5
Project 2 Algorithm General Overview of Algorithm: Iterative Process (20x): q=(I-PUU)/PUL q=q*Y Iterative Process (70x): Grab highest value unlabeled data for each labeled category. Remove from unlabeled data (PUL) and put in labeled data (Y). Touch up of True_LL data.

Project 2 Algorithm

Embed Size (px)

DESCRIPTION

General Overview of Algorithm: Iterative Process (20x): q=(I-PUU)/PUL q=q*Y Iterative Process (70x): Grab highest value unlabeled data for each labeled category. Remove from unlabeled data (PUL) and put in labeled data (Y). Touch up of True_LL data. Project 2 Algorithm. - PowerPoint PPT Presentation

Citation preview

Page 1: Project 2 Algorithm

Project 2 Algorithm

General Overview of Algorithm:

Iterative Process (20x):

q=(I-PUU)/PULq=q*Y

Iterative Process (70x):

Grab highest value unlabeled data for each labeled category.Remove from unlabeled data (PUL) and put in labeled data (Y).

Touch up of True_LL data.

Page 2: Project 2 Algorithm

Setup Step

Y=zeros(1330,70);c=zeros(70,1); %temporary variable used later

d=zeros(70,1); %number of new data for each category

...for(q1=1:70)

Y(q1,q1)=1;end..q=(I-PUU)/PULq=q*Y

Page 3: Project 2 Algorithm

Grab Highest Value in Category

For all 70 categories:max=0Run through all available unlabeled data

and find highest using max.Use 70 long hash vector for appropriatevalue and increase d(i), where i is thecategory number

Page 4: Project 2 Algorithm

Remove from Unlabeled, Add to Labeled

Basically remove the appropriate column for PUL, and add the value to the returned z value, which is based on True_LL.

Page 5: Project 2 Algorithm

Touchup

for i=1:1330if(z(i)==0&&i>1)

z(i)=z(i-1);end

end