Please enable JavaScript.

Coggle requires JavaScript to display documents.

AI ((ANN (Types (feed forward ANN (can be used for general purpose models…

- - - - feed forward ANN
        
        i/p layer
        
        o/p layer
        
        one or more hidden layer
        
        good approximaters
        
        can be used for general purpose models
        
        regression
        
        classification
      - CNN
        
        modeled on visual cortex
        
        filters
        
        perform feature/pattern recognition task on data
        
        usage
        
        image ,voice ,video recognition
      - RNN
        
        hidden layer
        
        has memory
        
        has internal state
        
        usages
        
        handwriting recognition
        
        speech recognition
        
        log analysis
        
        fraud detection
        
        cybersecurity
        
        character generation
    - - Feedforward
      - Feedback
        
        content addressable memories
    - - Learning strategies
        
        Supervised Learning
        
        Unsupervised Learning
        
        Reinforcement Learning
      - Algorithms
        
        Back Propagation Algorithm
        
        learns by example
    - - connection
        
        pathway of info
        
        weights
        
        controls the signal
    - - graphical structures
        
        probabilistic relationship btwn random variables
  - - - distributed
        
        Cordianted Matrix
        
        Block Matrix
        
        IndexedRowMatrix
        
        Row Matrix
      - vector
        
        collect multiple vectors
    - - local
        
        Dense
        
        Sparse
        
        Breeze?
      - Vector Arithmetic
      - Easier Representation
        
        Column Vector
        
        Better for ML
        
        Row Vector
    - - R package
      - light-weight frontend
    - - graphs and graph-parallel computation
      - extends spark RDD
      - Directed multgraph
        
        properties <- Each vertex & edge
      - fundamental operators
        
        subgraph, joinVertices, and aggregateMessages
  - - - feature
        
        n dimensional vector
        
        individual measurable property
        
        usually numeric
        
        feature engineering
        
        Extract or select feature
        
        automated
        
        Feature learning
      - starts from measured data
        
        builds derived values #
  - - - PCA
      - CFA
  - - - regression
        
        quantitative
      - classification
        
        classes
      - Generalized Linear Models?
      - Linear and Quadratic Discriminant Analysis?
  - - - Open?
      - Closed
  - - - features and labels