🚀 Introducing PantheonOS A Fully Open-Source Agent OS for Science
PantheonOS began as a research project in Qiu Lab @ Stanford and has since evolved into a vision to redefine data science in the era of AI—starting with computational biology, especially single-cell and spatial genomics.
🌟Applications open- LOGML 2025🌟
👥Mentor-led projects, expert talks, tutorials, socials, and a networking night
✍️Application form: logml.ai
🔬Projects: www.logml.ai/projects.html
📅Apply by 6th April 2025
✉️Questions? logml.committee@gmail.com
#MachineLearning #SummerSchool #LOGML #Geometry
The organisation and scientific advisory committees: @simofoti.bsky.social, @valegiunca.bsky.social, @pragya-singh.bsky.social, @daniel-platt.bsky.social, Vincenzo Marco De Luca, Massimiliano Esposito, Arne Wolf, Zhengang Zhong, Rahul Singh
We are currently recruiting mentors to lead up to 6 students on a week-long project at the intersection of geometry and ML. Mentors can be PhD students (not first years), Postdocs or lectures! Many projects result in top conferences and journal publications. Mentors expenses will be covered.
LOGML (London Geometry and Machine Learning) summer school is back and we are looking for mentors!
@logml.bsky.social aims to bring together mathematicians and computer scientists to collaborate on problems at the intersection of geometry and ML.
More information is available at www.logml.ai.
@simofoti.bsky.social @pragya-singh.bsky.social @valegiunca.bsky.social @daniel-platt.bsky.social
@mmbronstein.bsky.social @marinkazitnik.bsky.social
⭐️Mentor applications open⭐️
We're excited to announce that LOGML summer school will return in London: July 7-11 2025. We are seeking mentors to lead group projects at the intersection of geometry and machine learning. Find out more and apply:
logml.ai
ProCyon: A multimodal foundation model for protein phenotypes
Figure 1
Figure 2
Figure 3
ProCyon: A multimodal foundation model for protein phenotypes [new]
ProCyon: A multimodal foundation model for protein phenotypes www.biorxiv.org/content/10.1101/2024.12....
@imperialcollegeldn.bsky.social
@harvard.edu
@kingscollegelondon.bsky.social
@imperialbrains.bsky.social
I am happy to finally share ProCyon, a multimodal multiscale model that integrates protein sequences, structures, and natural language to predict and generate protein phenotypes.
Paper: www.biorxiv.org/content/10.1...
Blog post: kempnerinstitute.harvard.edu/research/dee...
ProCyon: A multimodal foundation model for protein phenotypes www.biorxiv.org/content/10.1101/2024.12.... 🧬🖥️🧪 https://github.com/mims-harvard/ProCyon
We tested it on 12 online tasks collected with Cognitron.
Compared to standard measures of RT and accuracy, IDoCT's measures of ability:
- have more interpretable latent cognitive factors
- are less sensitive to device
- have higher sensitivity and specificity
We tested the model on simulated data and IDoCT could reliably recover ground truth measures of trial’s difficulty, ability and visuomotor delay
IDoCT comes with a nice set of features:
- Robust: Works with as little as 100 participants
- Efficient: Scales up inexpensively to > 100K participants
- Flexible: Can work with potentially any online task collecting trial-by-trial responses
IDoCT derives specific estimates of ability, and visuomotor delay from trial-by-trial measures of reaction time (RT) and accuracy, while also providing data-driven trial’s difficulty scales that detect the most challenging aspects/dimensions of each task
🚨 It took two years but it finally happened!
Excited to share IDoCT - a novel computational model that can disentangle the motor and cognitive component from participants’ performance in online cognitive tasks - now published in Nature Digital Medicine.
