Let's add the reordering strategy: -rauto, and also make it the default.
The intended behaviour for now is to select the "best" reordering strategy available,
i.e. expand to sloane when boost is available, or force when not.
At a later stage, after some more data analsyis, -rauto could even autotune the best reordering based on the model.
In both cases, LTSmin should write to stderr what strategy it is has actually chosen, in order to provide reproducibility of the experiment in a future where the default option has changed/improved.
I'm a bit in doubt how to specify that one wants NO reordering:
"": will mean -rauto
"-r": would then mean: no reordering.
However, it is a bit strange, the naive user might expect -r to mean: please apply reordering...
Actually, we might want the "auto" value for more options in LTSmin. Let's start here...
Let's add the reordering strategy: -rauto, and also make it the default. The intended behaviour for now is to select the "best" reordering strategy available, i.e. expand to sloane when boost is available, or force when not. At a later stage, after some more data analsyis, -rauto could even autotune the best reordering based on the model. In both cases, LTSmin should write to stderr what strategy it is has actually chosen, in order to provide reproducibility of the experiment in a future where the default option has changed/improved.
I'm a bit in doubt how to specify that one wants NO reordering: "": will mean -rauto "-r": would then mean: no reordering. However, it is a bit strange, the naive user might expect -r to mean: please apply reordering...
Actually, we might want the "auto" value for more options in LTSmin. Let's start here...