NeurIPS 2020

Deep Smoothing of the Implied Volatility Surface

Meta Review

We thank your for your submission. The reviewers felt that this paper presents an interesting problem for ML community and tackles it in a novel way. However, during the discussion period reviewers found that there are some major concerns with this paper: 1. The motivation is not clear. Volatility (sigma_i) is inherently unobserved quantity. However, the methodology presented here seems to rely on its availability. But if one has a mechanism to produce sigma_i's what is the need for this method? 2. Reproducibility. Experimental section (and appendix) seem to be missing a lot of details. We strongly encourage the authors to make the code (including data preprocessing part) open. Please refer to the reviews for details. Personally, I also found the MAPE+RMSE part of the objective poorly motivated. It felt like multiple things has been tried and this just seemed to work best. This being said, reviewers felt that novelty of the proposed problem and solution outweight this. Please take a serious effort to address these comments in your final version.