Résumé de section

    • Lecture : slides part I - 2-5
    • Assignments
      1. Please, fill the following document (Student statistics).
      2. Once the document will be complete, estimate the following distributions:
        • P(Gl, Re, Ta, Lo) [from the statistics gathered] This is a joint distribution, check that its sum is one.
        • P(Gl, Re, Ta) [marginalization] This is another joint distribution summing to one
        • P(Lo | Gl, Re, Ta) [definition of conditional probability distribution] This is a conditional distribution, which is a set of joint distributions (one for each value of {GL, Re, Ta}, so you can check that each joint distribution sums to one.
        • et ainsi de suite pour P(Gl, Re), P(Ta | Gl, Re), P(Gl) et P(Re | Gl)
      3. Finally, choose one configuration (set of values) for {Gl, Re, Ta, Lo} and check for this configuration that P(Gl, Re, Ta, Lo) = P(Gl) * P(Re | Gl) * P(Ta | Gl, Re) * P(Lo | Gl, Re, Ta) [chain rule]
      4. Let's imagine that Re and Gl are independent, and that Lo is independent with Re and Ta given Gl. What would be the simplification of the chain rule decomposition by adding such assumptions ?