Please enable JavaScript.

Coggle requires JavaScript to display documents.

ISLR (Unsupervised learning (Clustering (What is similarity? (Disà (Types,…

- - - - What are the principal components?
      - What are the assumptions?
      - other intepretations?
    - - How to scale the variables
      - uniqueness of components
      - Proportion of variance explained
  - - - Usually domain specific
      - Choice of dissimilarity measure
      - Types
        
        Correlation
        
        Euclidean distance
        
        Hamming Distance
        
        Manhattan Distance
    - - Hierarchical clustering
        
        Bottom up (agglomerative clustering
        
        Linkage
        
        Single Linkage
        
        Average Linkage
        
        Complete linkage
        
        Centroid
        
        Algorithm
        
        Dendogram
      - K-means clustering
    - - Small decision big consequence
        
        Should the observations be standardized?
        
        Hierarchical clustering
        
        what dissimilarity measure to use
        
        what type of linkage?
        
        where to cut the dendogram to get clusters?
        
        K means: how many clusters to look for?
        
        Can that be automated?
        
        Interpreting the results
      - How to validate the clusters obtained?
      - Other considerations
- - - - all nodes connected to each other
    - - a single connected component which contains most of the nodes in the network
    - - like strong, but direction ignored
    - - directed path from any node to any other
  - - - importance
        
        betweenness centrality
        
        Forms bridges between nodes
        
        Eigenvalue centrality
        
        If many other nodes link to it
- - - - can it be used to link between articles? :question:
- - - - sigmoid commute time kernel
      - logarithmic exponential diffusion
      - logarithmic laplacian diffusion kernel
      - randomized shortest paths dissimilarity matrix
      - Free energy distance
    - - reallocation
      - Recomputation
  - - - graph cut
      - multi-way ratio cut
      - multi-way normalised cut
  - - - 1 local optimisation
      - 2 Coarsening
- - - - Proxy to relative importance
    - - measures instituional and economic influences as wel as individual people, ideas and artifcats