more information about sketch-toy-i.txt file

[lucyzhou7735-pixel]

Hi authors,

Thank you for your excellent work!

While implementing the results, I was wondering how to obtain the file names using the following code:

file_name = DATASET + '3_hop_train_toy/sketch-toy-' + str(i) + '.txt'

Could you please provide more details or codes on how these files are generated? It would also be very helpful if you could share example files for /processed_1_hop_trace.csv and processed_uuid_trace.txt.

Thank you in advance for your guidance.

Many thanks.

[lucyzhou7735-pixel]

Also, I was wondering, what would be the data for python token_and_resample.py --semi_data_path='../data/cadet/full_data/graph/graph_data_sampled/' ?

[lucyzhou7735-pixel]

Sorry to bother again, how can we generate file with 'truth_label'?

[Samfisheryu]

Thanks for the question. For reproducing the released results, you do not need to generate the sketch-toy-*.txt files manually. The released data under 'train/data//*.json' is already in the final training format [x, y, z], where z is the GraphChi sketch vector merged into each sample.

The sketch_toy.zip files are the precomputed sketch-toy-*.txt GraphChi sketch outputs. We generated them using a modified GraphChi implementation from Unicorn. They are only needed if you want to rerun preprocessing from raw logs or apply TIDAL to your own dataset.