* Implement the scorecard for measures
* For each edge that the measure supports for the specified data set
* Compute bias? (distance from ground truth to cost minimum)
* Perturb trajectory small amounts around ground truth and evaluate costs
* Include tests with ta=tb and ta~=tb
* Granularity (dilution of precision, distance until cost increases at all)
* Monotonicity (distance until cost begins to decrease)
* Cost (maybe smoothed around ground truth)
* Jacobian (sensitivity of cost to pertrubation from ground truth)
* Hessian (eigenvalues and consistency of their ratios, eigenvectors)
* Time to run findEdges
* Time to run evaluateEdgeCost initially
* Time to run evaluateEdgeCost repeated
Original issue reported on code.google.com by dddvis...@gmail.com on 13 Oct 2010 at 6:49
Original issue reported on code.google.com by
dddvis...@gmail.com
on 13 Oct 2010 at 6:49