It will work best for the robot that made the data as we have more controlled conditions than the mess we have in our current available data , but it can probably be transfered to others with a few select experiments and adjustments. Not sure about QM, I'm not a big expert there.
We need robitics before we can standardize the conditions enough to really make the ML shine. On the other hand we need the predictions and ML to make the robotics work.
Tillykke😉
Ah, I dont think Ill be in DK that date...
Sadly, many coutries didn't even bother to care about the scientists already living there. US scientists are not special, they just had funding.
Ja, underlige regler er der mange af. Men hele forskningsområdet har jo været udsultet over mange årtier, så det er lidt provokerende at der nu pludseligt skal gøres ekstra for specielle nationaliteter. Jeg selv endte med at udvandre, og det er nok ikke DK jeg flytter til hvis jeg flytter land igen.
Hvad med at tiltrække forskere fra Danmark?
1 week left to apply for this Assistant Professorship at Chalmers!
Join us in Gothenburg for the 3rd CHAIR Structured Learning Workshop! Very exciting line-up of speakers and we expect lots of engaging discussions, too. Attendance is free but you need to apply as we have limited slots.
6/ 2️⃣ Improved Custom Fingerprinting
We've made it easier to create and use your own molecular fingerprints.
3️⃣ Cleaner Code & UX Polish
• Typos fixed (yes, CheckSmilesSanitazion is now spelled right)
• Better subclass handling
• New hybrid logo for both light and dark background 🎨
5/ 🔬 Try a full example:
👉 scikit-mol.readthedo...
Whether you're validating models or deciding what predictions to trust — AD is now built-in.
8/ 📣 How do you use applicability domain estimation in chemML workflows?
Are these features helping your model validation or deployment? Let us know — and feel free to contribute or suggest more!
#cheminformatics #MachineLearning #OpenSource
4/ Includes:
• kNN, Leverage, and more
• Unified interfaces:
.predict() → in/out classification
.transform() → raw scores
.score_transform() → soft scores (0–1)
• Support for threshold fitting from validation sets
🧪
7/ 🙌 Huge thanks to @asiomchen, @Productivista, @mieczyslaw — and of course Olivier — for pushing the project forward!
We’re excited to see how the community uses these new tools. 👨🔬👩🔬
3/ 1️⃣ Applicability Domain Estimators
Thanks to a generous code donation from Olivier J. M. Béquignon's MLChemAD, we’ve added a suite of AD estimators—standardized and integrated into the Scikit-Mol framework. 🙏
2/ 🔬 What is Scikit-Mol?
An open-source library connecting #RDKit featurization with #Scikit-Learn workflows. It offers plug-and-play descriptors and fingerprints for ML in life sciences.
📦 github.com/EBjerrum/...
1/ 🧪 Scikit-Mol Update: New Features + Applicability Domain Estimation!
We've just released v0.6.0 — bringing feature-based AD estimators and a series of improvements to help you better validate and trust your chemistry + ML models. 🔍🧠
I remember I installed slackware Linux on my 486 laptop with 4 mb of ram using my brothers zipdrive. Had to hack the modelines to get X to work. The zipdisks had more storage than my harddrive! Nowadays its boringly easy.
C:/>
Who was he mocking?
Whats the timings for both of them?
Yes, it seems like an unnessecary complication.
This week's #RDKit blog post shows how to use the built-in drawing code to create a more complex figure.
greglandrum.github.io/rdkit-blog/p...
Men det her er min "proff" konto
ja
This is Marco Rubio explaining how the USA promised to defend Ukraine forever if they got rid of their nuclear arsenal left after the Soviet Union fell.
This is why lil marco was sinking into the couch. He was hoping we wouldn’t find it…so don’t RT right now this very second.
Thanks for the heads-up
