Search for: All records

Abstract Reliable projections of the South American Summer Monsoon (SASM) are critical for managing regional hydroclimatic risks, yet remain highly uncertain due to internal climate variability. Here, we reconstruct a robust historical SASM index ensemble from 1850 CE onward by integrating high-resolution paleoclimate proxies (tree rings and ice cores), historical documents, and instrumental observations. We further analyze future changes using large ensembles from the CESM2 and CanESM5 climate models. Our results demonstrate that multidecadal variability in the SASM is primarily driven by the Interdecadal Pacific Oscillation (IPO) and the associated changes in the Pacific Walker Circulation (PWC), whereas the influence of the tropical Atlantic sea surface temperature (SST) gradient is comparatively minor. By constraining these key Pacific modes, we reduce the uncertainty in projected SASM intensity by approximately 30%, highlighting their dominant role in shaping near-term monsoon trajectories. This study underscores the importance of improved simulation and representation of Pacific variability for advancing hydroclimate projections and informing climate adaptation strategies in tropical South America. 

This paper develops provable methods for managing trade-offs in machine learning serving systems. 

This work provides theoretical guidance for early-exit inference and studies recall-based exit rules for accuracy-latency trade-offs. 

The intrinsic alignments (IA) of galaxies, a key contaminant in weak lensing analyses, arise from correlations in galaxy shapes driven by tidal interactions and galaxy formation processes. Accurate IA modeling is essential for robust cosmological inference, but current approaches rely on perturbative methods that break down on nonlinear scales or on expensive simulations. We introduce IAEmu, a neural network-based emulator that predicts the galaxy position-position ( $\xi$), position-orientation ( $\omega$), and orientation-orientation ( $\eta$) correlation functions and their uncertainties using mock catalogs based on the halo occupation distribution (HOD) framework. Compared to simulations, IAEmu achieves ~3% average error for $\xi$and ~5% for $\omega$, while capturing the stochasticity of $\eta$without overfitting. The emulator provides both aleatoric and epistemic uncertainties, helping identify regions where predictions may be less reliable. We also demonstrate generalization to non-HOD alignment signals by fitting to IllustrisTNG hydrodynamical simulation data. As a fully differentiable neural network, IAEmu enables $ $10,000$$ speed-ups in mapping HOD parameters to correlation functions on GPUs, compared to CPU-based simulations. This acceleration facilitates inverse modeling via gradient-based sampling, making IAEmu a powerful surrogate model for galaxy bias and IA studies with direct applications to Stage IV weak lensing surveys. 

This paper studies differentially private revenue-maximizing single-item auction design and reduces private mechanism design to private quantile estimation. 

We investigate the problem of designing differentially private (DP), revenue- maximizing single item auction. Specifically, we consider broadly applicable settings in mechanism design where agents’ valuation distributions are indepen- dent, non-identical, and can be either bounded or unbounded. Our goal is to design such auctions with pure, i.e., (ω, 0) privacy in polynomial time.