Search

We developed a mathematical model to investigate the role of indirect transmission in the spread of infectious diseases, using the illustrative example of sarcoptic mange as a case study. This disease can be transmitted through direct contact between an infected host and a susceptible one, or indirectly when potential hosts encounter infectious mites and larvae deposited in the environment, commonly referred to as fomites. Our focus is on exploring the potential of these infectious reservoirs as triggers for emerging infection events and as stable reservoirs of the disease. To achieve this, our mean field compartmental model incorporates the epidemiological dynamics driven by indirect transmission via fomites. We identify different types of dynamics that the system can go into, controlled by different levels of direct and indirect transmission. Among these, we find a new regime where the disease can emerge and persist over time solely through fomites, without the necessity for direct transmission. This possibility of the system reveals an evolutionary pathway that could enable the parasite to enhance its fitness beyond host co-evolution. We also define a new threshold based on an effective reproductive number, that enables us to predict the conditions for disease persistence. Our model allows us to assess the potential effectiveness of various disease intervention measures by incorporating a feature observed in real systems. We hope this contributes to a better understanding of infectious disease outbreaks., Comment: Accepted in Physica A

Concept Bottleneck Models (CBMs) have emerged as a promising interpretable method whose final prediction is based on intermediate, human-understandable concepts rather than the raw input. Through time-consuming manual interventions, a user can correct wrongly predicted concept values to enhance the model's downstream performance. We propose Stochastic Concept Bottleneck Models (SCBMs), a novel approach that models concept dependencies. In SCBMs, a single-concept intervention affects all correlated concepts, thereby improving intervention effectiveness. Unlike previous approaches that model the concept relations via an autoregressive structure, we introduce an explicit, distributional parameterization that allows SCBMs to retain the CBMs' efficient training and inference procedure. Additionally, we leverage the parameterization to derive an effective intervention strategy based on the confidence region. We show empirically on synthetic tabular and natural image datasets that our approach improves intervention effectiveness significantly. Notably, we showcase the versatility and usability of SCBMs by examining a setting with CLIP-inferred concepts, alleviating the need for manual concept annotations.

Astronomical adaptive optics (AO) is a critical approach to enable ground-based diffraction-limited imaging and high contrast science, with the potential to enable habitable exoplanet imaging on future extremely large telescopes. However, AO systems must improve significantly to enable habitable exoplanet imaging. Time lag between the end of an exposure and end of deformable mirror commands being applied in an AO loop is now the dominant error term in many extreme AO systems (e.g., Poyneer et al. 2016), and within that lag component detector read time is becoming non-negligible (e.g., Cetre et al. 2018). This term will decrease as faster detector readout capabilities are developed by vendors. In complement, we have developed a modified Shack Hartmann Wavefront Sensor (SHWFS) to address this problem called the Focal-plane Actualized Shifted Technique Realized for a SHWFS (fastrSHWFS). The novelty of this design is to replace the usual lenslet array with a bespoke pupil-plane phase mask that redistributes the spot pattern on the detector into a rectangular array with a custom aspect ratio (in an extreme case, if the detector size can accommodate it, the array can be a single line). We present the fastrSHWFS concept and preliminary laboratory tests. For some detectors and AO systems, the fastrSHWFS technique can decrease the read time per frame compared to a regular SHWFS by up to 30x, supporting the goal of reduced AO lag needed to eventually enable habitable exoplanet imaging., Comment: submitted to SPIE Astronomical Telescopes and Instrumentation, paper number 13097-283

Many superconducting on-chip filter-banks suffer from poor coupling to the detectors behind each filter. This is a problem intrinsic to the commonly used half wavelength filter, which has a maximum theoretical coupling of 50 %. In this paper we introduce a phase coherent filter, called a directional filter, which has a theoretical coupling of 100 %. In order to to study and compare different types of filter-banks, we first analyze the measured filter frequency scatter, losses, and spectral resolution of a DESHIMA 2.0 filter-bank chip. Based on measured fabrication tolerances and losses, we adapt the input parameters for our circuit simulations, quantitatively reproducing the measurements. We find that the frequency scatter is caused by nanometer-scale line-width variations and that variances in the spectral resolution is caused by losses in the dielectric only. Finally, we include these realistic parameters in a full filter-bank model and simulate a wide range of spectral resolutions and oversampling values. For all cases the directional filter-bank has significantly higher coupling to the detectors than the half-wave resonator filter-bank. The directional filter eliminates the need to use oversampling as a method to improve the total efficiency, instead capturing nearly all the power remaining after dielectric losses., Comment: 11 pages, 4 figures. Accepted for publication by JLTP

