Question on Loss Computation

[anish-pratheepkumar]

Hello,

First of all, thank you for the really nice work on human motion prediction and for sharing the same with the open-source community!

I have a question about loss computation. The previous works like the one by 'Julieta Martinez Et al.' (https://github.com/una-dinosauria/human-motion-prediction) use Eucledian distance (L2 norm) for computing the evaluation error and the results are also presented in the same metric.

Could you please help me understand the idea behind the L1 distance mentioned in your paper (section 3.3 Short-Term Prediction)?

Thank you!