Question about the implements of Beta_coverage

[zealscott]

Question

Hi, thanks for your great implementation! I have a small question about the evaluation metric of beta_coverage, which implements here:

synthcity/src/synthcity/metrics/eval_statistical.py

Lines 615 to 620 in ab47bcf

	beta_coverage = np.mean(
	(
	(real_to_synth <= real_to_real)
	* (real_synth_closest_d <= closest_synth_Radii[k])
	)
	)

I want to know why real_to_synth <= real_to_real is necessary to determine the coverage of each real sample? Intuitively, if the real to syn distance is smaller than the radii, it should be considered covered by the generative distribution. Why we need to consider the cloest distance for real-to-real?

Thanks and looking forward to your reply!