Hi, I wouldn't call SM2 (Mnemosyne's algorithm) an optimisation algorithm, it's more like heuristic...
Cheers, Peter -----Original Message----- From: [email protected] <[email protected]> On Behalf Of [email protected] Sent: 22 May 2019 17:01 To: mnemosyne-proj-users <[email protected]> Subject: [mnemosyne-proj-users] Re: New SRS algorithm based on optimal stochastic control theory: Tabibian et al 2019 On Monday, February 11, 2019 at 3:34:57 PM UTC-5, Gwern Branwen wrote: > "Enhancing human learning via spaced repetition optimization", > Tabibian et al 2019: > https://www.pnas.org/content/early/2019/01/18/1815156116 > > > Understanding human memory has been a long-standing problem in various > > scientific disciplines. Early works focused on characterizing human memory > > using small-scale controlled experiments and these empirical studies later > > motivated the design of spaced repetition algorithms for efficient > > memorization. However, current spaced repetition algorithms are rule-based > > heuristics with hard-coded parameters, which do not leverage the automated > > fine-grained monitoring and greater degree of control offered by modern > > online learning platforms. In this work, we develop a computational > > framework to derive optimal spaced repetition algorithms, specially > > designed to adapt to the learners’ performance. A large-scale natural > > experiment using data from a popular language-learning online platform > > provides empirical evidence that the spaced repetition algorithms derived > > using our framework are significantly superior to alternatives. > > More popularized overview: http://learning.mpi-sws.org/memorize/ > > Dataset/code: https://github.com/duolingo/halflife-regression > > --------- > > It's unclear to me if this is superior in practice to any of > SuperMemo/Anki/Mnemosyne's current algorithms, since they don't > directly compare them (just to a uniform strawman baseline, and a > 'threshold' of unclear origin) or implement it on real-world users. > They are proud of their optimality guarantee, but of course that's > only optimal based on a specific set of assumptions and algorithm > classes, like being limited to a stochastic algorithm. (There might be > some other limits like being efficient asymptotically rather than at > all time-scales.) > > Nevertheless, it's cool that the result is *so* simple, and control > theory is a very rich mathematical area, so more realistic optimal > algorithms can probably be devised. (And the topic of 'point > processes' is relevant to me for my interest in various kinds of > 'anti' spaced repetition, for note-reviewing or movie-watching, which > I've mentioned before. > > -- > gwern > https://www.gwern.net Hey all! I've not posted here before, but I've been into SuperMemo and other spaced repetition algorithms for quite some time. This study is great. So the gist looks like it's looks like it's using an optimizing function with a feedback loop, where future guesses at the "rate of forgetting" are adjusted based on success/fail of previous guesses? I'm with you in asking whether this approach could be compared to something like Mnemosyne/Supermemo family algorithms rather than a uniform distribution, which really does seem like a straw man. Am I correct in thinking that Mnemosyne does the same sort of optimization relying on the user's "judgment of learning" (to use a Brainscape term) in place of an optimization function? -- You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/mnemosyne-proj-users/3c0ce8a8-5e8c-4f4d-8d1e-1be3f3b7534e%40googlegroups.com. For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/mnemosyne-proj-users/776a8a5089aa414b8989ab7df98acba1%40xmail101.UGent.be. For more options, visit https://groups.google.com/d/optout.
