This is an experimental webpage for Khan-Academy-style video highlights of select recitations. Answers to some popular questions from recitations or from office hours may appear here as well.
Warning: The material here is a bit informal, will be posted with slight delay, and is not meant to replace recitations.
09/09/2011: Recitation 1 Highlights: Gaussian random vectors, space and time complexities for manipulating distributions
09/16/2011: Recitation 2 Highlights: Alarm network - intuition behind Bayes Ball rules
09/16/2011: Recitation 2 Highlights: Bipartite graphs - examples of graphical models, representing distributions compactly with linear algebra
09/16/2011: Recitation 2 Highlights: Questions - Bayes ball, super-nodes, testing conditional independencies
09/23/2011: Recitation 3 Highlights: Warm-up 1 - I-map or D-map? (1)
09/23/2011: Recitation 3 Highlights: Warm-up 2 - I-map or D-map? (2)
09/23/2011: Recitation 3 Highlights: Warm-up 3 - Undirected graph to directed minimal I-map
09/23/2011: Recitation 3 Highlights: Questions
09/30/2011: Recitation 4 Highlights: Why conditioning is easy in info form (algebra)
09/30/2011: Recitation 4 Highlights: Visualizing the Elimination Algorithm
09/30/2011: Recitation 4 Highlights: Gaussian graphical model problem
09/30/2011: Recitation 4 Highlights: DAG elimination problem
09/30/2011: TA Office Hours: Why info matrix sparsity relates to pairwise Markov property
09/30/2011: TA Office Hours: Gaussian random vectors - marginal and block independence
For multivariate Gaussian x: x_1 indep. of x_2 and x_1 indep. of x_3 implies that x_1 indep. of (x_2, x_3)
10/07/2011: Recitation 5 Highlights: Probabilistic interpretation of forward-backward
10/07/2011: Recitation 5 Highlights: Computational complexity of convolution codes
10/07/2011: Recitation 5 Highlights: Elimination algorithm with observed nodes
10/14/2011: Recitation 6 Highlights: Convolution codes in factor graph form
10/14/2011: Recitation 6 Highlights: Nuances of factor graphs
10/14/2011: Recitation 6 Highlights: MAP Inference and computation of partition function are hard
10/14/2011: Recitation 6 Highlights: Review of complexity class terms
10/21/2011: Recitation 7 Highlights: The Viterbi algorithm
10/21/2011: Recitation 7 Highlights: Relating the Viterbi algorithm to the shortest path problem
10/21/2011: Recitation 7 Highlights: Node max marginals having unique maximal points implies a unique global MAP configuration
10/28/2011: Recitation 8 Highlights: Gaussian BP - High-level review
10/28/2011: Recitation 8 Highlights: Gaussian BP - Sparsity pattern of information matrix for a linear dynamical system
10/28/2011: Recitation 8 Highlights: Gaussians - Product of potentials in information form
10/28/2011: Recitation 8 Highlights: Gaussians - Sum of independent Gaussians is Gaussian
10/28/2011: Recitation 8 Highlights: Gaussians - Marginalization in information form
11/04/2011: Recitation 9 Highlights: Intuition for the junction tree property
11/04/2011: Recitation 9 Highlights: Loopy belief propagation and computation trees
11/18/2011: Recitation 10 Highlights: Why Sampling? - Justification of why sampling from p is useful
11/18/2011: Recitation 10 Highlights: Overview of the Metropolis-Hastings algorithm
11/18/2011: Recitation 10 Highlights: Overview of Gibbs sampling
11/18/2011: Recitation 10 Highlights: Particle Filtering through picture
12/02/2011: Recitation 11 Highlights: EM algorithm overview
12/02/2011: Recitation 11 Highlights: EM for HMM's -- Baum-Welch
12/10/2011: Maximum likelihood estimation for multivariate Gaussians
12/10/2011: Gaussian mixture models: intro and parameter estimation with EM