EECS 391/491: Final Exam Information Sheet
Logistics
- The final exam will be held Tuesday, 06 MAY, 8:30-11:30AM, in Bingham 103.
- The exam will open book, open notes.
- Coverage: all lectures; assigned reading in R&N (Chs.
1-7, 11-14, 16-21, 26) and Braitenberg; Problem Sets 1-12; Mini PSs 1-2.
- Study help: read/review, problems/solutions, chapter summaries, formula sheets
- 491: you will have to do two extra problems;
extra material includes G problems and assigned readings
Topics in More Detail
- Everything on the Midterm Exam Information
Sheet
- Acting under Uncertainty
- Probability: axioms, joint/marginal/conditional, independence, normalization
- Bayes' Rule
- Bayes' Nets: causal graph and CPTs
- Exact inference
- Approximate inference
- Fuzzy Logic
- Forms of Learning: supervised, unsupervised, reinforcement
- Inductive learning: examples, hypothesis, hypothesis space, consistent, Occam's razor
- Decision Trees: information (entropy), stumps
- Decision Tree Issues: missing data, continuous/integer vars.
- Regression in general:train/test, least-squares fits, over-fitting
- Computational Learning Theory: ensemble learning, boosting, PAC
- Hypothesis space: consistent, FN, FP, generalization/specialization
- Statistical Learning: MDL, ML, MAP, Naive Bayes
- Instance-based Learning: NN, kernels/RBFs
- Neural Networks: activation functions, thresholds, hidden units, forward propagation
- Perceptrons/TLUs: linearly separable, Perceptron Learning Rule
- Plus: back-propagation, support vector machines, examples
- One-shot decisions: preferences, utility, maximum expected utility, dominance
- Sequential decision-making: MDP (state/action/reward), policy, value function, optimal policy, optimal value function
- Value and Policy Iteration: policy evaluation (linear eqns), greedy wrt value function
- Reinforcement Learning: passive policy evaluation (MC, TD), Q-function, Q-learning
- Plus: credit assignment problem, exploration/exploitation tradeoff, n-armed bandits, function approximation
- Game Theory: games as tables, mixed/pure strategies, zero-sum, minimax/saddle/Nash equilibria
Author: M.S. Branicky. Created: 2008-04-08.
Modified: 2008-04-08.