Uri Shalit Profile picture
Machine learning researcher, working on causal inference and healthcare applications. Associate prof @TechnionLive @UriShalit@mastodon.social @urish.bsky.social
Nov 8, 2021 15 tweets 5 min read
Happy to share that our paper “On Calibration and Out-of-domain Generalization” is accepted to #NeurIPS2021!

Congratulations to the wonderful students who came up with the idea and led the work on this paper:
@wald_yoav @amir_feder @d_greenfeld

arxiv.org/abs/2102.10395

1/15 tl;dr: Making a classifier calibrated over multiple training domains is an easy and powerful way for better generalization to unseen domains
2/15
Jul 6, 2021 9 tweets 2 min read
You may have seen that the Israeli MoH claims that Pfizer vaccine efficacy has dropped to 64% in Israel, concurrent with the rise of the Delta variant.
The MoH has now published more details on the methodology. Briefly: negative binomial regression controlling for week and age. In my opinion this leaves *a lot* of potential confounders unaccounted for. Vaccination rates in Israel have converged to 70-90%, with older ppl generally more vaccinated.
Jan 31, 2021 12 tweets 5 min read
In recent days two new pieces of evidence have come out of Israel about the effect of the vaccination drive, showing good news about both individual-level and national level effects.
1/7
One is a study from Maccabi HMO (link further in the thread).
The other is one I’m a part of w/ @H_Rossman @GorfineMalka and @segal_eran , and is explained in the thread here:

2/7
Jan 20, 2021 19 tweets 8 min read
I’m seeing many discussions about the evidence for vaccine effectiveness in Israel. This is a thread with my thoughts on what we know and don’t know at this point. First some vaccination statistics:

1/18 Israel has been vaccinating at a fast pace - by today 78% of people aged 60+ have received at least 1 dose, 58% of 60+ are >14 days from their 1st dose, and 30% already received their 2nd dose.

2/18
Mar 18, 2020 9 tweets 4 min read
We’ve noticed something that looks like a potentially important data-entry problem regarding comorbidities in the “big” China CDC report on the Epidemiological Characteristics of COVID19. This report uses data from 72,314 patient records. (1/6) The China CDC report has the most widely used numbers I’ve seen for age and sex fatality rates. It was reported in brief form in JAMA, and fully in a China CDC weekly (link below, it loads slowly sometimes).
weekly.chinacdc.cn/en/article/id/…

(2/6)