6.867 Recitation Topics
September 17

   * The relation between optimization of functions of multidimensional
     parameters and the positive definiteness of the second derivative.
     Eigenvalue/eigenvector properties of symmetric matrices. Proof
     that the sample covariance is semi-positive definite.
   * A few properties of the multivariate Gaussian distribution
   * The maximum likelihood estimator and its properties in terms of
     bias and variance. Consistency, asymptotic efficiency. Example of
     a situation in which the set of parameters maximizing the
     likelihood is not unique, and of a situation in which the
     likelihood is unbounded.
   * Introduction to the MAP estimator in relation to the maximum
     likelihood estimator and an example.
   * Discussion on the fact that the ML estimator is vulnerable to
     overfitting. Review of cross-validation.