Recent gravitational wave observations include possible detections of black hole - neutron star binary mergers. As with binary black hole mergers, numerical simulations help characterize the sources. For binary systems with neutron star components, the simulations help to predict the imprint of tidal deformations and disruptions on the gravitational wave signals. In a previous study, we investigated how the mass of the black hole has an impact on the disruption of the neutron star and, as a consequence, on the shape of the gravitational waves emitted. We extend these results to study the effects of varying the compactness of the neutron star. We consider neutron star compactness in the 0.123 to 0.2 range for binaries with mass ratios of 3 and 5. As the compactness and the mass ratio increase, the binary system behaves during the late inspiral and merger more like a black hole binary. For the case with the highest mass ratio and most compact neutron star, the gravitational waves emitted, in terms of mismatches, are almost indistinguishable from those by a binary black hole. The disruption of the star significantly suppresses the kicks on the final black hole. The disruption also affects, although not dramatically, the spin of the final black hole. Lastly, for neutron stars with low compactness, the quasi-normal ringing of the black hole after the merger does not show a clean quasi-normal ringing because of the late accretion of debris from the neutron star., Comment: 24 pages, 15 figures, 6 tables

Given two images, we can estimate the relative camera pose between them by establishing image-to-image correspondences. Usually, correspondences are 2D-to-2D and the pose we estimate is defined only up to scale. Some applications, aiming at instant augmented reality anywhere, require scale-metric pose estimates, and hence, they rely on external depth estimators to recover the scale. We present MicKey, a keypoint matching pipeline that is able to predict metric correspondences in 3D camera space. By learning to match 3D coordinates across images, we are able to infer the metric relative pose without depth measurements. Depth measurements are also not required for training, nor are scene reconstructions or image overlap information. MicKey is supervised only by pairs of images and their relative poses. MicKey achieves state-of-the-art performance on the Map-Free Relocalisation benchmark while requiring less supervision than competing approaches.

This paper argues why FLAMINGO (Fast Light Atmospheric Monitoring and Imaging Novel Gamma-ray Observatory) is the perfect name for an array of very-high-energy Cherenkov telescopes. Studies which indicate pink is the most suitable pigment for the structures of Cherenkov telescopes have passed with flying colors. Pink optimizes the absorption and reflectivity properties of the telescopes with respect to the characteristic blue color of the Cherenkov radiation emitted by high-energy particles in the atmosphere. In addition to giving the sensitivity a big leg up, a pink color scheme also adds a unique and visually appealing aspect to the project's branding and outreach efforts. FLAMINGO has a fun and memorable quality that can help to increase public engagement and interest in astrophysics and also help to promote diversity in the field with its colorful nature. In an era of increasingly unpronounceable scientific acronyms, we are putting our foot down. FLAMINGO is particularly fitting, as flamingos have eyesight optimized to detect small particles, aligning with the primary purpose of Cherenkov telescopes to detect faint signals from air showers. We should not wait in the wings just wishing for new name to come along: in FLAMINGO we have an acronym that both accurately reflects the science behind Cherenkov telescopes and provides a visually striking identity for the project. While such a sea change will be no easy feet, we are glad to stick our necks out and try: FLAMINGO captures the essence of what an array of Cherenkov telescopes represents and can help to promote the science to a wider audience. We aim to create an experiment and brand that people from all walks of life will flock to., Comment: 4 pages, 2 figures

