recommender.md

Recommender System

Recommender Systems (RSs) are software tools and techniques providing suggestions for items to be of use to a user.

RSs are primarily directed towards individuals who lack sufficient personal experience or competence to evaluate the potentially overwhelming number of alternative items that a Web site, for example, may offer.

• https://github.com/hongleizhang/RSPapers
• https://github.com/YuyangZhangFTD/awesome-RecSys-papers
• https://github.com/daicoolb/RecommenderSystem-Paper
• https://github.com/grahamjenson/list_of_recommender_systems
• https://github.com/benfred/implicit

Collaborative Filtering

If we have collected user ${u}$'s evaluation to the item ${i}$ , $R_{[u][i]}$, and all such data makes up a matrix $R=(R_{[u][i]})$ while the user $u$ cannot evaluate all the item so that the matrix is incomplete and missing much data.

Matrix completion is to complete the matrix $X$ with missing elements, such as

$$ \min Rank(Z) \\ s.t. \sum_{(i,j):Observed}(Z_{(i,j)}-X_{(i,j)})^2\leq \delta $$

SVD is to factorize the matrix into the multiplication of matrices so that $$ \hat{R}=P^{T}Q $$

And we can predict the score $R_{[u][i]}$ via

$$\hat{R}{[u][i]}=\hat{r}{u,i}=\left<P_u,Q_i\right>=\sum_f p_{u,f}q_{i,f}$$

where $P_u, Q_i$ is the ${u}$th column of ${P}$ and the ${i}$th column of ${Q}$, respectively. And we can define the cost function

$$C(P,Q) = \sum_{(u,i):Observed}(r_{u,i}-\hat{r}{u,i})^{2}=\sum{(u,i):Observed}(r_{u,i}-\sum_f p_{u,f}q_{i,f})^{2}.$$

Additionally, we can add regular term into the cost function to void over-fitting

$$C(P,Q) = \sum_{(u,i):Observed}(r_{u,i}-\sum_f p_{u,f}q_{i,f})^{2}+\lambda(|P_u|^2+|Q_i|^2).$$

---

• https://www.zhihu.com/question/47716840/answer/110843844
• https://www.cnblogs.com/Xnice/p/4522671.html
• https://blog.csdn.net/turing365/article/details/80544594
• https://en.wikipedia.org/wiki/Collaborative_filtering
• https://bugra.github.io/work/notes/2014-04-19/alternating-least-squares-method-for-collaborative-filtering/

Deep Learning and RS

Deep learning is powerful in processing visual and text information.

• https://github.com/robi56/Deep-Learning-for-Recommendation-Systems
• https://github.com/wzhe06/Reco-papers
• https://github.com/shenweichen/DeepCTR
• https://github.com/cyhong549/DeepFM-Keras

Computational Advertisement

• https://github.com/wzhe06/Ad-papers