=================================================================================

Recommending items using a classification approach can:
          i) Provide personalization.
          ii) Capture context (e.g., time of day).

Recommending items using a simple count based co-occurrence matrix can:
          i) Provide personalization.

Recommending items using featurized matrix factorization can:
          i) Provide personalization.
          ii) Capture context (e.g., time of day).

============================================

A reader is enjoying a article with a representation: [2 3 2 1 2]. Which of the following articles would you recommend to the reader next?
          [7 0 2 1 0] * [2 3 2 1 2] = [14 0 4 1 0] = 19
          [1 7 0 0 2] * [2 3 2 1 2] = [2 21 0 0 4] = 27
          [1 0 0 0 7] * [2 3 2 1 2] = [2 0 0 0 14] = 7
          [0 2 0 0 7] * [2 3 2 1 2] = [0 6 0 0 14] = 20
      Then, the answer is:
          [1 7 0 0 2]