[23.10] FSRS optimizer returns same weights regardless of SM2 retention

Am fairly new to understanding FSRS, so please forgive me if I’m errant in my observation, but:

from what I can see the model weights given to me by the FSRS optimizer seem to be the same regardless of what value I enter for ‘SM2 retention’ and, additionally, regardless what I set for ‘desired retention.’

Additionally, though separately, with ‘desired retention’ set to 0.97 I’m seeing what seem to be some odd weights (though again I don’t fully understand their articulation). Specifically I’m given:

1.1228, 6.6034, 17.8637, 39.5135, 4.2433, 0.3504, 1.0120, 0.1778, 2.2007, 0.1011, 1.6502, 2.4072, 0.0366, 0.6052, 1.4239, 0.3788, 3.4351

These values lead to some very strange, undesireable intervals.

Any suggestions/advice would be appreciated (particularly as I don’t fully appreciate the weights enough to tinker with them).

SM-2 retention is not used for training. So it doesn’t affect the weights.

1 Like

Thank you for the response. Perhaps I misunderstand then; if SM-2 value does not affect weights, why is it included as an input option?

1 Like

Much obliged. I’ve been poking around quite a bit, but I seem to have overlooked that particular detail. Thank you much. :slight_smile:

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.