Search

We present Subtractive Training, a simple and novel method for synthesizing individual musical instrument stems given other instruments as context. This method pairs a dataset of complete music mixes with 1) a variant of the dataset lacking a specific stem, and 2) LLM-generated instructions describing how the missing stem should be reintroduced. We then fine-tune a pretrained text-to-audio diffusion model to generate the missing instrument stem, guided by both the existing stems and the text instruction. Our results demonstrate Subtractive Training's efficacy in creating authentic drum stems that seamlessly blend with the existing tracks. We also show that we can use the text instruction to control the generation of the inserted stem in terms of rhythm, dynamics, and genre, allowing us to modify the style of a single instrument in a full song while keeping the remaining instruments the same. Lastly, we extend this technique to MIDI formats, successfully generating compatible bass, drum, and guitar parts for incomplete arrangements., Comment: 5 pages, survey, edit pipeline figure, fix typos

On the route towards merging neutron stars and stripped-envelope supernovae, binary population synthesis predicts a large number of post-interaction systems with massive stars that have stripped off their outer layers. Yet, observations of such stars in the intermediate-mass regime below the Wolf-Rayet masses are rare. Using X-Shooting ULLYSES (XShootU) data, we discovered three partially stripped star + Be/Oe binaries in the Magellanic Clouds. We analyzed the UV and optical spectra using the PoWR model atmosphere code by superimposing model spectra corresponding to each component. The estimated current masses of the partially stripped stars fall within the intermediate mass range of 4-8 $M_{\odot}$. These objects are overluminous for their stellar masses, matching core He-burning luminosities. Their Be/Oe secondaries have much higher masses than their stripped primaries (mass ratio > 2). All three partially stripped stars show significant nitrogen enrichment and carbon and oxygen depletion on their surfaces. Additionally, one of our sample stars exhibits significant helium enrichment. Our study provides the first comprehensive determination of the wind parameters of partially stripped stars in the intermediate mass range. The wind mass-loss rates of these stars are found to be on the order of $10^{-7} M_\odot$ yr$^{-1}$, which is over ten times higher than that of OB stars of the same luminosity. Current evolutionary models characterizing this phase typically employ OB or WR mass-loss rates, which underestimate or overestimate stripped stars' mass-loss rates by an order of magnitude. Binary evolution models indicate that the observed primaries had initial masses of 12-17 $M_{\odot}$, making them potential candidates for stripped-envelope supernovae that form neutron stars. If they survive the explosion, these systems may become Be X-ray binaries and later double neutron stars., Comment: Accepted for publication in Astronomy & Astrophysics

Human feedback plays a critical role in learning and refining reward models for text-to-image generation, but the optimal form the feedback should take for learning an accurate reward function has not been conclusively established. This paper investigates the effectiveness of fine-grained feedback which captures nuanced distinctions in image quality and prompt-alignment, compared to traditional coarse-grained feedback (for example, thumbs up/down or ranking between a set of options). While fine-grained feedback holds promise, particularly for systems catering to diverse societal preferences, we show that demonstrating its superiority to coarse-grained feedback is not automatic. Through experiments on real and synthetic preference data, we surface the complexities of building effective models due to the interplay of model choice, feedback type, and the alignment between human judgment and computational interpretation. We identify key challenges in eliciting and utilizing fine-grained feedback, prompting a reassessment of its assumed benefits and practicality. Our findings -- e.g., that fine-grained feedback can lead to worse models for a fixed budget, in some settings; however, in controlled settings with known attributes, fine grained rewards can indeed be more helpful -- call for careful consideration of feedback attributes and potentially beckon novel modeling approaches to appropriately unlock the potential value of fine-grained feedback in-the-wild.

Global climate models typically operate at a grid resolution of hundreds of kilometers and fail to resolve atmospheric mesoscale processes, e.g., clouds, precipitation, and gravity waves (GWs). Model representation of these processes and their sources is essential to the global circulation and planetary energy budget, but subgrid scale contributions from these processes are often only approximately represented in models using parameterizations. These parameterizations are subject to approximations and idealizations, which limit their capability and accuracy. The most drastic of these approximations is the "single-column approximation" which completely neglects the horizontal evolution of these processes, resulting in key biases in current climate models. With a focus on atmospheric GWs, we present the first-ever global simulation of atmospheric GW fluxes using machine learning (ML) models trained on the WINDSET dataset to emulate global GW emulation in the atmosphere, as an alternative to traditional single-column parameterizations. Using an Attention U-Net-based architecture trained on globally resolved GW momentum fluxes, we illustrate the importance and effectiveness of global nonlocality, when simulating GWs using data-driven schemes., Comment: International Conference on Machine Learning 2024

