Open DominikPeters opened 5 months ago
That's a good idea. One issue is that the ordinal samplers currently return Numpy arrays that cannot really be used to store weak orders. In order for the weak order samplers to be usable with other features of the package (typically mixing a 30% weak order and 70% Mallows') we would need to change that...
@szufix, any thought?
It's now implemented:
Let me know what you think of this.
In our paper on defining IRV and STV for weak orders, we needed to sample random weak-order profiles. We did this using a "coin flip model" and a "radius model", see page 16.
This could be a useful addition to the
prefsampling
package.Our code for the coin-flip strategy: