I am trying to use Weka Multilayer Perceptron classifier to determine if a 14 by 14 picture is of a 4 or a 9. I am given 200 files text files containing a 0 or a 1 for each pixel of the picture being light or dark. I am supposed to convert each image in the dataset into a vector and create a Weka arff file using the converted vectors. My question is, is there some sort of arff vector attribute? The ridiculous arff I made with 197 attributes (including the class) and 200 data rows does output something but it is incredibly slow and I can't be sure of the result because of how strange the original file looks. What would be the best way to format the arff?
Asked
Active
Viewed 381 times
1 Answers
0
Why don't you write your own input reader if arff is too slow?
Weka is open source, you can change/extend it as you like.
Has QUIT--Anony-Mousse
- 76,138
- 12
- 138
- 194
-
Weka calculations on the data are slow, not the generation of the arff – Timothy Elbert Nov 26 '15 at 22:13
-
Weka is always slow, and you asked about ARFF, not about speeding up Weka (answer: rewrite) – Has QUIT--Anony-Mousse Nov 27 '15 at 06:53