https://arxiv.org/abs/1803.00567
explains the dual problem well
Peyré, G., & Cuturi, M. (2018, March 1). Computational Optimal transport. arXiv.org. https://arxiv.org/abs/1803.00567
tutorial_on_optimal_transport.pdf
2312.07511 A Hitchhiker’s Guide to Geometric GNNs for 3D Atomic Systems
March 1, 2025https://arxiv.org/abs/2312.07511
Duval, A., Mathis, S., V., Joshi, C. K., Schmidt, V., Miret, S., Malliaros, F. D., Cohen, T., Liò, P., Bengio, Y., & Bronstein, M. (2023, December 12). A Hitchhiker’s guide to Geometric GNNs for 3D atomic Systems. arXiv.org. https://arxiv.org/abs/2312.07511
Good intuition on spherical harmonics
Diffusion Tutorial
March 1, 2025Some tutorials on diffusion models:
[An Arxiv Tutorial]
https://arxiv.org/pdf/2403.18103
https://arxiv.org/abs/2403.18103
Chan, S. H. (2024, March 26). Tutorial on diffusion models for imaging and vision. arXiv.org. https://arxiv.org/abs/2403.18103
has master equation, forward & back SDE, relationship of SDE to p(x)
[Also, Some Useful Blogs]
https://baincapitalventures.notion.site/Diffusion-Without-Tears-14e1469584c180deb0a9ed9aa6ff7a4c https://yang-song.net/blog/2021/score/
https://lilianweng.github.io/posts/2021-07-11-diffusion-models/
One Day on Netflix is excruciating to watch. I loved it.
March 1, 2025tutorial on transformer
March 1, 2025https://www.datacamp.com/tutorial/how-transformers-work
How Transformers Work: A Detailed Exploration of Transformer Architecture
Explore the architecture of Transformers, the models that have revolutionized data handling through self-attention mechanisms.
Jan 9, 2024 · 15 min read
NYTimes.com: Francis Collins Retires From N.I.H., Saying Colleagues ‘Deserve the Utmost Respect’
March 1, 2025Francis Collins Retires From N.I.H., Saying Colleagues ‘Deserve the Utmost Respect’
Dr. Collins, a renowned geneticist, ran the National Institutes of Health for 12 years. His parting statement offered a pointed, if veiled, message to the Trump administration.
https://www.nytimes.com/2025/03/01/us/politics/francis-collins-nih-retires.html
GCM2 – Wikipedia
March 1, 2025https://en.wikipedia.org/wiki/GCM2
QT:{{”
The Drosophila ‘glial cells missing’ (gcm) gene is thought to act as a binary switch between neuronal and glial cell determination. The gcm protein and mammalian gcm homologs contain a conserved N-terminal gcm motif that has DNA-binding activity. See GCM1 (MIM 603715).[supplied by OMIM]
“}}
Wearables may offer clues to psychiatric diagnoses | Findings | Yale Alumni Magazine
March 1, 2025https://www.yalealumnimagazine.com/articles/6008-wearables-may-offer-clues-to-psychiatric-diagnoses
QT:{{”
A study by a Yale-led research team has shown that more accurate diagnosis of attention-deficit/hyperactivity disorder (ADHD) and other mental health conditions may be on the horizon, thanks to wearable sensors such as smartwatches. The study was published in the journal Cell.
ADHD, characterized by difficulty focusing, restlessness, and impulsive behavior, is the most commonly diagnosed behavioral disorder in children; it can lead to disruptions in learning and daily functioning. Early, accurate diagnosis and intervention can be critical in minimizing its impact. The problem, notes co–senior author Mark Gerstein, the Albert L. Williams Professor of Biomedical Informatics, is that “ADHD has traditionally been diagnosed
symptomatically, and there is an element of subjectivity to
categorizing human behavior. We wanted to see if wearable devices could increase diagnostic precision.”
The study analyzed clinical, wearable, and genetic data from 11,878 US adolescents (ages 9–14) recruited by the NIH Adolescent Brain Cognitive Development Consortium project. Information collected included measurements of heart rate, calorie expenditure, activity intensity, steps, and sleep intensity.
The researchers determined that correctly processed smartwatch data could be used as a “digital phenotype.” (Phenotype is the observable expression—physical characteristics, behaviors—of someone’s genotype, which is each person’s unique DNA sequence and their environment.) “We found,” says Gerstein, “that from the sensor readings we could quite accurately determine if someone has ADHD.”
The researchers used the data to train AI models to identify two psychiatric disorders. They further determined which measurements were most useful in characterizing them. Heart rate was the most important predictor for ADHD, while sleep quality and stage were more useful for identifying anxiety. Based on the patterns across the wearable features, the researchers were able to pinpoint genes and genetic variants associated with ADHD.
Gerstein notes the significance of making the connection to genotype. “We found that the smartwatch measurements can better relate disorders to genetics than just correlating them directly with clinical diagnoses,” he says. “Finding more genetic variants and genes related to the disorders could uncover molecular mechanisms and give us pathways to new treatments.”
“}}
linkstream2 microblog 