Computer-Oriented Approaches to Pattern Recognition

by William S. Meisel.

Introduction:
This is an enjoyable textbook first textbook in pattern recognition that is very easy to read and follow. The problems are enjoyable and interesting. I would recommend it highly. In addition, I have implemented many of the problems from Meisel's book in Matlab below.

Download Problem Solutions (Part 1 of 1).

Chapter 3 (Introduction to Optimization Techniques):
• prob_3_1.m (gradient decent comparing various learning rate schedules)
  
• prob_3_2.m (gradient decent on a non-smooth function)
  
• prob_3_3.m (gradient decent on a least squares problem)
  
Chapter 4 (Linear Discriminant Functions and Extensions):
• prob_4_3.m (search for a linear separating hyperplane with various learning rate schedules)
  
• prob_4_5.m (search for a linear separating hyperplanes for a three class problem)
  
• prob_4_6.m (search for a linear separating hyperplane using the many-at-a-time algorithm)
  
• prob_4_7.m (search for a linear separating hyperplane using the one-at-a-time algorithm)
  
Chapter 8 (Cluster Analysis and Unsupervised Learning):
• prob_8_3.m (the Sebestyen-Edie clustering algorithm)
  
• prob_8_4.m (the similarity matrix method for clustering)
  
• prob_8_6.m (an example for iterative cluster adjustment)
  
• Auxiliary routines developed for this chapter:
  • cluster_N_plot.m (iterative cluster adjustment)
    
  • mDSTLNN.m (average of the squared distance to the L-th nearest neighbor)
    
  • findNearestCluster.m (finds the nearest "cluster" based on the distance to each centroid the centroid normalized by the standard deviation)
    
Chapter 9 (Feature Selection):
• prob_9_3.m (project samples to minimize the ratio of the intraset (same class) distances over the interset (different class) distances)
  
• prob_9_4.m (a piecewise linear transformation such that the projected samples are linearly separable ... unfinished)
  
• prob_9_5.m (rank each feature using Fisher's criterion (the features mean difference over sum of their variances))
  
• Auxiliary routines developed for this chapter:
  • plotTruthFAHistograms.m (plots histograms of two features ... for visualization of discrimination performance)