search

zhilingluo / NgramCNN

A Ngram based convolution neural network for graph objects classification
8 stars 6 forks source link

Query regarding Data Augmentation in prepossing of dataset using NgramCNN model. #1

Closed sahertariq07 closed 5 years ago

sahertariq07 commented 5 years ago

I'm using NgramCNN mode to train on chemoinformatics dataset. As reported in the paper that graphs can be augmented in preprocessing stage. But unable to produce extend graph number. So i tried to change "duplicate_limit" parameter in the code and still nothing change. I have also tried on the given example "mutag dataset". Output is given below. Dataset: mutag Operation: 1 Parameters: Kernel width: 7 graph num is 188 max node: 28 max label: 2 max degree: 1.2222222222222223 min degree: 1.0 avg degree: 1.0943860392762486 max node: 28 min node: 10 avg node: 17 max edge: 33 min edge: 10 avg edge: 19 org graph number 188 extend graph number 0 start writing disk done

Kindly tell me how to extend graph number using this model as reported. Any suggestions will be appreciated in advance.

sahertariq07 commented 5 years ago

I got a mistake in code that it prints "extend graph number 0" because wrong variable is given to print. So I'm going to close this Issue.