Please enable JavaScript.

Coggle requires JavaScript to display documents.

4th Data Analytics (Machine learning and pattern recognition (Pattern…

- - - - Pattern
        
        A description of an object
        
        A set of measurements
        
        A set of features
      - Recognition
        
        Classifying an object to a predefined pattern
      - Pattern Recognition
        
        Pattern recognition is an important component of intelligent systems. They are used for
        
        Decision making
        
        Object and Pattern classification
        
        Data Preprocesing
        
        Applications
        
        Image processing, Segmentation, and Analysis | Computer vision | Radar signal analysis and classification | Face recognition | Speech Recognition | Fingerprint identification | Character recognition | Handwriting analysis | Cardio signal analysis | Medical diagnosis | Data mining | Image databases | Seismic (earth-quick) analysis
        
        Commercial pattern recognition machines
        
        Automated speech recognition | Fingerprint Identification | Optical character Recognition | DNA sequence identification | Blood cells | Printed text | Iris pattern recognition
    - - Feature extraction
        Transforming the existing features in to a lower dimensional space
        
        The feature extractor should characterize the object such that the feature values of each class are very different from others.
        
        But in practice this requirement is not met since
        
        Features tend to over represent training data
        
        Thus the noise is also modeled to the classifier with the useful information
        
        Feature extraction methods
        
        Template Matching
        
        Contour Profile
        
        Zoning for binary contours
        
        For each zone of the contour is followed and a directional histogram is obtained by analyzing the adjacent pixels in 3x3 neighborhood
        
        Limitation: when the contour curve was close to zone borders, small variations in the contour curve could lead to large variations in the extended features
        
        Projection Histogram
        
        Horizontal
        
        Vertical
        
        Zoning Methods
        
        Skeleton Methods
        
        Zoning for skeletons
        
        Steps
        
        Distinguish individual line segments
        
        Labeling line segment information
        
        Line type normalization
        
        Formation of feature vector through zoning
        
        Limitation
        Features cannot be made rotation independent
        
        Further improvements
        The presence or absence of junctions or endpoints in each zone can be used as additional features
        
        Zhang-Suen Thinning Algorithm
      - Feature
        
        Feature is distinctive aspect, quality or characteristic
        
        Pattern is a composite of traits or features characteristic of an individual
        
        In classification, a pattern is a pair of variables
        {x,w} where
        
        x is a collection of observations or features (feature vector)
        
        w is the concept behind the observation (label)
        
        Feature vector
        a collection of features
        
        What makes a "good" feature vector
        The quality of a feature vector is related to its ability to discriminate examples from different classes
        
        Examples from same class should have similar feature values
        
        Examples from different classes have different feature values
        
        Feature properties
        
        Liner separability
        
        Non-linear separability
        
        Multi modal
        
        Highly correlated features
      - Classifier
        
        Classifier makes decisions using features about the class that it belong to
        
        The goal of the classifier is to partition feature space in to class labeled decision regions.
        Borders between decision regions are called decision boundaries
      - Feature Selection
        Selecting a subset of the existing features without a transformation
        
        Search methods
        
        Branch and Bound algorithm
        
        https://youtu.be/hG7IVK_waNQ
        
        Beam Search
    - - The design of an automatic pattern recognition system general involves several major problem areas.
        
        Concerned with the representation of input data which can be measured from the objects to be recognized
        
        Concerns the extraction of characteristic features or attributes from the received input data
        
        Reduction of the dimensionality of pattern vectors
        
        Determination of decision procedures which are needed in the identification and classification process
- - - - Apriori Algorithm
        
        https://youtu.be/Hk1zFOMLTrw
        
        Explains the first example in lecture3
        
        For many frequent-itemsets algorithms, main-memory bottleneck is critically concerned because we read baskets, and then count a huge number of different itemsets as well as swap counts in/out. Therefore, frequent itemset generation is still computationally expensive.
        A first candidate to resolve these problems is A-Priori algorithm.
        
        Generating association rules
        
        Apriori scans the database several times, depending on the size of the longest frequent itemset
      - Partitioning Algorithm
      - Sampling Algorithm