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:

# In short CIKM 2015 article, one of the core reference on HIN-based recommendation, they consider a weighted variation which allows them to deal with rated cases. # Summary ### 1. Introduction ##### context and problem - more and more research on HIN for data mining: similarity search, clustering, classification - recently on reco: 2 (Jamali et al), 7 (Luo et al), 15 (Sun et al): interesting because possibility to integrate various kind of information (cf Fig 1), ex: user-movie-user is more or less a CF model - pb1: no values on links while recommendations are often associated with ratings (ex: Tom and Mary on Fig1, who have seen different movies but very different ratings) => necessary to generalize the formalism of HIN in order to account for the link weights - pb2: problem to combine information from different meta-paths, a good method of weight learning should allow to obtain personalized weights, weights should contribute to explanation, but if we personalize recommendation data sparsity problems get worst - contributions: extend HINs to weighted case ; propose SemRec (semantic path based RS) which flexibly include information of heterogeneous nature ; define consistency rule of similar users to circumvent data sparsity pb ; empirically study 2 datasets: Yelp and Douban ### 2. Heterogeneous network framework for recommendation ##### 2.1 Basic concepts - HIN for the weighted case (as usual, but with weights on one or more relation) ; illustration from Fig 2a - extended meta-paths to paths with attributes: links weights must be in a given range (give illustration) ##### 2.2 Recommendation on HIN - Table1: semantic example associated to meta-paths - Discussion about how different RS models will use meta-paths ##### 2.3 Similarity measure based on weighted meta-path - go through literature reco models based on paths in HIN (but no WHIN): PathSim (12), PCRW (4), HeteSim(10) ; cannot be simply transposed as they have no notion of weight - they adapt to this context with "atomic meta path": meta-path where weights take specific values - illustration on Fig3 on the score decomposition in the same fashion as PathSim (which counts meta-paths) ; notice the normalization step ### 3. The SemRec solution ##### 3.1 Basic idea - principle: evaluate similarity between users based on weighted and unweighted MP then infer scores from similar users - preference weights are given to different MP - difficulties: combine recommendations generated by different MP - pb1: important bias due to the fact that some types of paths are more numerous than others => similarity based on paths does not necessarily reflect similarity between objects => some kind of normalization to avoid that - pb2: we should personalize recommendations for better efficiency, but sparser data => recommendation by user groups with same preferences ##### 3.2 Recommendation with a single path - presentation of the method for one path (before generalizing) - supposing ratings from 1 to N - we have user similarity matrix for this type of specific path - compute Q_u,i,r: intensity of user u evaluating r item i from the similarity sum between users according to meta-path P_l - score predicted is the weighted average of ratings over Q_u,i,r ##### 3.3 Recommendation with multiple paths - now if we use several MP... - 3.3.1: compute weights to ratings corresponding to each type of path by minimizing mean squared error between actual and predicted scores - 3.3.2: personalized learning: each user as a weight vector, then same principle: minimizing MSE but with different weights with different users - 3.3.3: add a regularization process: learning difficult when we have few data => we use similar users ; regularization term to have the weight of a user similar to its neighbors average weight => eq 9 general form of the optimization goal ; optimization by projected gradient until convergence ##### 3.4 Discussion - general form of the objective function is L_3 in eq9 - if parameter lambda_1=0 => L_2 (equation 6) - if parameter lambda_1=infinity => L_1 (equation 4) - lambda_1 controls the level of personalization (the lower, the more personalized) - complexity analysis ### 4. Experiments ##### 4.1 Datasets - Douban (13400 u, 12700 i (= movies), 1M ratings de 1 à 5) - Yelp (16200 u, 14300 i (= buisnesses), 200K ratings de 1 à 5) - Douban clearly denser than Yelp - cf Tab.2 ##### 4.2 Metrics - accuracy evaluated with RMSE and MAE ##### 4.3 Comparison methods - 4 variants of their own model: SemRecReg (comprehensive), SemRecSgl (one type of MP), SemRecAll (same weight for everyone), SemRecInd (personalized weights, but no regularization (?)) - and methods from the literature: PMF (classic MF), SMF (PMF + reg), CMF (MF with HIN structure), HeteMF (MF + reg based on entity similarity) - no MP longer than 4 ##### 4.4 Effectiveness experiments - different settings training/test: 20/80, 40/60, 60/40, 80/20 for Douban ; 60/40, 70/30, 80/20, 90/10 for Yelp (because sparser) - Tab4: SemRecReg always have better performances in all conditions - other trends: SemRec with multiple paths in general better than SemRec with simple paths ; sparsity implies that SemRecInd is worse than SemRecAll in most circumstances (maybe I misunderstood something before) ; regularization has beneficial effects ##### 4.5 Study on cold start problem - cold start translated here by smaller training and larger test - Fig4: SemRecReg is clearly more performing in this context ##### 4.6 Study of weight preferences - explore the importance of weighted meta-paths (versus unweighted) - we can give some "slack" by imposing less constraints on scores (for example rating +/- 1) - Fig6: very demonstrative, perfs are clearly better when constraints are harder ### 5. Related work - ref 11 : HIN for data mining (Sun et Han, 2012) - ref 12 : similarity measures in HIN with PathSim (Sun et al. 2012) - ref 10 : similarity measures in HIN with HeteSim (Shi et al. 2014) - ref 2 : HIN for RS (Jamali and Lakshmanan, 2013) - the cold start question, difference depending on the technique used (MF, CF...) - ref 15 : closest to this paper, HeteRec (Yu et al., 2014) but do not use weighted paths
Read the paper, add your comments…

