离上次写间隔重复记忆算法研究的入门教程《间隔重复记忆算法:e 天内,从入门到入土。》已经过去半个多月了,不知道读者朋友们入门间隔重复记忆算法了吗?这次就主要分享一下我在研究过程[1]中用到的资源和阅读过的文献,帮助这个领域的探索者更快地上手。
科普
墨墨团队制作的间隔重复记忆模型科普,适合零基础入门:
【墨墨科普】几个公式,拯救你的记忆。_哔哩哔哩_bilibili综述
Gwern 所著的间隔重复综述,涵盖了大部分间隔重复实证研究。
高效学习的间隔重复数据集
FSRS-Anki-20k(15 亿条)
叶峻峣:目前世界最大的人类记忆行为数据集发布墨墨记忆行为数据(2.2 亿条)
采集自墨墨背单词 BMMS 系统的记忆行为数据,是首个包含时序特征的间隔重复数据集。
Replication Data for: A Stochastic Shortest Path Algorithm for Optimizing Spaced Repetition SchedulingDuolingo(1300 万条)
源自多邻国的用户学习日志。
Replication Data for: A Trainable Spaced Repetition Model for Language Learningmnemosyne(86 万条)
mnemosyne project 的开源数据,但是我没有从他们官方渠道找到相关的数据,而是从一篇提及 mnemosyne 的论文中找到的附件。
leitnerqAnki(22 万条)
我自己的 Anki 复习数据。
叶峻峣:Anki 复习数据清洗与分析(附源码和数据)开源代码
墨墨记忆算法的开源代码:
maimemo/SSP-MMC: A Stochastic Shortest Path Algorithm for Optimizing Spaced Repetition Scheduling (github.com)朋友和我基于 DSR 模型开发的间隔重复算法:
叶峻峣:自由间隔重复调度算法——FSRSDuolingo 算法的开源代码:
duolingo/halflife-regression (github.com)Memorize(一种基于标记点过程和随机最优控制的复习算法)
Networks-Learning/memorize: Code and real data for "Enhancing Human Learning via Spaced Repetition Optimization", PNAS 2019 (github.com)deeptutor(一种基于强化学习的复习算法)
rddy/deeptutor: Spaced repetition through deep reinforcement learning (github.com)leitnerq(一种基于队列网络优化和 Leitner box 的复习算法)
rddy/leitnerq: A queueing network model for spaced repetition (github.com)SM-15(非官方的社区简化实现)
slaypni/SM-15: Spaced repetition for memorizing tons of things. (github.com)研究文献
太多了,以后有机会慢慢分享这些论文的具体内容吧。这里只给一个清单:
Just a moment...A Stochastic Shortest Path Algorithm for Optimizing Spaced Repetition Scheduling | Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data MiningImproving Artificial Teachers by Considering How People Learn and Forget | 26th International Conference on Intelligent User InterfacesAdaptive Forgetting Curves for Spaced Repetition Language Learning | SpringerLinkTADS: Learning Time-Aware Scheduling Policy with Dyna-Style Planning for Spaced Repetition | Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information RetrievalBroccoli: Sprinkling Lightweight Vocabulary Learning into Everyday Information Diets | Proceedings of The Web Conference 2020Enhancing human learning via spaced repetition optimization | PNASUsing deep reinforcement learning for personalizing review sessions on e-learning platforms with spaced repetition | Semantic ScholarTeaching Multiple Concepts to a Forgetful LearnerDAS3H: Modeling Student Learning and Forgetting for Optimally Scheduling Distributed Practice of SkillsDeep Reinforcement Learning of Marked Temporal Point ProcessesAccelerating Human Learning with Deep Reinforcement Learning | Semantic ScholarA Trainable Spaced Repetition Model for Language LearningUnbounded Human Learning | Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data MiningPredicting and Improving Memory Retention: Psychological Theory MatterA Multiscale Context Model of MemoryUsing a model to compute the optimal schedule of practice. - PsycNET (apa.org)Practice and Forgetting Effects on Vocabulary Memory: An Activation‐Based Model of the Spacing Effect - Pavlik - 2005 - Cognitive Science - Wiley Online LibraryTwo components of long-term memory - PubMedOptimization of repetition spacing in the practice of learning - PubMed相关会议/期刊
大部分都不在 CCF 上,想要刷顶会/顶刊的可以不看。
叶峻峣:【随笔】AI+教育交叉方向学术会议调研/吐槽(2022)研究笔记
叶峻峣:从一个记忆伪概念,到记忆研究的难题。叶峻峣:复习算法发展的两大方向叶峻峣:从 Anki 算法说起,探索记忆的状态空间叶峻峣:从 Duolingo 机器学习算法说起,浅析记忆数据的特征工程叶峻峣:Anki 制卡对复习记忆效率的影响分析叶峻峣:系统化思维角度的记忆研究叶峻峣:我是如何在本科期间发表顶会论文的?(内含开源代码和数据集)叶峻峣:间隔重复记忆算法:e 天内,从入门到入土。拓展阅读
SuperMemo 系列算法的研发历史:
《间隔重复的历史》间隔重复记忆系统的用处:
mmjang:【三万字长文】量子物理学家是如何使用 Anki 的?怎么用好间隔重复记忆系统:
叶峻峣:如何写出好卡片:利用间隔重复创造理解间隔重复的新方向:
叶峻峣:我们如何才能开发出变革性的思想工具?这么多内容,够你看一壶了。我也没想到我居然看了这么多文献和代码,写了这么多笔记,可能这就是自由学习[2]的奥妙吧,笑死。
希望这些研究资源能够帮助到你。
叶峻峣
2022 年 9 月 5 日
参考
1. 我是如何在本科期间发表顶会论文的? ./543325359.html2. 自由学习 ./272543239.html