Adam Oberman

#professor #appliedmath #DeepLearning #AdversarialNetworks

[Adam Oberman @oberman_adam]

@SimonPrinceAI Hi, I just found your book, and I'm using the chapters on Generative Modeling to teach a few lectures in my course. Thanks!

[Adam Oberman @oberman_adam]

Behind the door-to-door campaign viewed as key to raising Montreal's vaccination rate
https://t.co/f4fsgmAP5i by ⁦@AlisonNorthcott⁩

[Alison Northcott @AlisonNorthcott]

A look inside one of Montreal’s ICUs. Grateful to the patients and health-care workers who shared their stories. w/ @JessRubinger and cameraman @Martelsi

[André Picard @picardonhealth]

From Jim Crow to Jim Code: Using artificial intelligence in healthcare is alluring, but we must keep algorithmic #racism out of health care’s #AI toolkit, by @AlbergaHannah
https://t.co/oPFktclHjX via @GlobeDebate #cdnhealth #cdnpoli #SystemicRacism

[CIFAR @CIFAR_News]

Canada welcomes 29 newly-appointed Canada CIFAR AI Chairs at @AmiiThinks @Mila_Quebec @VectorInst - congratulations to all! #AICan
https://t.co/dBhAj5PtVp

[McGill Science @McGillScience]

Congratulations to the four McGill researchers recognized by the @RSCTheAcademies, including oceanographer Alfonso Mucci, who was awarded the Miller Medal, and statistician Christian Genest, recipient of the Synge Award.
@EPS_McGill
https://t.co/RtjqTX5bBE

[Justin Solomon @JustinMSolomon]

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!🔥🎊🎈🎉

[Wendell Potter @wendellpotter]

Amid America's #COVID19 disaster, I must come clean about a lie I spread as a health insurance exec: We spent big $$ to push the idea that Canada's single-payer system was awful & the U.S. system much better. It was a lie & the nations' COVID responses prove it. The truth: (1/6)

[Adam Oberman @oberman_adam]

@im_td Thanks for mentioning this!

[Tim Davidson @im_td]

Regularized Neural ODEs (RN-ODE) - elegant improvement for one of the more exciting research directions in ML .. proposes two simple loss regularization terms to encourage well-behaved solutions .. cool work!

paper: https://t.co/la8IWygpeE

[Adam Oberman @oberman_adam]

https://t.co/JEMulfaSYA

[Adam Oberman @oberman_adam]

Excited about our new paper: Adversarial robustness which scales to ImageNet Scaleable input gradient regularization for adversarial robustness https://t.co/jGmlKm7Yhq https://t.co/ho9eLzl8GG

[Erik Brynjolfsson @erikbryn]

"By changing a few pixels on a lung scan, for instance, someone could fool an A.I. system into seeing an illness that is not really there"

#GANs #MachineLearning
#adversarialnetworks #2MA

https://t.co/Zjzn6MDB9O

[Adam Oberman @oberman_adam]

@rsalakhu Do the reviewers see the author comments before then?

[Denny Britz @dennybritz]

A good summary of criticisms of Deep Learning for Vision (https://t.co/TCcC6cxDCq). IMO one overarching issue is that research is done to beat benchmarks and publish papers, rarely “regularized” by real-world problems with data characteristics that may be significantly different.

[Gautam Kamath @thegautamkamath]

Let me add the @SimonsInstitute workshop to this thread, which has a ton of great videos: https://t.co/TlV5BuySsw 4/3

[Gautam Kamath @thegautamkamath]

Ilias Diakonikolas and Daniel Kane organized a workshop at @TTIC_Connect, and the slides are now uploaded (https://t.co/scqZ5Ezr0F). Ilias also gave a tutorial at @SimonsInstitute (video & slides https://t.co/M20qug9fxt) 3/3

[Adam Oberman @oberman_adam]

@ylecun For good theory, we need precise questions. What matters besides accuracy? Adversarially robust? Compressible? Interpretable? Confidence? Any pointers to a review discussing open problems?

[Adam Oberman @oberman_adam]

@SebastienBubeck Looking forward to seeing the papers. Thanks for your hard work!