Advance AI notes

By the end of the semester the students will be able to:
understand the search and decision making techniques used in modern artificial intelligence
apply artificial intelligence techniques in their own code
understand the societal and ethical implications of artificial intelligence


History and positioning of AI
Motivating AI. Dangers of AI and AGI.
Early history
Expert systems

Uninformed search
Reflex agents
Search problems
Depth first and breadth first search
Uniform cost search

Informed search: A* search and heuristics
Informed search methods
Heuristics
Greedy search
A* search
Graph search

Game playing and adversarial search
Types of games
Adversarial search, minimax
The problem of depth
Evaluation functions
Alpha Beta pruning

Expect Imax search and utilities
Expect Imax search
Refresher about probabilities
Utilities and rationality

Markov decision processes 1
Defining MDPs: policies and utilities
Optimal policy, value of state, value of Q-state

Markov decision processes 2
Policy iteration

Reinforcement learning 1
Reinforcement learning as a twist on MDPs

Reinforcement learning 2
Exploration vs. exploitation, regret
Generalization across states
Policy search

Probability
Random variables
Joint and marginal distributions, conditional distribution

Markov models
Markov chains
Conditional independence
Stationary distributions

Hidden Markov models
Hidden Markov models
Example: robot localization

Particle filters and applications of HMMs
Particle filters
Robot localization with particle filters
Dynamic Bayes nets

Classification, principles of machine learning, naïve Bayes
Classification
Model-based classification
Naive Bayes
Spam filter example
Generalization and overfitting
Parameter estimation

Introduction to deep learning
History and impact

Machine learning background of deep learning
History and impact
Machine learning background
Loss functions: squared, cross-entropy, soft max
Optimization, stochastic gradient descent
Backpropagation

Feedforward neural networks
Feedforward networks
Stochastic gradient descent

Convolutional neural networks
Convolutions
Convolutional filters in neural networks
Pooling layers