The last paper from my PhD thesis got finally accepted at ICML!
Since this was a single author paper, this was the most difficult to publish from all other papers I have published in the past – even when comparing to the PRL paper that is supposed to be one of the most difficult venues to publish in physics. And even though I intentionally have been submitting the work to conferences with double blind reviewing policies, my older workshop papers, that were accepted during the end of my thesis, were revealing enough for the reviewers to know who was behind the paper, while the senior area chairs had full visibility anyways, and my threshold ratings were getting pushed (twice!) towards rejection ( some of them with extra push ;) ).
When I finished my PhD I was so fed up with that work that I thought I will just leave it there as a workshop paper, and focus on the new projects I started during my post-doc, that were new and more exciting. I figured I could probably publish part of this work together with my latent state inference project that came subsequently. But things didn’t turn as expected… It became clear that I would never see my post-doc papers at a timepoint that would be usuful for my career (and now it is questionable whether I will see any at all :) ), while my latent state inference project got semi-scooped.
So since I needed some time to work on new projects to publish, the only way forward was to push this work forward. Inbetween I actually noticed that more people in machine learning were taking into account the “geometry of the data”, which is one of the central intuitions of my approach, so I figured people might find it intresting. At the end it all ended fine!
I am really grateful to everyone who helped me for this, and especially to my PhD advisor, Manfred Opper, but also to the reviewers, who throughout this loong review process, shaped the resulting manuscript!