https://bsky.app/profile/aidd.bsky.social

We have moved to bsky.app/profile/aidd.b… No posts will be at twitter/X anymore.

34th International Conference on Artificial Neural Networks #ICANN2025 Call for Papers, Special Sessions, Workshops & Tutorials Kaunas, Lithuania Sep 9 - 12, 2025 See: e-nns.org/icann2025/call… #ENNS #AI #NeuralNetworks #ML #MachineLearning

new: "Human-in-the-loop active learning for goal-oriented molecule generation" jcheminf.biomedcentral.com/articles/10.11…

new: "Be aware of overfitting by hyperparameter optimization!" jcheminf.biomedcentral.com/articles/10.11…

Hyperparameter optimization is frequently employed in machine learning, but optimization of a large space of parameters could result in overfitting.Use of pre-optimized hyperparameters can give similar or better performance up to 10000x faster: jcheminf.biomedcentral.com/articles/10.11… @jcheminf

@ELLISforEurope ML4Science workshop co-organised by our PI @fra_grisoni & @AiddOne PI @gklambauer was a great event to learn advances in ML & promote @AiddOne results: ai-dd.eu/lectures. @AiChemist_DN fellows @xasanovadina @5_Marcel_7 attend poster session. Berlin eve.

A huge congratulations to Dr. @JULIANCREMER15 for successfully defending his PhD - we are on a roll! Dr. Cremer will continue his outstanding work on #denovo #drugdesign, which has received many accolades and an impressive amount of citations, as a postdoc at @pfizer. 👨‍🔬👏🎉

Bio-xLSTM: Generative modeling, representation and in-context learning of biological and chemical sequences xLSTM also shines for DNA, proteins and small molecules -- can handle large-range interactions and huge context! P: arxiv.org/abs/2411.04165

Menzerath's law (big construct formed by small components) can be seen as antifragility of structure coming from fragility of the parts

AiChemist PI Alessandra Roncaglioni aichemist.eu/irfmn-supervis…, head of the Environmental Chemistry and Toxicology Laboratory of @MarioNegriIRCCS in Milan, explains the use of Artificial Intelligence for the analysis of drugs. Enjoy listening, if you parli Italiano 😉.

"Can Publishing Neural Networks Expose Confidential Training Data?" Fabian Krüger will dive into the details during the AiChemist seminar on 09.12 at 14:00 - all are welcome to join! Link to be posted on aichemist.eu/news at 10:00 CET the same day.

Does this work excite you? Are you a prospective PhD, postdoc, or technical assistant? Look no further! Thanks to a generous donation from OpenPhilanthropy, we have multiple openings in our group, reach out to us here forms.fillout.com/t/eoGA7AhnAKus

Our modeling work with Dr. Igor Tetko @AiddOne , which focuses on Plasma Protein Binding (PPB), has been published in the official journal of EUFEPS @eufeps . doi.org/10.1016/j.ejps… It is interesting to see that LogIt improves prediction accuracy for high-PPB compounds.

Igor Tetko explains AiChemist and @AiddOne projects to PIs of helmholtz-munich.de/en/molecular-t… from @HelmholtzMunich and speaks about excellent career perspectives of Doctoral Candidates trained within @MSCActions

Chemists often combine many different techniques to elucidate structures. @adrian_mirza_ has been building a system that mimics this using machine-learning models and genetic algorithms.

Understanding the difference between Standard Deviation (SD) and Standard Error (SE) is crucial for accurate data interpretation. SD measures the variability within your data, indicating how spread out the individual data points are from the mean. In contrast, SE measures the uncertainty around the sample mean as an estimate of the population mean. It reflects the precision of the mean, with SE decreasing as the sample size increases, making your estimate more reliable. The relationship between SD and SE is given by the formula: SE = SD / √(sample size). While SD remains relatively constant with larger samples, SE diminishes, highlighting the reduced uncertainty in the mean estimate. A common mistake in research is using the “±” notation without specifying whether it refers to SD or SE, leading to potential misinterpretation of the data. Clear distinction is essential for transparency and accuracy in reporting. Key Takeaways: • Use SD to describe data variability. • Use SE to indicate the precision of the mean. • Always specify which measure you are reporting.

There’s a new Practical Cheminformatics post, “Some Thoughts on Dataset Splitting,” (with code and a robot cartoon) at practicalcheminformatics.blogspot.com/2024/11/some-t…

What was a busy week! After lecture and participation to PhD defense at Shenzhen Institute of Advanced Technology, I had seminar at Peking Union Medical College followed by lecture at Beijing University of Chemical Technology to @AiddOne partner, Prof. Yan ai-dd.eu/aixia

ELLIS ML4Molecules DEADLINE IN TWO DAYS! Consider submitting your current works as extended abstracts or short papers to ML4Molecules 2024: DEADLINE IS NOVEMBER 1. moleculediscovery.github.io/workshop2024/

Congratulations our PI with another great success!