wwydmanski @wwydmanski@qoto.org

Medycy z akcji #lekarzeDlaUkrainy pomogli już prawie 2000 uchodźców!
Jeżeli jesteś lekarzem/lekarką i chcesz dołączyć do inicjatywy, zarejestruj się tutaj: https://lekarzedlaukrainy.pl/pl/register

The original comparison of Adam vs AdamW is quite memorable.

On a related note, I absolutely recommend reading through reviews of that publication
https://openreview.net/forum?id=rk6qdGgCZ

#MachineLearning #DataScience

If you happen to be at @ISMBinfo conference, feel invited to drop by our poster "Clusters of interactions between operational taxonomical units in human gut #microbiome" about using #graphs and #NLP for function mining in taxons!
Done with @valentyn_bez, @pawel_labaj, and more

A tak zmieniał się średni #czynsz za mieszkanie 3 pokojowe w #Kraków w ostatnim czasie

https://flathub.pl/app/pl/blog/czynsz-za-mieszkanie-3-pokojowe-w-krakowie-w-2022
#nieruchomosci

#dalle's little sibling, #dallemini, while not potent as the full version is still capable of producing some interesting images.
Prompt: "happy microbe, digital art"

#Docker is an obvious tool in IT, but don't let your information bubble fool you - it's a very underused tool in #research.
Publishing your code on github is great for #replicability, but with docker you can go one step further and make it futureproof

Spotify seems to be dead set on making me a fan of the Doors, but you won't see me complaining
https://open.spotify.com/track/0JNZGIavoUrdup1NsgJOQs?si=

Try out your best approaches, and see you at the leaderboards!
BTW, just a reminder: @AI21Labs's models are legal in the #DILI challenge

#NLP #DataScience #conference
---
RT @CAMDA_conf
Final call for abstracts. Present your best work at @ISMBinfo: Take the Camda Challenge – Literature AI, Anti-microbial Resistance, #COVID19 Disease Maps – or share late breaking method papers! Extended abstracts 20 May. 👉http://camda.info @IARAInews @e…
https://twitter.com/CAMDA_conf/status/1524887674246545408

The deadline has been extended to 20th May.
Help @US_FDA in doing what comes after #drugdiscovery - keeping track of the drug safety.

Design an #NLP algorithm that will relieve humans of this task, submit it to CAMDA, and let's see each other on @iscb's ISMB!
#MachineLearning
---
RT @CAMDA_conf
Validation Leaderboards now live! Disrupt #DrugDevelopment by Literature #AI for #DILI: Build a digital twin of @US_FDA experts and identify r…
https://twitter.com/CAMDA_conf/status/1524379583658827776

Together with our friends from #FDA and David Kreil’s and Sepp Hochreiter's IARAI we've been working on an #NLP-oriented #challenge that is focused on tracking safety of the drug after clinical trials.
Consider a doctor that wants to help their patient through pharmacological treatment. Even if it doesn't cause any immediate adverse reaction, it still involves a major, long-term risk - #DILI, drug-induced liver injury.
Even though the #toxicity of the drugs is modelled and tested thoroughly during clinical trials, this method isn't 100% accurate and the danger related to them might be observed only later.
Right now, there is only one way of guarding against it in a real world scenario - constantly reading through publications regarding the drugs that you want to prescribe. Of course, there are many information banks which contain and aggregate this information, but they are still human-curated - which means that this problem isn't solved, it's just dumped on somebody else.
We want the participants to prove once and for all that humans can be relieved if this task.
The final, validation round has just began! You still have time until 16.05 to submit your results to ISMB 2022.

https://bipress.boku.ac.at/camda2022/
If you know anyone who might be interested, it'd be fantastic if you could spread the word.
#machinelearning

Together with FDA and @HochreiterSepp's IARAI we've been working on an #NLP-oriented challenge that is focused on tracking safety of the drug after clinical trials.
Submit before 16th may to join us at @CAMDA_conf 2022 in Wisconsin!
https://bipress.boku.ac.at/camda2022/

#MachineLearning

The second - that the model's github implementation _actually_ works out of the box! It should be a standard, but for some reason it rarely works for me

The first one is the dual #benchmark - comparing all models both default and tuned #hyperparameters.
Sure, it doesn't make much difference for production deployment of the model, but good defaults are very convenient during #EDA and early experiments

Popov et al.'s NODE might not be the newest #publication, but I love two things about it. 🧵

https://arxiv.org/abs/1909.06312

#GPT3 is cool, but if you're fine with a considerably cheaper bootleg version, then AI21's Jumbo is for you

https://studio.ai21.com/playground