AI CS Core

AI-A: Fundamental Issues

  1. Overview of AI problems, Examples of successful recent AI applications
  2. Definitions of agents with examples (e.g., reactive, deliberative)
  3. What is intelligent behavior?
    1. The Turing test and its flaws
    2. Multimodal input and output
    3. Simulation of intelligent behavior
    4. Rational versus non-rational reasoning
  4. Problem characteristics
    1. Fully versus partially observable
    2. Single versus multi-agent
    3. Deterministic versus stochastic
    4. Static versus dynamic
    5. Discrete versus continuous
  5. Nature of agents
    1. Autonomous, semi-autonomous, mixed-initiative autonomy
    2. Reflexive, goal-based, and utility-based
    3. Decision making under uncertainty and with incomplete information
    4. The importance of perception and environmental interactions
    5. Learning-based agents
    6. Embodied agents
      1. sensors, dynamics, effectors
  6. AI Applications, growth, and Impact (economic, societal, ethics)

AI-B: Search

  1. State space representation of a problem
    1. Specifying states, goals, and operators
    2. Factoring states into representations (hypothesis spaces)
    3. Problem solving by graph search
      1. e.g., Graphs as a space, and tree traversals as exploration of that space
      2. Dynamic construction of the graph (you’re not given it upfront)
  2. Uninformed graph search for problem solving
    1. Breadth-first search
    2. Depth-first search
      1. With iterative deepening
    3. Uniform cost search
  3. Heuristic graph search for problem solving
    1. Heuristic construction and admissibility
    2. Hill-climbing
    3. Local minima and the search landscape
      1. Local vs global solutions
    4. Greedy best-first search
    5. A* search
  4. Space and time complexities of graph search algorithms

 

AI-C: Fundamental Knowledge Representation and Reasoning

  1. Types of representations
    1. Symbolic, logical
      1. Creating a representation from a natural language problem statement
    2. Learned subsymbolic representations
    3. Graphical models (e.g., naive Bayes, Bayes net)
  2. Review of probabilistic reasoning, Bayes theorem (cross-reference with DS/Discrete Probability)
  3. Bayesian reasoning
    1. Bayesian inference

 

 

AI-D: Machine Learning

  1. Definition and examples of a broad variety of machine learning tasks
    1. Supervised learning
      1. Classification
      2. Regression
    2. Reinforcement learning
    3. Unsupervised learning
      1. Clustering
  2. Fundamental ideas:
    1. no free lunch
    2. undecidability
    3. sources of error in machine learning
  3. Simple statistical-based supervised learning such as Naive Bayes, Decision trees
    1. Focus on how they work without going into mathematical or optimization details; enough to understand and use existing implementations correctly
  4. The overfitting problem and controlling solution complexity (regularization, pruning – intuition only)
    1. The bias (underfitting) – variance (overfitting) tradeoff
  5. Working with Data
    1. Data preprocessing
      1. Importance and pitfalls of
    2. Handling missing values (imputing, flag-as-missing)
      1. Implications of imputing vs flag-as-missing
    3. Encoding categorical variables, encoding real-valued data
    4. Normalization/standardization
    5. Emphasis on real data, not textbook examples
  6. Representations
    1. Hypothesis spaces and complexity
    2. Simple basis feature expansion, such as squaring univariate features
    3. Learned feature representations
  7. Machine learning evaluation
    1. Separation of train, validation, and test sets
    2. Performance metrics for classifiers
    3. Estimation of test performance on held-out data
    4. Tuning the parameters of a machine learning model with a validation set
    5. Importance of understanding what your model is actually doing, where its pitfalls/shortcomings are, and the implications of its decisions
  8. Basic neural networks
    1. Fundamentals of understanding how neural networks work and their training process, without details of the calculations
  9. Ethics for Machine Learning
    1. Focus on real data, real scenarios, and case studies.
    2. Dataset/algorithmic/evaluation bias