Comments:

# In short Aim at introducing a diversity notion for recommendation which combines different existing notions of diversity (intra-list diversity, coverage, redundancy), and then apply re-ranking technique. # Summary ### Introduction * approach based on genre Intra-List Similarity * they aim at 3 different properties: genre coverage, (non-)redundancy, list size awareness * dataset: movies rec with Netflix prize ### Related Work ##### Diversity in Recommender Systems * Herlocker et al : accuracy alone insufficient to assess satisfaction * McNee et al : defining properties related to satisfaction (coverage, diversity, novelty, serendipity) * ref 14 (Pu et al) : increasing diversity increases satisfaction * ref 22 (Ziegler et al) : introduce Intra-List Diversity * ref 7 (Clarke et al) : ILD limited when considering query results, as queries are short and ambiguous * ref 15 (Santos et al) : propose to cover a maximum of subtopics in the first results (as for a web research) ##### Measuring and enhancing diversity * frameworks to improve diversity largely rely on re-ranking * usual approach: greedy selection, assumes the definition of an objective function (see algo1, à la Ziegler), pairwise framework, measure based on the ILS (or ILD); in ref 21 (Zhang and Hurley) same kind of strategy * framework intent-aware: optimization of coverage (particularly to circumvent ambiguity problems), ref15 proposes xQuAD for example * framework proportionality aims at covering topics proportionally to the user interest, ref 9 (Dang and Croft) for example ### Characterizing genres * what characterizes a genre * following limitations (hierarchy of meaning, unbalanced distribution, overlap between genres, ...) * dataset Netflix: 100M ratings (1 to 5), 480.000 users, around 18000 movies; genres extracted from IMDB => info on 9300 movies (meaning 83% of the ratings) ### Measuring genre diversity in recommendation lists * a diversity measure should capture genre coverage (covering a maximum of genres, proportionally to user interest) * redundancy (important that items in the list cover a genre but also that other items do not cover this genre) * size-awareness (the previous two should take into account the size of the rec list, e.g. if the list is short only most important genres) * limitations of the literature: Ziegler's ILS, ref5's MMR are pairwise notions which are not well suited to evaluate notions such as a genre generality * intent-aware frameworks (refs 2 and 15) do not fully account for the idea that it is important that items do not cover a genre represented in the list, assumes that genres are independent from each other * ref9 (Dang and Croft) use the notion of proportionality to the user interest but do not penalize redundancy * no existing method take the length of the list into account ### Binomial framework for genre diversity * general principle: random distribution is considered as reference for optimal => model likelihood for a genre to randomly appear in a list according to a binomial distribution ##### Binomial diversity metric * selection of an item from a genre is seen as a Bernoulli test * n.b.: theoretically selection without replacement, practically nearly equivalent to selection with replacement * formal definitions: item i covers genre G(i) ; k_g^s = number of success on set s that item has genre g ; p_g" is proportion of interactions of a user with genre g (local importance) ; p_g' is proportion of interactions of all users with genre g (global importance) ; p_g = (1-alpha).p_g' + alpha.p_g" is the expected probability of a genre g to be in rec list R * coverage score: product of the probabilities for the genres not represented in R not to be selected randomly following the Bernoulli process (eq9) * non-redundancy score: measures how probable it is that a genre appears at least k times in R (so it's a kind of remaining tolerance) (eq10) * binomial diversity = coverage . non-redundancy * BinomDiv has appropriate properties: maximizes coverage as a function of p_g, penalizes over-representation of genres, adapts to the list length with the number of tests to do to create R ##### Binomial re-ranking algorithm * greedy re-ranking to optimize a trade-off function between relevance and diversity (eq13), parametrized by lambda ##### Qualitative analysis * results in Table 3: see how various diversity metrics behave in 4 different specific ranking situations ; principal conclusion is that BinomDiv is the only one which works all for all these situations * results in Table 4: (item based kNN + reranking) ; we observe the qualitative results of the reranking, depending on the user tastes ### Experiments * Two experiments with two datasets: Netflix prize + imdb genres (83M ratings, 480K users, 9300 movies, 28 genres) * MovieLens 1M ##### Setup * 5-fold cross-validation * RS rank all movies above a given threshold (grade) for the user considered + 1000 random movies of the dataset * RS tested: item-based CF kNN ; CF implicit Matrix Factorization ; item popularity ; random * reranking optimization is done with a grid search on lambda (trade-off diversity/relevance parameter) * diversity evaluation with all index in the literature (EILD, ERR-IA, CPR) + subtopic recall + subtopic per item * relevance evaluation with nDCG ##### Results for baseline diversity * Tab5: résults without diversification reranking (reminder: alpha reflects personalization degree) * random: very low relevance ; strong diversity * popularity: better relevance ; weaker diversity * personalized RS tend to have weaker non-personalized diversity scores but improve when the user history is taken into account ##### Results for diversified results * Tab6: résults after reranking, cutoff 20 items ; alpha =0.5 ; best lambda found with grid search * all diversifications => accuracy decreases * any diversification process is best when diversity evaluation is realized with it * xQuAD and ERR-IA tend to accumulate genres without penalizing redundancy * ERR-IA and CPR-rel correlated to SPI (subtopics per item) * Fig3: view improvement to baseline * BinomDiv can improve to baseline for nearly every diversity metric * general conclusion: BinomDiv able to bring more coverage while limiting redundancy * Tab7: explores size-awareness by changing cut-off value, diversification relative to lambda always best with the corresponding size [?]
Read the paper, add your comments…

