Fabian Schwarz
fabianmrblack.bsky.social
Fabian Schwarz
@fabianmrblack.bsky.social
Computational Chemistry & Materials Science | Postdoc Uppsala University 🇸🇪 | Former PhD Student @LNM ETHZ 🇨🇭 | Sverige | 🗣️: 🇨🇭🇩🇪🇬🇧🇸🇪
Underlines the importance of publishing and "reusing" good large datasets.
October 2, 2025 at 9:34 PM
Based on my experience, only fine-tuning foundational models and ignoring the computational cost of the foundational model (training + trading data + test data), one can save a lot of computational resources. Training from ground up & building training and test set is very "expensive", however.
October 2, 2025 at 9:30 PM
Microsoft released their DFT functional with chemical accuracy. It (or some other model) might be used to generate training data instead of CC.
September 11, 2025 at 7:58 PM
June 15, 2025 at 10:10 AM
In German: @realscientists.de , English account @realscientists.bsky.social seems inactive.
June 15, 2025 at 10:08 AM
Wenn es Forschungsideen sind, finde ich das ganz gut. Bin auch schon Jahre später auf Ideen von Kollegen oder vom Supervisor zurückgekommen. Natürlich muss man das dann auch so kommunizieren, dass man nicht erwartet, dass diese Idee jetzt sofort (oder überhaupt irgendwann) umgesetzt wird.
January 1, 2025 at 4:00 PM