wwydmanski @wwydmanski@qoto.org
AI guy. Tooting interesting publications and statistics that catch my eye
Joined May 2021
*sometimes*
---
RT @GregBodwin
In academia you sometimes have to work 7 days a week, but the freedom to choose which 7 days is unparalleled
https://twitter.com/GregBodwin/status/1408576508634009603
As an unordered service, we have also thoroughly disinfected two random spots on each of our swabbing locations.
#gCSD2021 with @metasub is nearly over!
---
RT @pawel_labaj
Most of @MCB_UJ @JagiellonskiUni teams are done with @metasub #gCSD2021 in @krakow_pl Thank You All!!!
https://twitter.com/pawel_labaj/status/1406982443077357573
RT @MCB_UJ
Handrails in our cities can sometimes have more biodiversity than rainforests. Will we learn about new species by studying Kraków's microbial fingerprint during global City Sampling Day 2021? Look out for MCB researchers on the streets, Monday June 21st.
If you've ever wondered whether licking the handrails can be good for your health, the aftermath of #gCSD2021 should be of interest to you.
We're going to check that so you don't have to - together with @mason_lab, @MCB_UJ, and @JagiellonskiUni
Meta-dataset is a nice way of testing how robust your new, groundbreaking model is _actually_.
https://github.com/google-research/meta-dataset
Splitting a dataset into training and test set is cool and all, but but did you try splitting your metadataset into training and test datasets?
If not, don't worry, Google Research has got you covered.
RT @dermayank
Difference between machine learning and AI:
If it is written in Python, it's probably machine learning
If it is written in PowerPoint, it's probably AI
Similarity of concepts of a knowledge graph is a frequently encountered problem in text analysis for which Zhu et al. propose an approach that goes beyond the casual path calculation.
If you need more resolution of your metrics you should give it a try:
RT @IARAInews
Dear competitors! We are inviting you to join the forum for our #Traffic4cast21 competition! There you can already find the competition details with first explorations on the temporal shift due to Covid-19 as well as on the spatial properties and shift.
👉https://www.iarai.ac.at/traffic4cast/join-us-for-access/?redirect_me=https://www.iarai.ac.at/traffic4cast/forums/forum/competition/competition-2021/
The dynamic time warping is a technique that is often used for alignment of time series data, with many usages in areas such as biometry.
Surprisingly, it can also be used in computing text similarity!
Let Matuschek et al. tell you about it:
https://www.researchgate.net/publication/221524439_Measuring_text_similarity_with_dynamic_time_warping/figures
RT @mason_lab
Cities have their own distinct microbial fingerprints @metasub https://www.sciencemag.org/news/2021/05/cities-have-their-own-distinct-microbial-fingerprints
RT @mason_lab
I'm elated to release four @metasub papers today, with new species, data interfaces, & online resources. The first is in @CellCellPress, "A global metagenomic map of urban microbiomes and antimicrobial resistance" led by @dcdanko @Daniela_Bezdan https://www.cell.com/cell/fulltext/S0092-8674(21)00585-7" a 🧵 /1
RT @MCB_UJ
Global study of 60 cities’ microbes finds each has a signature microbial fingerprint. A largest-ever global metagenomic study of urban microbiomes.
https://www.cell.com/cell/fulltext/S0092-8674(21)00585-7#secsectitle0360
@krakow_pl @MPK_Krakow @razemwruchu @metasub @pawel_labaj @mason_lab
RT @pawel_labaj
There is still time to take part in #CAMDAchallenge! Don't miss the deadline on May 31st! meet @kyrpides @fcicca on July 28-30 http://camda.info discuss with @dpkreil @xdopazo @mason_lab @TomaszLabMCB @RafalMostowy @metasub about #COVID19 #AMR #Cancer #metagenomics #AI https://twitter.com/CAMDA_conf/status/1397226041148645393
The validation data for the "Literature #AI for Drug Induced Liver Injury" #CAMDAchallenge is live!
Feel free to join! If you've got any questions you can ask them here or on our forum.
Jay Alammar's overview of BERT-based models shows us some not so obvious conclusions - like the fact that the best way of getting a contextual word embedding isn't taking the output layer of the model.
Our article with @SzymonSzott about deep reinforcement learning for contention window optimization is available in the IEEE Xplore repository.
RT @josh_tobin_
In the real world, data distributions are almost never static.
So if you want to use ML in production, your goal should be to build a continual learning machine, not a static model.
AI guy. Tooting interesting publications and statistics that catch my eye
Joined May 2021
