Clustering task with t2vec

[zealscott]

Hi,

I have a cluster task with porto dataset. I use your best_model.pt to get the trajectory representation(about 1w data), and use kmeans to get cluster result.

But compare with other traditional measurement(like LCSS/EDR), the acc/nmi is really bad (0.2 vs 0.6), I am confused about the results. Can you tell me why could it happen?

Thanks a lot!