Metric (‒m)

Rmsd

A metric is used when computing how similar two states are. By default, the root mean square difference between the states is used:

score = sqrt(sum (w_i * diff_i^2))

where w_i are the weights for variable i and diff_i is the difference between the state for variable i.

This is the default, but it can also be specified by:

wxgen sim -db db.nc -n 21 -t 365 -v 0,1,2 -w 1,2.5,0.001 -m rmsd -o rmsd.nc

Exp

Another metric is the exponential metric, which is computed as follows:

score = exp(-sum(w_i * abs(diff))

wxgen sim -db db.nc -n 21 -t 365 -v 0,1,2 -w 1,2.5,0.001 -m exp -o exp1.nc
wxgen sim -db db.nc -n 21 -t 365 -v 0,1,2 -w 2,5,0.002 -m exp -o exp2.nc

This score does not change the relative ranking of the states (so it has no impact on the -p top<N> policy), but it allows you to alter the relative probability of picking the different states. For -p random this means that the higher the weights, the less likely it is to pick a state that has a poor match.

Mad

Another metric is the mean absolute distance, which is defined as follows:

score = (sum (w_i * abs(diff_i)))

wxgen sim -db db.nc -n 21 -t 365 -v 0,1,2 -w 1,2.5,0.001 -m mad-o mad1.nc
wxgen sim -db db.nc -n 21 -t 365 -v 0,1,2 -w 2,5,0.002 -m mad -o mad2.nc

Other

[In the future, other metrics may be implemented]