Module 6: Dimensionality Reduction
  Lecture 28: Dimensionality Reduction
 

                                            

 

 

Reduction to one dimension
  • Hashing
  • Space-filling curves
  • Choosing single dimension
 
  • Choose dimension with largest spread or largest variance
  • Pruning and proximity-preserving properties are generally violated
  • Loss of information and distortion are too high
  • In general, not very useful