Hello, thanks for the article. I wonder: in this line, "self.log_var = nn.Linear(self.num_hidden, self.num_hidden)", shall we use "log_var" or "log_std"?

In the original paper (Auto-Encoding Variational Bayes, Kingma et al., 2013), the authors said "where the mean and s.d. of the approximate posterior, μ and σ, are outputs of the encoding MLP". It means one should use "log std".

In another paper (Variational Graph Auto-Encoders, Kipf et al., 2016, note: same last author "Max Welling"), they also used "log std". The implementation here: https://github.com/tkipf/gae/blob/master/gae/model.py#L112

What do you think? 🤔