tong-zeng

Hi @IrvingZJU It seems the formula is different in the newest version of the paper, here you go https://arxiv.org/pdf/1703.04977.pdf

Hi @zhackzey I think as [Issue #3](https://github.com/Hui-Li/multi-task-learning-example-PyTorch/issues/3) reported, The uncertainty parameter is `σ^-2`, and the uncertainty penalty is `logσ`, when taking the exp in the code, `exp(-logσ) = σ^-1` which...

Hi @zhackzey Just realized that the author corrected the formula in a new version of the paper, https://arxiv.org/pdf/1703.04977.pdf

I'd also like to know how the thresholds of 1532 and 1200 are determined, thank you.