Wellcome to Alt-Tab's library,

  You can find here all papers liked or uploaded by Alt-Tab
  together with brief user bio and description of her/his academic activity.


Some information about this user

Comments:

Read the paper, add your comments…

Comments:

# In short Short article considering HIN-based recommendation. The contribution consists in using not only meta-paths in the HIN, but also "enhanced" meta-paths which are related to more elaborate motifs (typically triangles) # Summary ### 1. Introduction - usually HIN-based RS rely on the number pf meta-paths from u to i, the more there are, the higher the recommendation - Figure 1 summarizes situations that the authors want to distinguish: (u1,b4) and (u1,b3) are equivalent in terms of (P2) meta-paths, but not in terms of 3-nodes based "trust" patterns - patterns here are considered using 3 nodes patterns (à la Milo et al) as represented in Figure 2, meta-paths based on these motifs are called Motif Enhanced Meta Path (MEMP) - the concept is explored on 2 datasets: Epinions and CiaoDVD ### 2. Framework - to compute similarity using MP, we build the adjacency matrix W_{Ai,Aj} where Ai and Aj are types, then we build the commuting matrix Cp which is a product of the adjacency matrices corresponding to path P => similarities are based on counting obtained by matrix products ##### MoHINRec framework - they define adj matrix based on the same principle, but an edge is considered if it belongs to a pattern of interest (illustrated on Figure 3) - equation 2: they merge different matrices (obtained with usual MPs or MEMPs) with an alpha-weighting - then they test this with state-of-the-art HIN-based RS (actually, one they have designed in another work) ### 3. Experiments and analysis - 2 datasets, described in Table 1: - Epinions: ~22K users, 300K items, 900K ratings - CiaoDVD: ~17K users, 16K items, 72K ratings - evaluation metrics: MAE, RMSE (accuracy based) - baselines for comparison: RegSVD (MF with regularization), SoReg (MF using social links for regularization), SocialMF (MF with social trust propagation), FMG (IN-based RS, state-of-the-art) - supposedly better than the others - Settings: 8/1/1 train-validation-test, 5 experiment series with different splits ##### Performance comparison: - Summary in Table 2: MoHINRec outperforms the others - depending on the motif considered, we should vary the alpha coefficient to obtain the best possible performance (it is not clear for me how they tune alpha) - performances vary depending on the dataset and morif considered - FMG is still more efficient than othher benchmark methods ### 4. Related work ##### 4.1 HIN based recommendation ##### 4.2 motifs (very complex networks based: Milo et al., recent works by Leskovec et al.
Read the paper, add your comments…

Comments:

### Short summary ##### Plan - procedure for article selection - recommender systems overview - review itself: diversity measure ; impact of diversification on recommendation ; diversification methods - conclusion and perspectives ##### Procedure for article selection - search on Google Scholar with keywords selection - doubles elimination - selection of articles without additional payments - clustering into three groups of articles (based on review plan) ##### RS overview (standard, aiming at people new to the field) - dates back to Salton and McGill, 1980 (ref 1) - usual standard techniques: word vectors, DT, Naïve Bayes, kNN, SVM - applications: digital TV, web multimedia (YouTube, Shelfari (now merged into Goodreads), Facebook, Goodreads), personalized ads, online shopping - the general process of recommendation: past users activity collection ; create user model ; present recommendation information ; feedback collection (distinguishing explicit and implicit recommendation) - important challenges: data sparsity (working with "mostly empty user-items datasets") ; cold start (new users or items in the dataset ; overfitting (actually, rather in the sense of overspecialization) ##### Diversification - Table 1 summarizes diversity measures - Bradley-Smyth 2001: average dissimilarity between all pairs of items - Fleder-Hosanagar 2007: Gini - explore with a model how diversity evolves through recommendation cycles - Clarke et al. 2008: combined measure (ambiguity, redundancy, novelty...) - Vargas et al. 2011: intralist diversity - Hu-Pu 2011: perceived diversity (questionnaire) - Vargas et al. 2012: in the line of Clarke et al. 2008 - Castagnos et al. 2013: in the line of Bradley-Smyth 2001 - develop experiments with users - L'Huillier et al. 2014: idem - Vargas et al. 2014: binomial diversity (mixing coverage and redundancy) - Table 2 summarizes how diversity affects recommendation - usually: F-measure, MAE, NMAE - some articles show that diversification by reranking is possible without affecting too much accuracy (ex: Adomavicius et Kwon, 51) - some address the question of trade-off between diversity and accuracy (52: Hurley-Zhang 2011, 55: Aytekin-Karakaya 2014, 56: Ekstrand et al, 2014, 57: Javari-Jalili, 2014) - pb seen as multi-objective, looking for Pareto efficient ranking (58: Ribeiro et al 2015) - Table 3 summarizes diversification algorithms - many methods are reranking from accuracy-based ranking (59: Ziegler et al. 2005, 51 et 61: Adomavicius-Kwon 2012, 2011, 62: Premchaiswadi et al 2013) - then various strategies, depending on the method, on the type of data, whether the authors question temporal aspects - underlying idea is that the original algorithm (typically CF) is already diverse, just needs reordering ##### Conclusions - no consensus on a diversity metric - increasing diversity does not necessarily means sacrifice accuracy - various challenges: not enough live studies ; work in psychology would be useful ; how to use systems which have a lot of different types of items ; how to diversify during the reco process (and not a posteriori)
Read the paper, add your comments…
Pages: 1 2 3 4 5 6