search

allenhaozhu / SSGC

Implementation for Simple Spectral Graph Convolution in ICLR 2021
Other
82 stars 17 forks source link

Where can I find the final version of the code and the final version of the paper? #5

Open coraler opened 3 years ago

coraler commented 3 years ago

As discussed in these issues, these codes and experimental results are not adopted in the final version of paper? So could you tell me .Thank you.

allenhaozhu commented 3 years ago

As discussed in these issues, these codes and experimental results are not adopted in the final version of paper? So could you tell me .Thank you.

Sorry this code was written when I submitted supplement materials. Thus the bug only affects parameter analysis in Table 8 and 9 basically. you can replace sgc_precompute function in utils.py as the following code: `def sgc_precompute(features, adj, degree, alpha): t = perf_counter() feature_ori = features feature_set = torch.zeros_like(features) for i in range(degree): features = torch.spmm(adj, features) feature_set += (1-alpha)features + alphafeature_ori

feature_set/=degree+1

      feature_set /= degree
      precompute_time = perf_counter()-t
      return feature_set, precompute_time`

Because new hyper-parameters need to update, so I need few days (I hope tomorrow to update Github).