Ablation study for k (trajectories) and m (samples) parameters?

[sanowl]

Hi! Great work on MOTIF.

Quick question: Did you experiment with different values for k (number of trajectories) and m (number of samples) beyond k=4, m=8?

Would love to see how performance/computational cost varies with these parameters.

Thanks!

[purbeshmitra]

Hi @sanowl . No, we haven't experimented with different k, m values. The numbers we chose are just for convenience of training. You are right, it would indeed be worth doing a full ablation study.