I wrote #beginner level book teaching Deep Learning - its goal is to be the easiest intro possible In the book, each lesson builds a neural component *from scratch* in #NumPy Each *from scratch* toy code example is in the Github below #100DaysOfMLCode github.com/iamtrask/Grokk…

Also this week, Tim Cernak’s group present a key step analysis to complement retrosynthetic software predictions, then showcase it by minimizing step count in stemoamide synthesis science.org/doi/10.1126/sc…

Formal Cross-Coupling of Amines and Carboxylic Acids to Form sp3–sp2 Carbon–Carbon Bonds U-M College of Pharmacy University of Michigan Michigan Chemistry Total Synthesis and Methodology Highlights #CrossCoupling #Amines #CarboxylicAcids #CCBond #Formation pubs.acs.org/doi/10.1021/ja…

Scientific colour maps 8.0 fabiocrameri.ch/colourmaps #useBatlow #ScientificVisualisation #visualisation #Science #ColourPalette #DataVis #GraphicDesign #OpenSource #Accessibility #Colour #AcademicCommunityContribution #ACC 1/n

Can scientists be taught how to choose good projects? I took a fantastic class taught by Michael Fischbach last quarter where the foundational hypothesis is: Yes, they can! And I continually come back to these core insights as I think about project choice and direction in my 2nd year of

Honored for our perspective on low-data machine learning (centered on transfer- and active learning) for organic synthesis in JCIM & JCTC Journals to be highlighted Organic Process Research & Development. The paper can be seen here: pubs.acs.org/doi/10.1021/ac…

The Zimmerman Group group is at #ACSFall2023 come check out the amazing science from our group.

Happy to have given my first "in-person" presentation at #ACSFall2023 #CINF ML and AI for Organic Chemistry! Was great getting to know other people that share similar views. Thanks to those who asked questions and showed interest - means a lot to an imposter :)

Anti-hype LLM reading list. Pretty good list. gist.github.com/veekaybee/be37…

So happy to see good feedback on my latest Attention video! It took me a lot of thinking, scribbling, and asking the experts, but I was so happy to finally wrap my head around a pictorial explanation of the Key, Query, and Value matrices. Check it out! youtube.com/watch?v=UPtG_3…

On the simplicity of protein sequence-function relationships "We find that main and pairwise effects, together with a simple form of global nonlinearity, account for a median of 96% of total variance." Yeonwoo Park Brian Metzger JoeThornton biorxiv.org/content/10.110…

This looks like a good new dataset for evaluating machine learning protein engineering models: doi.org/10.1101/2023.0… The authors create 3706 alpha-amylase sequence variants based on presence/absence of buriedness, contact residues, active site distance, secondary structure.

New paper!! We found a pattern in how NNs extrapolate: as inputs become more OOD, model outputs tend to go towards some “average”-like prediction. What is this “average”-like prediction? Why does this happen? Can we leverage this to better handle OOD inputs? (Spoiler: Yes!) 🧵:

1/6 Molecular set representation learning Do we need molecular topology or are sets of vectors of atomic invariants (~ECFP0) enough? They can be; and enhance existing GNNs; and be used for binding affinity, molecular property, and rxn yield prediction. chemrxiv.org/engage/chemrxi…

Why do Random Forests perform so well off-the-shelf & appear essentially immune to overfitting?!? I’ve found the text-book answer “it’s just variance reduction 🤷🏼‍♀️” to be a bit too unspecific, so in our new pre-print arxiv.org/abs/2402.01502, Alan Jeffares & I investigate..🕵🏼‍♀️ 1/n

Excited to share an awesome collab between the Narayan Lab @ UMich ang the gomes group! Check out our preprint on CATNIP, using ML to help chemists identify a compatible biocatalyst for small molecule synthesis! Huge thanks to Daniil Boiko for making it all possible!🎉chemrxiv.org/engage/chemrxi…

Check out our latest article by Eunjae Shim now published in Chemical Science on label ranking, introduced as an efficient #MachineLearning algorithm to predict high yielding and effective reaction conditions for a variety of substrates. pubs.rsc.org/en/content/art…

Calling everyone interested in learning biocatalysis! Come to the Narayan lab to learn the biocat basics. Please apply here - docs.google.com/forms/d/e/1FAI… #Biocat #GoBlue