yes, it provides guidance for Finite Difference Methods (FDM) and ensures they are written to be compatible with JAX's differentiation engine.

Why use JAX for physics?

JAX allows you to differentiate through the entire simulation, making it possible to optimize physical parameters or initial conditions using gradients.

Can I use this for Neural Networks?

Yes, it specifically covers Physics-Informed Neural Networks (PINNs) and hybrid models that combine data with physical laws.

What is the JAX-PDE skill?

It is a specialized capability for Claude Code focused on implementing Partial Differential Equation (PDE) solvers and differentiable physics models using JAX.

JAX Differentiable Physics & PDEs

Name: JAX Differentiable Physics & PDEs
Author: tondevrel

bytondevrel

•

Data Science & ML

Provides specialized guidance for implementing differentiable physics simulations and solving partial differential equations using JAX.

This skill enables Claude to assist researchers and engineers in building advanced scientific computing applications by leveraging JAX's automatic differentiation for physical systems. It provides implementation patterns for both traditional numerical methods like Finite Difference Methods (FDM) and modern machine learning approaches such as Physics-Informed Neural Networks (PINNs). By focusing on 'optimization-friendly' solvers, it facilitates inverse design, sensitivity analysis, and the creation of hybrid neural-physical models for domains like fluid dynamics, climate modeling, and robotics.

Key Features

01Physics-Informed Neural Network (PINN) architecture design

02Optimization-friendly Finite Difference Methods (FDM)

03High-performance computation using JAX vmap and lax.scan

041 GitHub stars

05Differentiable solver implementation (Euler, Runge-Kutta)

06Inverse problem solving and physical parameter estimation

Use Cases

01Solving Navier-Stokes or Heat equations on GPU-accelerated grids

02Developing differentiable simulations for robotics control and trajectory optimization

03Estimating unknown physical properties from experimental observation data

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tondevrel/scientific-agent-skills jax-pde

For use in Claude.ai and ChatGPT

Download Skill