It may be interesting to start with a high exploration rate then gradually back exploration rate down to zero as time passes. E.g. a deadline for determining a winner.
It would also be interesting to gradually eliminate weakest variants as we approach high statistical significance.
It may be interesting to start with a high exploration rate then gradually back exploration rate down to zero as time passes. E.g. a deadline for determining a winner.
It would also be interesting to gradually eliminate weakest variants as we approach high statistical significance.