| A |
Probabilistic reasoning (FC)
- Slides (✎ OCT 26, ✎ OCT 31, ✎ NOV 04 and ✎ NOV 07) Last update NOV 07
- Exercises (Last updated NOV 16 - ✎ NOV 21 and ✎ NOV 23, with JC and N in LEC I)
Hand-in by NOV 27 at 23:59:59 Fortaleza time
- Results in [0,10] (Scores are PRELIMINARY: Submissions will be subjected to a 2nd round of evaluation)
|
- Probability refresher (Definitions, rules, tables, conditional probability)
- Probabilistic reasoning
- Prior, likelihood and posterior
|
| B |
Graph concepts (FC)
- Slides (✎ NOV 09) Last update NOV 09
|
- Definitions
- Numerical encoding (edge lists, adjacency matrices, clique matrices)
|
| C |
Belief networks (FC)
- Slides (✎ NOV 11, ✎ NOV 14 and ✎ NOV 16)
- Exercises (✎ NOV 28 and ✎ NOV 30, with JC and N in LEC I, and ✎ DEC 02 with FC)
Hand-in by DEC 11 (was DEC 04) at 23:59:59 Fortaleza time
Results in [0,10].
|
- Structure (independencies and specifications)
- Uncertain and unreliable evidence
- Belief networks (conditional independence, collisions, path manipulations for independence, d-separation, graphical and distributional in/dependence, Markov equivalence, expressibility of belief networks)
- Causality (Simpson's paradox, do-calculus, influence diagrams)
|
| D |
Graphical models (FC)
|
- Graphical models
- Markov networks (Markov properties, Markov random fields, Hammersley-Clifford theorem, Conditional independence using Markov networks, lattice models)
- Chain graphical models
- Factor graphs (Conditional independence)
- Expressiveness of graphical models
|
| E |
Inference in trees (FC) |
- Marginal inference (Variable elimination in a Markov chain and message passing, the sum-product algorithm of factor graphs, dealing with evidence, computing the marginal likelihood, loops)
- Forms of inference (max-product, finding the $N$ most probable states, the most probable path and the shortes path, mixed inference)
- Inference in multiply connected graphs (Bucket elimination, Loop-cut conditioning)
|
| F |
The Junction tree algorithm (FC) |
- Clustering variables (reparameterisation)
- Clique graphs (Absorption, absorption schedule on clique graphs)
- Junction trees (The running intersection property)
- Constructing a junction tree for singly-connected distributions (moralisation, forming a clique graph, forming a junction tree from a clique graph, assigning potentials to cliques)
- Junction trees for multiply connected distributions (triangulation algorithm)
- The junction tree algorithm (remarks on the algorithm, computing the normalisation constant of a distribution, marginal likelihood, examples, Shafer-Shenoy propagation)
|