NVIDIA Tensor Cores and AMD Matrix Cores (together called Matrix Accelerators) are of growing interest in high-performance computing and machine learning owing to their high performance. Unfortunately, their numerical behaviors are not publicly documented, including the number of extra precision bits maintained, the accumulation order of addition, and predictable subnormal number handling during computations. This makes it impossible to reliably port codes across these differing accelerators. This paper contributes a collection of {\em Feature Targeted Tests for Numerical Properties} that that help determine these features across five floating-point formats, four rounding modes and additional that highlight the rounding behaviors and preservation of extra precision bits. To show the practical relevance of FTTN, we design a simple matrix-multiplication test designed with insights gathered from our feature-tests. We executed this very simple test on five platforms, producing different answers: V100, A100, and MI250X produced 0, MI100 produced 255.875, and Hopper H100 produced 191.875. Our matrix multiplication tests employ patterns found in iterative refinement-based algorithms, highlighting the need to check for significant result variability when porting code across GPUs.

Cross-modal retrieval is the task of retrieving samples of a given modality by using queries of a different one. Due to the wide range of practical applications, the problem has been mainly focused on the vision and language case, e.g. text to image retrieval, where models like CLIP have proven effective in solving such tasks. The dominant approach to learning such coordinated representations consists of projecting them onto a common space where matching views stay close and those from non-matching pairs are pushed away from each other. Although this cross-modal coordination has been applied also to other pairwise combinations, extending it to an arbitrary number of diverse modalities is a problem that has not been fully explored in the literature. In this paper, we propose two different approaches to the problem. The first is based on an extension of the CLIP contrastive objective to an arbitrary number of input modalities, while the second departs from the contrastive formulation and tackles the coordination problem by regressing the cross-modal similarities towards a target that reflects two simple and intuitive constraints of the cross-modal retrieval task. We run experiments on two different datasets, over different combinations of input modalities and show that the approach is not only simple and effective but also allows for tackling the retrieval problem in novel ways. Besides capturing a more diverse set of pair-wise interactions, we show that we can use the learned representations to improve retrieval performance by combining the embeddings from two or more such modalities.

Recently, interpretable machine learning has re-explored concept bottleneck models (CBM). An advantage of this model class is the user's ability to intervene on predicted concept values, affecting the downstream output. In this work, we introduce a method to perform such concept-based interventions on pretrained neural networks, which are not interpretable by design, only given a small validation set with concept labels. Furthermore, we formalise the notion of intervenability as a measure of the effectiveness of concept-based interventions and leverage this definition to fine-tune black boxes. Empirically, we explore the intervenability of black-box classifiers on synthetic tabular and natural image benchmarks. We focus on backbone architectures of varying complexity, from simple, fully connected neural nets to Stable Diffusion. We demonstrate that the proposed fine-tuning improves intervention effectiveness and often yields better-calibrated predictions. To showcase the practical utility of our techniques, we apply them to deep chest X-ray classifiers and show that fine-tuned black boxes are more intervenable than CBMs. Lastly, we establish that our methods are still effective under vision-language-model-based concept annotations, alleviating the need for a human-annotated validation set.

Machine learning, particularly neural networks, has rapidly permeated most activities and work where data has a story to tell. Recently, deep learning has started to be used for solving differential equations with input from physics, also known as Physics Informed Neural Networks (PINNs). We present a study showing the efficacy of PINNs for solving the Zerilli and the Regge-Wheeler equations in the time domain to calculate the quasi-normal oscillation modes of a Schwarzschild black hole. We compare the extracted modes with those obtained with finite difference methods. Although the PINN results are competitive, with a few percent differences in the quasi-normal modes estimates relative to those computed with finite difference methods, the real power of PINNs will emerge when applied to large dimensionality problems., Comment: 20 pages, 6 figures

This is a comprehensive review of the Greedy Randomized Adaptive Search Procedure (GRASP) metaheuristic and its hybridization with Path Relinking (PR) over the past two decades. GRASP with PR has become a widely adopted approach for solving hard optimization problems since its proposal in 1999. The paper covers the historical development of GRASP with PR and its theoretical foundations, as well as recent advances in its implementation and application. The review includes a critical analysis of variants of PR, including memory-based and randomized designs, with a total of ten different implementations. It describes these advanced designs both theoretically and practically on two well-known optimization problems, linear ordering and max-cut. The paper also explores the hybridization of GRASP with PR and other metaheuristics, such as Tabu Search and Scatter Search. Overall, this review provides valuable insights for researchers and practitioners seeking to utilize GRASP with PR for solving optimization problems., Comment: 28 pages, 13 figures

