Want to tinker with the 200M+ protein structures from AlphaFold in your favorite DataFrame format (pandas!)⁉️
— Sebastian Raschka (@rasbt) August 6, 2022
Thanks to a kind contribution by @arian_jamasb, this is now possible in the latest BioPandas dev version as of this morning!https://t.co/CO6DnCcEb2 https://t.co/K4edTcqs27 pic.twitter.com/2fO7bBv172
When it comes to tabular data
— Radek Osmulski 🇺🇦 (@radekosmulski) August 5, 2022
✅ it is equal parts art and science at the highest of levels
✅ who can run better experiments and better capture their outcomes, wins
A stellar talk by @Giba1!
Come for high-level insights, and stay for a great explanation of unique techniques! pic.twitter.com/YXC7zhtC1S
You are likely using DistilBert vs Bert due to the smaller memory footprint.
— Sebastian Raschka (@rasbt) August 5, 2022
But check this out:
🦾 7x speed-up via quantization (32 bit floats -> 8 bit ints) alone.
Cost? A meager <1% F1 score decrease 🤷♂️
If you are not already quantizing... https://t.co/xoXE0BbUW4 pic.twitter.com/GLRff3jALE
Such an inspiring reply to the shallow-minded question, “What's the point of being a tenured professor compared to being a research scientist in top companies and groups like DeepMind?” (https://t.co/PcA1yFSaaS)
— hardmaru (@hardmaru) August 5, 2022
Wonder why they deleted their account. https://t.co/oOU8vVXvP8 pic.twitter.com/7BOeZ3h3lq
A very good paper I came across this morning by the @DeepMind researchers. For the past five years Transformers have been one of the most dominant approaches to Deep Learning problems, especially in the #NLP domain.
— Bojan Tunguz (@tunguz) August 5, 2022
1/5 pic.twitter.com/XRQODHdQn3
One of the coolest things I've learned in the last few years is streaming.
— Chip Huyen (@chipro) August 5, 2022
I used to find it confusing. Phrases like “time-variant results”, “time travel”, “materialized” didn't help.
This post is my understanding of streaming. Feedback appreciated!https://t.co/6dPJ1My73Y pic.twitter.com/QtdMhQt4WV
Interesting machine tidbit: Most companies don't deploy "a" machine learning model.
— Sebastian Raschka (@rasbt) August 4, 2022
🤯 40% of companies have more than 50 machine learning models in production.
And at big companies with >25k employees, 41% have >100 models in production!
(Algorithmia: https://t.co/yCkdrqkQ4b)
And last but not least, I continue to be amazed how freely people share their insights on @kaggle
— Radek Osmulski 🇺🇦 (@radekosmulski) August 4, 2022
The quality of some of the posts is amazing
Here is a very good analysis of the noisiness of the metric for instancehttps://t.co/13JX1yb3YI
6/8 pic.twitter.com/hLztTOV1S9
How to avoid leakage in preprocessing or splitting your data?
— Radek Osmulski 🇺🇦 (@radekosmulski) August 4, 2022
This talk is a superb resource on this topic
By the way, this highlights the value of @kaggle -- it is only in a competitive setting that such nuanced but important DS concepts come to life!https://t.co/ocvPV24LhJ
Masked Vision and Language Modeling for Multi-modal Representation Learning
— AK (@_akhaliq) August 4, 2022
abs: https://t.co/zpOExcblUH pic.twitter.com/siunmulnng
I put together a quick guide to help anyone interested in publishing research datasets on @kaggle https://t.co/K74rwzVtOy
— meg.ehh 🇨🇦 (@MeganRisdal) August 3, 2022
LMK if you have feedback and reach out if you're interested in working together! ☺️
❤️🔥 Machine learning is maybe the most fun but also the most challenging field
— Sebastian Raschka (@rasbt) August 1, 2022
🌱 Building and maintaining math, programming, and software engineering foundations is one part.
🔥 The other part is ALWAYS catching up & learning something new every day or getting left behind.
