1. Regression of logic
    1. Divided into binomial and multinomial
    2. The logarithmic probability of an event occurring as a ratio of its probability of not occurring is a linear function of the input X



  2. Maximum entropy model (note is discriminant
  3. Model)
    1. When learning probabilistic models, of all possible probabilities (distributions), the model with the highest entropy is the best model
    2. Maximum entropy model learning
      1. Equivalent to constrained optimization problem



      3. The characteristic function