In \cite{FGHP}, the first author and his collaborators proved an equivariant Tamagawa number formula for the special value at $s=0$ of a Goss--type $L$--function, equivariant with respect to a Galois group $G$, and associated to a Drinfeld module defined on $\Bbb F_q[t]$ and over a finite, integral extension of $\Bbb F_q[t]$. The formula in question was proved provided that the values at $0$ of the Euler factors of the equivariant $L$--function in question satisfy certain identities involving Fitting ideals of certain $G$--cohomologically trivial, finite $\Bbb F_q[t][G]$--modules associated to the Drinfeld module. In \cite{FGHP}, we prove these identities in the particular case of the Carlitz module. In this paper, we develop general techniques and prove the identities in question for arbitrary Drinfeld modules. Further, we indicate how these techniques can be extended to the more general case of higher dimensional abelian $t$--modules, which is relevant in the context of the proof of the equivariant Tamagawa number formula for abelian $t$--modules given by N. Green and the first author in \cite{Green-Popescu}. This paper is based on a lecture given by the first author at ICMAT Madrid in May 2023 and builds upon results obtained by the second author in his PhD thesis \cite{Ramachandran-thesis}.

We perform a large-scale hydrodynamic simulation of a massive star cluster whose stellar population mimics that of the Cygnus OB2 association. The main-sequence stars are first simulated during 1.6 Myr, until a quasi-stationary state is reached. At this time the three Wolf-Rayet stars observed in Cygnus OB2 are added to the simulation, which continues to 2 Myr. Using a high-resolution grid in the centre of the domain, we can resolve the most massive stars individually, which allows us to probe the kinetic structures at small (parsec) scales. We find that, although the cluster excavates a spherical "superbubble" cavity, the stellar population is too loosely distributed to blow a large-scale cluster wind termination shock, and that collective effects from wind-wind interactions are much less efficient than usually assumed. This challenges our understanding of the ultra-high energy emission observed from the region., Comment: 15 pages, 12 figures

The fact that neutrinos carry a non-vanishing rest mass is evidence of physics beyond the Standard Model of elementary particles. Their absolute mass bears important relevance from particle physics to cosmology. In this work, we report on the search for the effective electron antineutrino mass with the KATRIN experiment. KATRIN performs precision spectroscopy of the tritium $\beta$-decay close to the kinematic endpoint. Based on the first five neutrino-mass measurement campaigns, we derive a best-fit value of $m_\nu^{2} = {-0.14^{+0.13}_{-0.15}}~\mathrm{eV^2}$, resulting in an upper limit of $m_\nu < {0.45}~\mathrm{eV}$ at 90 % confidence level. With six times the statistics of previous data sets, amounting to 36 million electrons collected in 259 measurement days, a substantial reduction of the background level and improved systematic uncertainties, this result tightens KATRIN's previous bound by a factor of almost two., Comment: 61 pages, 20 figures, 2 tables

Machine learning (ML) and Artificial Intelligence (AI) have fueled remarkable advancements, particularly in healthcare. Within medical imaging, ML models hold the promise of improving disease diagnoses, treatment planning, and post-treatment monitoring. Various computer vision tasks like image classification, object detection, and image segmentation are poised to become routine in clinical analysis. However, privacy concerns surrounding patient data hinder the assembly of large training datasets needed for developing and training accurate, robust, and generalizable models. Federated Learning (FL) emerges as a compelling solution, enabling organizations to collaborate on ML model training by sharing model training information (gradients) rather than data (e.g., medical images). FL's distributed learning framework facilitates inter-institutional collaboration while preserving patient privacy. However, FL, while robust in privacy preservation, faces several challenges. Sensitive information can still be gleaned from shared gradients that are passed on between organizations during model training. Additionally, in medical imaging, quantifying model confidence\uncertainty accurately is crucial due to the noise and artifacts present in the data. Uncertainty estimation in FL encounters unique hurdles due to data heterogeneity across organizations. This paper offers a comprehensive review of FL, privacy preservation, and uncertainty estimation, with a focus on medical imaging. Alongside a survey of current research, we identify gaps in the field and suggest future directions for FL research to enhance privacy and address noisy medical imaging data challenges., Comment: 31 pages, 5 figures, 3 tables, Journal preprint

Offshore wind farms have emerged as a popular renewable energy source that can generate substantial electric power with a low environmental impact. However, integrating these farms into the grid poses significant complexities. To address these issues, optimal-sized energy storage can provide potential solutions and help improve the reliability, efficiency, and flexibility of the grid. Nevertheless, limited studies have attempted to perform energy storage sizing while including design and operations (i.e., control co-design) for offshore wind farms. As a result, the present work develops a control co-design optimization formulation to optimize multiple objectives and identify Pareto optimal solutions. The graph-based optimization framework is proposed to address the complexity of the system, allowing the optimization problem to be decomposed for large power systems. The IEEE-9 bus system is treated as an onshore AC grid with two offshore wind farms connected via a multi-terminal DC grid for our use case. The developed methodology successfully identifies the Pareto front during the control co-design optimization, enabling decision-makers to select the best compromise solution for multiple objectives.