Expand description
Checking the predicted spread against the empirical one.
§The question
The spectrum of the normalised Jacobian yields a prediction: the spread
of the pose along direction i is σ_noise / σ'ᵢ. Is that true?
The answer comes from an experiment rather than an argument: the same scene is registered a thousand times with independent noise and a random initial guess, and the spread of the converged poses is compared with the prediction.
§Why this gates everything else
Until the prediction has been checked against reality, the conditioning report is a set of numbers with the right units and an unknown relation to the world. Measuring how fast such a report is produced would be premature.
Structs§
- Direction
Outcome - The outcome for a single direction of the spectrum.
- Scene
Outcome - The outcome for one scene.
- Trial
Config - Experiment parameters.