Expertium
Expertium
I've been thinking that maybe D is not the bottleneck, in other words, it's not the biggest obstacle on the path to decreasing RMSE. Perhaps trying to improve D doesn't...
I think there is actually a neural network that outputs a "similarity score" based on the card's text. @L-M-Sherlock I can't find it, but I think I saw it when...
> These effects are impossible to account for Of course if something happened outside of Anki, it's impossible to account for. But if something happened within Anki - such as...
So out of curiosity, I decided to combine the three most promising (on paper) ideas: 1) Optimizable values for grades instead of 1, 2, 3, 4 2) Initial D is...
> These discussions are about the optimizer or the model. I guess you mean the hyper-parameters? He meant the optimizer itself, the stuff that adjusts parameters and finds the optimum....
Oh, nevermind, it already has something similar: `"Remember language on a per deck basis": "True".` Though I think it should remember the destination field too. Plus, it would be nice...
Hey, it's me again. Any plans to make the add-on remember the destination field for each note type?
So, has it been implemented? I can't find it
I suppose we can release it in the future. Btw, if you include it in the table, it will be a little misleading. Right now it says 708,151,820 reviews, but...
Alternatively, we could use the version with new parameters and call it FSRS v5, or at least alpha v5.