Comments:

# In short Short article about the use of cross-domain recommendation (ie use recommendation profile on a type of products to recommend products of another type) and in particular the gain when considering cold-start problems. The article is short and esay to read, experimental results do not allow to identify a clear trend, in particular concerning diversity related results. # Summary ### Introduction ##### usual solution to cold start pbs: * ask users directly * exploit additional information for example combining collaborative filtering with content-based rec ##### literature on the second possibility (cross-domain RS) * ref 15 (Winoto, Tang): conjecture that CD-RS may degrade accuracy but improve diversity (they test this assumption in the paper) * ref 11 (Sahebi, Brusilovski): quality of rec improve when domains are semantically connected ##### article assumption profile size (quantity of info) and diversity in source domain have an impact on accuracy in target domain with CD-rec ##### 3 research questions * improvement in terms of accuracy of CD rec for cold-start users? * is CD rec really useful to improve diversity? * what is the impact of size and diversity of the user profile in the source domain for the rec in the target domain? ### Experimental setting ##### Datasets: * Facebook likes on music and movies, metadata from DBPedia * likes only so dataset with positive feedback only * typical data entry: id, name, category, timestamp * disambiguation with DBPedia (technically challenging) ##### Rec algorithms evaluated * popularity based pop * CF user-based using nearest neighbor with Jaccard (unn) * CF item-based using nearest neighbor with Jaccard (inn) * CF item-based using matrix factorization (imf) * hybrid HeteRec (see ref 16) * hybrid PathRank (see ref 8) ##### Evaluation methodology * user-based 5-fold cross-validation strategy (see ref 7) * elaborate preprocessing (only users with > 16 likes..) * after preprocessing: music is 50K users, 5800 artists, 2M likes ; movies is 27K users, 4000 movies, 875.000 likes * quality estimators: Mean Reciprocal Rank for accuracy ; Intra-list diversity and Binomial diversity (see ref 14, Vargas et al.) for diversity ; also catalogue coverage ### Results ##### most results in Tab1 ##### Cross-domain recommendation accuracy (RQ1) * specific focus on cold-start situations (profile in target domain unknown or close to unknown) * case1: Music source, Movies target * CD-unn most accurate for users with cold start but perf strongly decreases as soon as target profile grows, would be caused by the choice of Jaccard as similarity metric (which is unreliable in cold start situations) * in terms of accuracy, only inn and imf benefit from Music feedback * in terms of coverage, only unn benefit from Music feedback * case2: Movies source, Music target * CD-unn again less performing when increasing profile size * coverage: same trend as case1 * summary: CD rec may be useful in cold start situation; some methods are much more efficient when using only source domain rather than source domain + a few info from target domain (result which should be explored more) ##### Cross-domain recommendation diversity (RQ2) * binomial diversity and ILD follow similar trends * case1: in general, CD rec brings less diversity * case2: opposite trend, most CD rec brings more diversity ##### Size and diversity of source domain user profiles (RQ3) * groups users by 20 likes intervals in the source domain, and ILD quartiles * compute average MRR as a function of these categories * results on Fig1, focusing on cold start targets (few or no likes in the target domain) * observation: improvement with profile size in source domain (left panel) [not surprising] * observation: best results obtained for very focused in terms of diversity * interpretation: RS chosen have a hard time finding inter-domain correlations, in particular from Music to Movies * conclusion: user profile in source and target domains are important for rec * remark: CD-inn has better perf than other RS in many scenarios considered
Read the paper, add your comments…
Pages: 1 2 3 4 5 6