| name | potential-fitter |
| description | Solve graph Laplacian for V(state) from reversible edge log-ratios.
Estimates potential landscape from detailed-balance constraints.
Use when working on dynamics tasks.
|
| metadata | [object Object] |
Potential Fitter
Solve graph Laplacian for V(state) from reversible edge log-ratios.
Inputs
| Parameter |
Type |
Required |
Default |
Description |
transition_matrix |
object |
Yes |
- |
Transition probability matrix from estimator |
anchor_state |
string |
No |
- |
State to anchor at V=0 (or null for auto-select) |
regularization |
number |
No |
0.01 |
Ridge regularization parameter |
Outputs
| Field |
Type |
Description |
potential |
array |
List of {state_id, V, stderr} entries |
fit_quality |
object |
RMSE of log-ratio residuals, reversible edge count |
non_equilibrium_drives |
array |
Edges with large residuals (cycles) |
Usage
python scripts/potential-fitter.py [arguments]
Generated from `skills/dynamics/potential-fitter.yaml`