I have a question about Distance Weighted Sampling.

Open wjddyd66 opened this issue 4 years ago • 0 comments

First of all, thank you very much for coding as Pytorch.

I have a question in that code.

I have a question about the expression for getting log_weights in the Distance Weighted Sampling Class.

First, in this paper, the expression is written as follows: $q(d) \propto d^{n-2}[1-\frac{1}{4}d^2]^{\frac{n-3}{2}}$ $\text{log}(q(d)) \propto (n-2)\text{log}(d)+\frac{n-3}{2}{log}(1-\frac{1}{4}d^2])$

The current code is as follows. log_weights = ((2.0 - float(d)) * distance.log() - (float(d-3)/2)torch.log(torch.clamp(1.0 - 0.25(distance*distance), min=1e-8)))

Is this code correct to code the formula for giving weight?: $\text{log}(q^{-1}(d)) \propto (2-n)\text{log}(d)+\frac{3-n}{2}{log}(1-\frac{1}{4}d^2])$

Jul 12 '21 10:07 wjddyd66