Integrated Calendar

Atmospheric mineral dust is a well-established source of nutrients to marine ecosystems,yet its contribution to terrestrial plant nutrition has long been underestimated, largely due tothe assumption that nutrient acquisition occurs predominantly through root uptake fromsoils. Here, we present evidence from controlled greenhouse experiments under ambientand elevated CO₂, laboratory simulations of leaf microenvironments, isotopic andgeochemical tracing, and field fertilization experiments conducted in both a Mediterraneanecosystem and a tropical forest in Puerto Rico, demonstrating that plants can directlyacquire nutrients through their leaf surfaces following atmospheric dust deposition. Usingrare earth elements and Nd isotopes, we distinguish nutrients derived from soils from thosedelivered by deposited atmospheric particles. Laboratory simulations show that mildlyacidic leaf surfaces, together with organic acids secreted by leaves, enhance mineraldissolution and facilitate foliar uptake of dust-borne nutrients. In a pioneering Mediterraneanfield experiment explicitly designed to isolate foliar uptake, we quantified the bioavailablefraction of key nutrients supplied by dust, including P, Fe, Mn, and Cu, and observed clearenrichment of multiple micronutrients in leaf tissues following dust application. These fieldbasedmeasurements enabled the construction of a global geospatial framework integratingdust deposition with soil nutrient fluxes, indicating that dust-derived inputs can constitute ameaningful fraction of total nutrient supply across large regions, and that during dustevents, short-term foliar inputs can rival or exceed soil-derived fluxes. Complementary fieldobservations in a tropical forest in Puerto Rico further reveal foliar nutrient responsesconsistent with direct dust uptake. Building on these results, we outline a pathway forincorporating foliar dust uptake into Earth system representations of terrestrial nutrientcycling by explicitly accounting for atmospheric nutrient inputs at the canopy level and theirinteraction with soil-derived fluxes. Together, these findings identify foliar dust uptake as anoverlooked but consequential nutrient acquisition pathway and highlight its relevance inhighly weathered, nutrient-limited tropical forests, where atmospheric inputs may play acritical role in regulating nutrient availability and carbon–nutrient interactions.

Dear Colleagues,As part of the Multidisciplinary Vesicle Program Webinar Series, we are pleased to invite you to a special webinar entitled: "Introduction to Analytical Ultracentrifugation (AUC)" This session will provide an overview of Analytical Ultracentrifugation (AUC) and its applications in the characterization of extracellular vesicles, nanoparticles, macromolecular complexes and other biological systems. The webinar will highlight the principles of sedimentation analysis, methodological considerations and the advantages of AUC as a powerful label free analytical platform for assessing size distribution, heterogeneity, aggregation state and sample purity. The session is intended for researchers interested in advanced biophysical characterization approaches and scalable analytical solutions for EV and nanoparticle research. 

Information Theory views learning as universal prediction under log loss, characterized through regret bounds. We propose a framework that provides non-uniform, model dependent bounds utilizing an effective notion of architecture-based model complexity. This complexity is defined by the probability mass or volume of the set of all models in the vicinity of the target model \theta_0, in an informational distance. This volume might be hard to evaluate, yet by local analysis it is related to spectral properties of the expected Hessian or the Fisher Information Matrix at \theta_0, leading to tractable approximations. We argue that successful architectures possess abroad complexity range, enabling learning in highly over-parameterized model classes. The framework sheds light on the role of inductive biases, the effectiveness of the stochastic gradient descent (SGD)algorithm (but also other algorithms), and phenomena such as flat minima. It unifies online, batch, supervised, and generative settings, and applies across the stochastic-realizable and agnostic regimes. Moreover, it provides insights into the success of modern machine-learning architectures, such as deep neural networks and transformers, suggesting that their broad complexity range naturally arises from their layered structure. These insights open the door to the design of alternative architectures with potentially comparable or even superior performance.

Stable strontium isotopes δ88/86Sr) are increasingly used to study the modern marine Sr budget and changes in the marine carbonatefactory over geological time. Both are linked to the long-term carboncycle through processes such as continental weathering,hydrothermal alteration of oceanic crust, and carbonate burial.This seminar focuses on the role of the common evaporite mineralsgypsum (CaSO4·2H2O) and anhydrite (CaSO4) in the stable Srisotope budget of the ocean. I will address three main questions:(1) Does large-scale evaporite deposition, such as during theMessinian Salinity Crisis, affect the stable Sr isotope composition ofseawater? (2) Does dissolution of ancient gypsum and anhydriteaffect the δ88/86Sr value of rivers? And (3), can the δ88/86Sr value ofancient seawater be reconstructed from gypsum and anhydrite?Finally, I will present ongoing work that builds on these studies andaims to quantify atmospheric CO2 consumption by carbonateweathering in river catchments using stable Sr isotopes.

In multiclass classification with bandit feedback, a learner only observes whether its predicted label was correct, rather than the true label itself.  This simple change turns a basic supervised learning problem into a much harder exploration problem: how can one learn accurately over a large set of possible labels while receiving only minimal feedback?  In this talk, I will discuss a recent line of work studying the statistical complexity of bandit multiclass classification.  Perhaps surprisingly, we show that bandit feedback can be essentially free: with the right exploration strategy, one can match the full-information sample complexity rates up to logarithmic factors. The key technical ingredient is a new algorithmic approach to low-variance exploration that scales with the sparsity of the rewards, rather than with their ambient dimension.  Beyond resolving a substantial gap in prior work, these ideas also extend naturally to richer prediction problems such as list classification, combinatorial feedback models, and contextual bandits; somewhat unexpectedly, they have also played a role in resolving the tight sample complexity of standard full-information multiclass classification.

Based on joint work with PhD student Liad Erez and collaborators Fan Chen, Alon Cohen, Steve Hanneke, Yishay Mansour, Shay Moran, Sasha Rakhlin, and Qian Zhang.

The search for life beyond Earth is entering a new era, ushered in by the James Webb Space Telescope (JWST) and upcoming missions like Europa Clipper and JUICE. As high-resolution data emerge, data-driven methods can help detect subtle patterns and integrate diverse observations. I’ll present two studies that exemplify this approach.In the first study, we apply a spectral decomposition framework to JWST NIRSpec observations of Europa. We isolate spatial-spectral modes of variability across nine diagnostic bands and three jointly-analyzed observation geometries of the leading hemisphere. We identify anomalous ice textures enriched in volatiles across several geologically active terrains, with implications for ocean— surface exchange processes that may inform our understanding of Europa’s habitability.In the second study, we propose a new class of biosignature based on the statistical structure of organic assemblages. Using ecodiversity metrics to compare amino acid profiles, we find that biotic samples are consistently more diverse than their abiotic counterparts. This distinction also holds for the deep geologic record and for fatty acids. Relying solely on relative abundances, it is applicable to all planetary missions capable of measuring molecular abundances.

Human social communication relies on complex vocal behavior, social cognition, and neural mechanisms that remain difficult to study experimentally in naturalistic settings. In this talk, I will present recent work establishing the common marmoset as a powerful model for studying the neural basis of social communication and behavior. First, I will describe our recent Science paper demonstrating vocal labeling of conspecifics by nonhuman primates. I will then discuss ongoing computational work using generative spoken language models to uncover latent structure and potential syntactic organization in marmoset vocal communication. Finally, I will present new unpublished findings revealing social spatial tuning in hippocampal neurons during freely moving natural social interactions, using generalized additive models (GAMs) to characterize neural coding in complex behavioral environments. Together, these results suggest that marmosets provide a unique experimental platform for investigating the evolution and neural basis of human social communication and cognition.