We characterize universal features of the sample-to-sample fluctuations of global geometrical observables, such as the area, width, length, and center-of-mass position, in random growing planar clusters. Our examples are taken from simulations of both continuous and discrete models of kinetically rough interfaces, including several universality classes, such as Kardar-Parisi-Zhang. We mostly focus on the scaling behavior with time of the sample-to-sample deviation for those global magnitudes, but we have also characterized their histograms and correlations.

Emerging deep learning workloads urgently need fast general matrix multiplication (GEMM). To meet such demand, one of the critical features of machine-learning-specific accelerators such as NVIDIA Tensor Cores, AMD Matrix Cores, and Google TPUs is the support of mixed-precision enabled GEMM. For DNN models, lower-precision FP data formats and computation offer acceptable correctness but significant performance, area, and memory footprint improvement. While promising, the mixed-precision computation on error resilience remains unexplored. To this end, we develop a fault injection framework that systematically injects fault into the mixed-precision computation results. We investigate how the faults affect the accuracy of machine learning applications. Based on the error resilience characteristics, we offer lightweight error detection and correction solutions that significantly improve the overall model accuracy if the models experience hardware faults. The solutions can be efficiently integrated into the accelerator's pipelines.

We consider the isochrone curves in first-passage percolation on a 2D square lattice, i.e. the boundary of the set of points which can be reached in less than a given time from a certain origin. The occurrence of an instantaneous average shape is described in terms of its Fourier components, highlighting a crossover between a diamond and a circular geometry as the noise level is increased. Generally, these isochrones can be understood as fluctuating interfaces with an inhomogeneous local width which reveals the underlying lattice structure. We show that once these inhomogeneities have been taken into account, the fluctuations fall into the Kardar-Parisi-Zhang (KPZ) universality class with very good accuracy, where they reproduce the Family-Vicsek Ansatz with the expected exponents and the Tracy-Widom histogram for the local radial fluctuations.

LISA, the Laser Interferometer Space Antenna, will usher in a new era in gravitational-wave astronomy. As the first anticipated space-based gravitational-wave detector, it will expand our view to the millihertz gravitational-wave sky, where a spectacular variety of interesting new sources abound: from millions of ultra-compact binaries in our Galaxy, to mergers of massive black holes at cosmological distances; from the beginnings of inspirals that will venture into the ground-based detectors' view to the death spiral of compact objects into massive black holes, and many sources in between. Central to realising LISA's discovery potential are waveform models, the theoretical and phenomenological predictions of the pattern of gravitational waves that these sources emit. This white paper is presented on behalf of the Waveform Working Group for the LISA Consortium. It provides a review of the current state of waveform models for LISA sources, and describes the significant challenges that must yet be overcome., Comment: 239 pages, 11 figures, white paper from the LISA Consortium Waveform Working Group, invited for submission to Living Reviews in Relativity, updated with comments from community

We characterize the long-term state of the 1D Dirac vacuum stirred by an impenetrable object, modeled as the ground state of a finite free-fermionic chain dynamically perturbed by a moving classical obstacle which suppresses the local hopping amplitudes. We find two different regimes, depending on the velocity of the obstacle. For a slow motion, the effective Floquet Hamiltonian presents features which are typical of the Gaussian orthogonal ensemble, and the occupation of the Floquet modes becomes roughly homogeneous. Moreover, the long term entanglement entropy of a contiguous block follows a Gaussian analogue of Page's law, i.e. a volumetric behavior. Indeed, the statistical properties of the reduced density matrices correspond to those of a random Slater determinant, which can be described using the Jacobi ensemble from random matrix theory. On the other hand, if the obstacle moves fast enough, the effective Floquet Hamiltonian presents a Poissonian behavior. The nature of the transition is clarified by the entanglement links, which determine the effective geometry underlying the entanglement structure, showing that the one-dimensionality of the physical Hamiltonian dissolves into a random adjacency matrix as we slow down the obstacle motion.

