Naive question: why are people excited about normalizing flows/neural ODEs? Aren't they essentially a rehash of GANs that's even harder to train? I've been looking at many papers and still can't figure out what justifies their popularity, so would love some enlightened input!
@_onionesque Haha I like that explanation! I guess rehash is not well chosen, but I what I meant is I have seen people use it mostly for generative models and was wondering what their specific advantage was in that case.
@n_keriven@JuddLegum@fchollet 'left leaning' doesn't have the same meaning in France and in the US... just the basic fact of being pro healthcare or for free education makes you a dangerous socialist here 😬
Trying to convince myself that I don't care about #NeurIPS2020 decisions but I have already checked CMT at least 5 times today 'just in case' so I guess I'll stop lying to myself.
I was having a great saturday morning watching the magnificent Pyrénées on #TourdeFrance until Thibaut Pinot started falling behind in Port de Balès. So I'm going back to bed in total denial.
Wasserstein barycenters are useful in stats/ML but typical algorithms discretize the domain, leading to low-dimensional/coarse approximations. With PhD student Lingxiao Li, postdoc Aude Genevay @a_gnv, and @MITIBMLab's Misha Yurochkin we construct a continuous barycenter!🔥🎊🎈🎉
Wasserstein barycenters are useful in stats/ML but typical algorithms discretize the domain, leading to low-dimensional/coarse approximations. With PhD student Lingxiao Li, postdoc Aude Genevay @a_gnv, and @MITIBMLab's Misha Yurochkin we construct a continuous barycenter!🔥🎊🎈🎉
@JustinMSolomon absolutely disliked the review form for neurips this year... I felt like I had to repeat the same thing in different entries and couldn't structure my review properly
Positions confirmed! Still hoping to hear from more great candidates at the intersection of ML and geometric data. Please send me a quick email if you’re interested or have questions! And, share with any interested colleagues or students.
@n_keriven And here I thought lack of statistical significance was a big deal... this is revolting — how are people like him still allowed to lead research teams?