Image Classification Using Biomimetic Pattern Recognition with Convolutional Neural Networks Features

Pseudocode 1

Pseudocode of the CNN-BPR algorithm.

(1) TRAINING PROCESS

INPUT: labeled training data as , K is the total of classes.

; % the raw training data are sent into CNN to get extracted feature vectors

; % the extracted feature vectors are mapped into high-dimensional space to

be % covered by CGC class by class

for i 1 to K do

  ; % Calculate the distance between any of two points in class

  ; % find the closest two points from D ⁽ⁱ⁾, marked as and

  ; % delete the marked points

; % FindPtoN is a function used to find the minimum distance sum

    % from F to and

; %  , and constitute the first plane triangle θ ₁

    % P ₁ is the coverage of with the covering radius called ψ3

    % neuron, and indicates the distance between and

;

; % ExcludeP is a function used to exclude points from F ⁽ⁱ⁾ covered by

;

 while   % repeat the steps above until F ⁽ⁱ⁾ is empty

  ;

  ;

  ;

  ;

 end

; % the final CGC of class is the union of each ψ3 neuron

end

OUTPUT: ; % the set of CGC of all classes

(2) CLASSIFICATION PROCESS

INPUT: is an image to be classified

;

; %   is the distance between and the coverage of neuron in class

OUTPUT:   % the class that belongs to