Previous Article Next Article
  • Probability matrices in MMSBM. We show the inferred values for the probability matrices p from the MovieLens 100K dataset. Left to Right, the five matrices correspond to the ratings r=1,2,3,4,5. For each one of them, the rows and columns correspond to the user’s and item’s groups; here K=L=10. Each element, shown as a heat map, gives the probability pkℓ(r) that a user in group k gives a rating r to an item in group ℓ. The matrices are normalized such that ∑r∈Spkℓ(r)=1,∀k,ℓ. Note that there is no ordering of the probability matrices that would make them diagonal.
    Antonia Godoy-Lorite et al. PNAS doi:10.1073/pnas.1606316113
    Probability matrices in MMSBM. We show the inferred values for the probability matrices p from the MovieLens 100K dataset. Left to Right, the five matrices correspond to the ratings r=1,2,3,4,5. For each one of them, the rows and columns correspond to the user’s and item’s groups; here K=L=10. Each element, shown as a heat map, gives the probability pkℓ(r) that a user in group k gives a rating r to an item in group ℓ. The matrices are normalized such that ∑r∈Spkℓ(r)=1,∀k,ℓ. Note that there is no ordering of the probability matrices that would make them diagonal.




Chat now!