What is Differentiable Simulation?
Differentiable simulation combines physics-based simulation with automatic differentiation to solve inverse problems.
Instead of:
- Forward Problem: Given parameters → predict observations
- Traditional Inverse: Trial-and-error parameter search
We use:
- Differentiable Simulation: Compute gradients through simulation → gradient-based optimization
Key insight:
If we can compute \( \frac{\partial \text{simulation}}{\partial \text{parameters}} \), we can use gradient descent to find optimal parameters.
©
|
Cornell University
|
Center for Advanced Computing
|
Copyright Statement
|
Access Statement
CVW material development is supported by NSF OAC awards 1854828, 2321040, 2323116 (UT Austin) and 2005506 (Indiana University)
CVW material development is supported by NSF OAC awards 1854828, 2321040, 2323116 (UT Austin) and 2005506 (Indiana University)