In order for telescopes to obtain good and precise images they need to see through atmospheric turbulence. To accomplish this and compensate for atmospheric turbulence we use Adaptive Optics technologies. In this thesis we analyze the variations in phase delays across phase plates which simulate atmospheric turbulence in order to characterize them and determine how well these phase plates reproduce the phase delay variation of the atmosphere. This experiment was conducted using the Quadrature Polarization Interferometer (QPI) testbed in the Lab of Adaptive Optics (LAO) at the University of California Santa Cruz (UCSC). Using the QPI lab setup allowed us to be able to develop and refine a final algorithm for determining the phase delay of the phase plates. The characterization of the phase plates was accomplished by using the interference patterns between a test and reference path and measuring their pathlength variations. This was achieved by calculating the intensity of the two paths of the Helium-Neon (HeNe) laser, modifying those values in order to obtain phase values between the beam, and from the phase values we are able to determine the path length variations, or phase delays, of the phase plates. This allows us to determine Fried's parameter, $r_o$, by analyzing any given separation on the phase plate and determining the path length variations between those separations., Comment: Bachelor's thesis

Boson stars are potential sources of gravitational radiation. To improve our comprehension of these exotic objects, we present the study of geodesics within the space-times of stable, collapsing, and migrating boson stars. We focus on timelike geodesics that are initially circular or reciprocating. We verify that orbits initially bound within a stable boson stars persist in their bound states. For a collapsing boson star, we show that orbits initially bound and reciprocating finally either become unbound or plunge into the newly formed black hole, depending on their initial maximal radii. For initially circular geodesics, orbits with small radii plunge into the black hole; orbits with intermediate radii become unbound; and large radii orbits linger around the black hole with nonvanishing eccentricities. For the migrating case, a black hole does not form. In this case, the reciprocating orbits span a wider radial range. For initially circular geodesics, orbits with small radii become unbound, and orbits with large radii remain bound with nonvanishing eccentricities. This geodesic study provides a novel approach to investigate the gravitational collapse and migration of boson stars., Comment: 6 pages, 5 figures, comments are welcome

Numerical relativity waveforms are a critical resource in the quest to deepen our understanding of the dynamics of, and gravitational waves emitted from, merging binary systems. We present 181 new numerical relativity simulations as the second MAYA catalog of binary black hole waveforms (a sequel to the Georgia Tech waveform catalog). Most importantly, these include 55 high mass ratio (q >= 4), 48 precessing, and 92 eccentric (e > 0.01) simulations, including 7 simulations which are both eccentric and precessing. With these significant additions, this new catalog fills in considerable gaps in existing public numerical relativity waveform catalogs. The waveforms presented in this catalog are shown to be convergent and are consistent with current gravitational wave models. They are available to the public at https://cgp.ph.utexas.edu/waveform., Comment: 11 pages, 9 figures

Due to turbulence in the atmosphere images taken from ground-based telescopes become distorted. With adaptive optics (AO) images can be given greater clarity allowing for better observations with existing telescopes and are essential for ground-based coronagraphic exoplanet imaging instruments. A disadvantage to many AO systems is that they use sensors that can not correct for non-common path aberrations. We have developed a new focal plane wavefront sensing technique to address this problem called deformable mirror (DM)-based pupil chopping. The process involves a coronagraphic or non-coronagraphic science image and a deformable mirror, which modulates the phase by applying a local tip/tilt every other frame which enables correcting for leftover aberrations in the wavefront after a conventional AO correction. We validate this technique with both simulations (for coronagraphic and non-coronagraphic images) and testing (for non-coronagraphic images) on UCSC's Santa Cruz Extreme AO Laboratory (SEAL) testbed. We demonstrate that with as low as 250 nm of DM stroke to apply the local tip/tilt this wavefront sensor is linear for low-order Zernike modes and enables real-time control, in principle up to kHz speeds to correct for residual atmospheric turbulence., Comment: Conference Proceeding for 2023 SPIE Optics & Photonics, Techniques and Instrumentation for Detection of Exoplanets XI

