search

sbustreamspot / sbustreamspot-train

Convert training graphs to shingle vectors and compute the best clustering
Apache License 2.0
1 stars 4 forks source link

An error when convert the StreamSpot training graphs to shingle vectors #5

Open yemai666 opened 1 year ago

yemai666 commented 1 year ago

image Hi，I have some questions as shown below。 Train_data.csv is that I selected all edges of 240 benign graphs, about 1.2GB of data, I wanted to use them for training, but there was a mistake in the figure above, could you please tell me why?

yemai666 commented 1 year ago

The purpose is that I want to use 240 benign graph for train instead of 375。 The Train_data.csv has 6476280 edges of 240 benign graph i chose，Is it because I don't need to input all the edges of the 240 graphs?

yemai666 commented 1 year ago

I am using 64GB of memory，but this error still occurs

emaadmanzoor commented 1 year ago

Ah this usually happens because there is not enough contiguous memory on your machine for allocation. The best way to be sure is to look at the core dump or debug using GDB. This code is also quite old: it might be possible to rewrite it more efficiently today.

yemai666 commented 1 year ago

Thanks for your reply! Am I doing the right thing by using all the edges of the 240 graphs for training? And how much memory did you use for the experiment with 75% data

emaadmanzoor commented 1 year ago

I don't remember since I last use this a while ago, sorry!