輔助任務增強的知識追蹤方法
關鍵詞:智慧教育;知識追蹤;輔助任務;注意力機制
中圖分類號:TP391 文獻標志碼:A 文章編號:1001-3695(2025)07-023-2096-07
doi: 10.19734/j . issn.1001-3695.2024.10.0448
Abstract:Knowledge tracingisaneffctiveway toachievesmarteducation,asitasistsstudents intimelyadjustingtheir learning strategiesbyasesing theirlearning status.Theexisting knowledgetracing methods stillhaveshortcomings indealing withthe nternalrelationshipbetweenexercisesandknowledgeconceptsandobtaining students’personalized knowledge level. Toaddressthisissue,thispaperproposed anauxiliarytask-enhanced knowledge tracing(ATEKT)method.Firstly,this paper usedthegroupcalculationmethodandfrequencycalculationmethodtocalculatethedificultyof theexercise,thediscriminationoftheexercise,andthenumberofanswers totherelevantknowledgepoint exercises,andotainedtheirembeddedrepresentation.Secondly,itpresentedtwoauxiliarytasks,exercisetaging predictiontasksandpriorknowledge predictiontasksto dynamicallyaess therelationshipsbetweenexercisesandknowledge points,whileestimatingstudents'knowledgelevelsfrom historicalresponserecords.Finall,theauxiliarytaskslearnedthecharacteristicsof theoptimizationprocesstoenhance knowledge tracing model predicts.Comparedwithsevenknowledge tracing modelsonthreepublicdatasetssuchas ASSIST2009,ASSIST2017 and EdNet,the AUC and ACC of the proposed method are increased by 2.4% to 15.5% and (204 2.0% to 9.9% respectively.The experimental results show that the proposed method can efectively extract information related to knowledgetracing tasks,and itsperformanceismoresuperior.Meanwhile,ablationexperimentsalsodemonstrate the effectivenessofdifferentauxiliarytasks.
Keywords:intelligent education;knowledge tracing;auxiliary task;attention mechanisn
0 引言
智慧教育是指利用現代信息技術、教育方法和心理學原理等多種手段,實現教育過程個性化、智能化和高效化的一種新型教育模式1。(剩余15715字)
- 多模態(tài)行人重識別研究綜述...
- 語義通信在邊緣算力網絡中的應用...
- 基于同態(tài)加密和零知識證明的區(qū)塊...
- HyperledgerFabr...
- PMoE:在P-tuning中...
- 基于大語言模型的多任務生成式重...
- 基于圖文對比融合的圖像人物情感...
- 基于深度特征交互與層次化多模態(tài)...
- 反向聚焦細粒度多模態(tài)語義對齊的...
- 基于CLIP文本特征增強的剪紙...
- 基于完整超圖神經網絡的捆綁推薦...
- 基于高階鄰域信息交互的自監(jiān)督異...
- 基于超圖和分層頻譜濾波器的序列...
- 針對圖像指代分割的訓練后量化策...
- 基于信息互補與交叉注意力的跨模...
- 基于強化學習協同進化算法求解柔...
- 融合實體鄰域信息的時序知識圖譜...
- 互補盲點策略和U型Transf...
- SP-POMDP:堆疊物體抓取...
- 基于果蠅協同算法求解雙目標混裝...
- 優(yōu)化時間窗改進Dijkstra...
- 帶頻繁區(qū)域的空間并置模式挖掘方...
- 輔助任務增強的知識追蹤方法...
- 基于沖突避讓的多智能體有效旁路...
- 基于特征融合的音頻偽造檢測方法...
- 基于多視圖舌象特征融合的中醫(yī)證...
- 多元異構耦合網絡中競爭性輿情信...
- 基于增強控制流圖與孿生網絡架構...
- 獎勵回溯DQN驅動的多QoS工...
- 基于QUIC的擁塞控制算法動態(tài)...
- CN2Conv:面向物聯網設備...
- 面向物流數據共享的可撤銷屬性加...
- 一種具有多級安全目標的動態(tài)對稱...
- 基于雙向數據流分析與圖抽象嵌入...
- 基于比特切片技術與指令集的LE...
- 基于隨機投影與改進min-ma...
- 結合自適應局部圖卷積與多尺度時...
- 基于圖元變換的建筑彩繪紋樣圖像...
- 雙流特征增強與融合的弱監(jiān)督時序...
- 多尺度降噪自編碼器的遮擋行人重...
- 基于深度正則化的三維高斯人體重...
- 基于雙曲空間的無監(jiān)督視頻異常檢...