Please enable JavaScript.

Coggle requires JavaScript to display documents.

AI (ML (Learn from examples & experience instead of relying on hard…

- - - - Instance Based Learning
        
        Examples
        
        Learning by analogy
        
        Assumes that if two situations are known to be similar in some aspects, it is likely that they will be similar in others
        
        Nearest neighbor (NN) learning
        
        1-Nearest Neighbor: To classify an object X, find similar instances to X in training data and assign X to the class of the nearest neighbor
        
        K-Nearest Neighbor Learning: Instead of comparing to only 1 NN, we consider K NNs
        
        Weighted regression
        
        Case-based reasoning
      - Evolutionary Learning
      - Example
        
        scikit-learn.org, Google's TensorFlow,
        Anaconda, Iris data, Re-training,
        Google's Inception,
        
        Step 1: Collect Training Data (Features are inputs to the classifier + Labels are the outputs of the classifier)
        
        Step 2: Use above data to train a classifier
        
        Step 3: Learning Algorithm. (clf = clf.fit(features, labels) (fit = find patterns in data)
        
        Step 4: Test classifier using "Test Data"
        
        Classifier (imagine: a set of rules)
        
        Types of Classifiers
        
        Decision Tree
        
        K-Nearest Neighbor (see Euclidean Distance)
        
        Neural Network (e.g. TensorFlow Playground)
        
        Deep Learning
        
        How are classifiers made? Using a model (e.g. straight line)
        
        Methods
        
        fit
        
        predict
  - - - Discovery
      - Pattern recognition
      - Clustering
        
        Types
        
        Heirarchical clustering
        
        Conceptual clustering
        
        Partitional clustering
        
        K-means Clustering
        
        Each cluster is associated with a centroid (center point).
        
        Each point is assigned to the cluster with the closest centroid.
        
        Applications
        
        Pattern or Object detection
        
        Outlier detection
        
        Dimensionality reduction
  - - - Active Reinforcement Learning (Goal: Learn model, come up with probabilities)
      - Passive Reinforcement Learning (Goal: Evaluate given policy(
- - - - Backtracking
      - Breadth Space Search
      - Depth First Search
    - - Goal Driven ~ Backward : Take the goal to solve and see what rules could be used to generate this goal and determine what conditions must be true to use them
      - Data Driven ~ Forward : Begin with given info, states, legal moves or rules, search by applying rules to produce new states until goal state is found
    - - AND op ~ problem decomposition
      - OR op - alternative solution paths
    - - N - set of nodes or states
      - A - set of arcs between nodes
      - S - non empty subset of N, contains the start state(s)
      - GD - non empty subset of N, contains the goal state(s)
    - - S - finite set of states in a connected graph
      - I - finite set of input values
      - F - state transition function
  - - - Predicate Calculus
        
        Modus Ponens: If the sentences P and P -> Q are known to be true , then infer Q
        
        Modus Tollens: If P -> Q is known to be true and Q is false, we can infer ~P
        
        Elimination: P ^ Q lets us conclude P and Q are tue
        
        Introduction: If P and Q are true, then P ^ Q is true
        
        Universal Instantiation: If any universally quantified variable in a true sentence is replaced by any appropriate term from the domain, the result is a true sentence. If a is from the domain of X, for every X in p(X) lets us infer p(a).
      - Sentence formation
        
        Disjunction (OR)
        
        Negation
        
        Implication (A -> B)
        
        Equivalence (E)
        
        For all X (Reverse A)
        
        For Every X (Mirror E)
        
        Conjunction (AND)
      - PROLOG
      - Propositional Calculus (Simple)
    - - Inductive Reasoning
        
        Allows inferring A from B, where A does not follow necessarily from B
      - Analogical Reasoning
      - Logical Reasoning
        
        Deductive
        
        Derives A (conclusion) from B (assumption) only where A is a formal logical consequence of B (given the truth of B). From All X (man(X) -> mortal(X)) and man(socrates) we derive mortal(socrates).
        
        Monotonic
      - Model-based Reasoning
      - Case-based Reasoning
      - Abductive Reasoning
        
        Allows inferring B as an explanation of A. From B -> A and A, it is possible to infer B.
      - Statistical Reasoning