当前位置:网站首页>Introduction to sakt method
Introduction to sakt method
2022-07-07 14:10:00 【Try more, record more, accumulate more】
Network architecture and embedded interpretation :
SAKT The Internet : At each timestamp , The attention weight is estimated only for each of the preceding elements . key 、 Values and queries are extracted from the embedding layer shown below . When the first j The first element is the query element and the i When elements are key elements , Note that the weight is a i j a_{ij} aij.
Embedded layer : Embed the current exercise the student is trying and his past interactions . At each mark t + 1 t+1 t+1 when , Use exercises to embed the current problem e t + 1 e_{t+1} et+1 Embedded in the query space , Use interaction to embed elements that will interact in the past x t x_t xt Embedded in key and value spaces .
The method is introduced in detail :
Model purpose : According to the students 1 To t moment Answer the exercises of ,( Interaction sequence X = x 1 , x 2 , . . . , x t X = x_1, x_2, ..., x_t X=x1,x2,...,xt,) Forecast on t + 1 t+1 t+1 moment , exercises e t + 1 e_{t+1} et+1 Response of ( That is, predict the real situation , The right probability ).
Interactive tuples : x t = ( e t , r t ) x_t = ( e_t, r_t) xt=(et,rt) : t t t Time exercises e t e_t et Answer of r t r_t rt Composed of . x t x_t xt When numbering , Use both to express ,: y t = e t + r t × E y_t = e_t + r_t × E yt=et+rt×E , E E E Is the number of topics , You can see the interaction number , Wrong answer The time is the same as the title number y t = e t y_t = e_t yt=et, When the answer is correct , Number plus the total number of topics y t = e t + E y_t = e_t + E yt=et+E.
Embedded layer description :
The interaction sequence needs to be divided , Ensure that the length of all interaction sequences is consistent , Many are truncated , Short fill .
Therefore, the interaction sequence is composed of y = ( y 1 , y 2 , . . . , y t ) y = (y_1, y_2, ...,y_t) y=(y1,y2,...,yt) Turn into s = ( s 1 , s 2 , . . . , s n ) s = (s_1,s_2,...,s_n) s=(s1,s2,...,sn).
Train an interactive embedding matrix : M ∈ R 2 E × d M ∈ R^{2E×d} M∈R2E×d, among d It's a potential dimension , Used to get interactive embedding . s i s_i si The embedding of is expressed as M s i M_{s_i} Msi
Practice embedding a matrix : E ∈ R E × d E ∈ R^{E×d} E∈RE×d, Users get exercises embedded . e i e_i ei The embedding of is expressed as E e i E_{e_i} Eei
Location code :
In order to encode the sequence sequence , Introduce parameters P ∈ R n × d P ∈ R^{n×d} P∈Rn×d, Add to interactive embedding , Form a new code . P i P_i Pi Add to section i i i An interactive embedding vector , Form an interactive embedding vector with position coding .
From the attention level
Q: Exercises embedded
K: Answer interactively embedded
V : Answer interactively embedded
Using the attention mechanism of scaling dot product
The current exercise interacts with each previous answer Have a relationship , Calculate the attention weight .
long position
Capture information from different subspaces .
Causal relationship
Because of the sequence , Unable to know the information of the predicted topic , So use the causality layer to mask the weights learned from future interactions .
Feedforward layer
In order to add nonlinearity to the model and consider the interaction between different potential dimensions , We use a feedforward network .
Residual connection
Use low-level information
Prediction layer
The probability of getting the prediction
Network training
Cross entropy
边栏推荐
- 2022-7-7 Leetcode 34.在排序数组中查找元素的第一个和最后一个位置
- 一个简单LEGv8处理器的Verilog实现【四】【单周期实现基础知识及模块设计讲解】
- Laravel5 call to undefined function openssl cipher iv length() 报错 PHP7开启OpenSSL扩展失败
- What are the principles for distinguishing the security objectives and implementation methods that cloud computing security expansion requires to focus on?
- The reason why data truncated for column 'xxx' at row 1 appears in the MySQL import file
- How does MySQL control the number of replace?
- .net core 关于redis的pipeline以及事务
- Mathématiques avancées - - chapitre 8 différenciation des fonctions multivariables 1
- Build a secure and trusted computing platform based on Kunpeng's native security
- Did login metamask
猜你喜欢
Social responsibility · value co creation, Zhongguancun network security and Information Industry Alliance dialogue, wechat entrepreneur Haitai Fangyuan, chairman Mr. Jiang Haizhou
566. 重塑矩阵
AI talent cultivation new ideas, this live broadcast has what you care about
Advanced Mathematics - Chapter 8 differential calculus of multivariate functions 1
Navicat run SQL file import data incomplete or import failed
Selenium库
TPG x AIDU | AI leading talent recruitment plan in progress!
gvim【三】【_vimrc配置】
docker部署oracle
Parsing of XML files
随机推荐
Common response status codes
高等数学---第八章多元函数微分学1
Attribute keywords aliases, calculated, cardinality, ClientName
請問,在使用flink sql sink數據到kafka的時候出現執行成功,但是kafka裏面沒有數
现在网上开户安全么?那么网上开户选哪个证券公司?
Vmware 与主机之间传输文件
Excerpt from "misogyny: female disgust in Japan"
"Song of ice and fire" in the eleventh issue of "open source Roundtable" -- how to balance the natural contradiction between open source and security?
一个简单LEGv8处理器的Verilog实现【四】【单周期实现基础知识及模块设计讲解】
Move base parameter analysis and experience summary
Beginner XML
Is the compass stock software reliable? Is it safe to trade stocks?
Co create a collaborative ecosystem of software and hardware: the "Joint submission" of graphcore IPU and Baidu PaddlePaddle appeared in mlperf
Mysql怎样控制replace替换的次数?
VSCode 配置使用 PyLint 语法检查器
Take you to master the three-tier architecture (recommended Collection)
作战图鉴:12大场景详述容器安全建设要求
高等數學---第八章多元函數微分學1
Toraw and markraw
IP address home location query