Dense environments hosting compact binary mergers can leave an imprint on the gravitational-wave emission which, in turn, can be used to identify the characteristics of the environment. To demonstrate such scenario, we consider a simple setup of binary black holes with an environment consisting of a scalar-field bubble. We use this as a proxy for more realistic environments and as an example of the simplest physics beyond the standard model. We perform Bayesian inference on the numerical relativity waveforms using state-of-the-art waveform templates for black-hole mergers. In particular, we perform parameter estimation and model selection on signals from black-hole mergers with different mass ratios, total mass and loudness, and hosted by scalar-field bubbles of varying field amplitude. We find that sub-dominant gravitational wave modes emitted during the coalescence and ringdown are key to identifying environmental effects. In particular, we find that for face-on signals dominated by the quadrupole mode, the environment is only detectable if both the ringdown and the late inspiral/early merger fall in the detector band, so that inconsistencies can be found between the inferred binary parameters and those of the final black hole. For edge-on mergers we find that the environment can be detected even if only the ringdown is in band, thanks to the information encoded in the quasi-normal mode structure of the final black-hole.

Depression is a growing issue in society's mental health that affects all areas of life and can even lead to suicide. Fortunately, prevention programs can be effective in its treatment. In this context, this work proposes an automatic system for detecting depression on social media based on machine learning and natural language processing methods. This paper presents the following contributions: (i) an ensemble learning system that combines several types of text representations for depression detection, including recent advances in the field; (ii) a contextualization schema through topic and affective information; (iii) an analysis of models' energy consumption, establishing a trade-off between classification performance and overall computational costs. To assess the proposed models' effectiveness, a thorough evaluation is performed in two datasets that model depressive text. Experiments indicate that the proposed contextualization strategies can improve the classification and that approaches that use Transformers can improve the overall F-score by 2% while augmenting the energy cost a hundred times. Finally, this work paves the way for future energy-wise systems by considering both the performance classification and the energy consumption.

The purpose of this research was to know the degree of administrative knowledge, the degree of training of human resources, the degree of commitment of administrators and the degree of effectiveness of the administration for information security risk based on ISO/IEC 27001.The population consisted of 81 subjects (66 administrators and 15 ITD personnel). Those evaluated were employers of the administrative office of the university and also staff of the Information Technology Department (ITD). To make the comparisons, three groups of managers were formed according to classifications of administrative staff, the classification was as follows: (a) first-line manager, (b) middle management and (c) top management. About the results, it can be corroborated that administrative staff with a lower rank have more problems in making the best decisions in relation to the implementation of an ISMS, it should be noted that the first-line manager is the one who has more contact with the students and is the one who is less involved in the implementation of an ISMS. It can also be inferred that the institution\'s planners are not fully trained in the institution\'s information security efforts. This in turn prevents the generation of proposals for initiatives to implement an ISMS. With this shortcoming, it is possible that security breaches could be generated.

This research shows the analysis of multiple factors that inhibit the implementation of an Information Security Management System (ISMS). The research data were collected from 143 respondents from two universities in northeastern Mexico, in faculties of engineering in related areas. In this study, the Information Security Management System Measurement Instrument (IM-ISMS) was validated. A scale of 24 items was obtained, divided into four factors: organizational policies and regulations, privacy, integrity and authenticity. The results of this study agree with the results found by [10] in which they pre-sent a model that complies with ISO/IEC 27002:2013 controls and security and privacy criteria to improve the ISMS. [48], Mentioned that the implementation of controls based on ISO standards can meet the requirements for cybersecurity best practices.A scale of 24 items was obtained, divided into four factors: organizational policies and regulations, privacy, integrity and authenticity. This version of the instrument meets the criteria established for its validity (KMO, Bartlett's test of sphericity). An extraction was performed by the minimum residuals method, an oblique rotation was performed by the promax method, when performing the rotation 17 of the 24 items were grouped in the corresponding factor. The final reliability of the scale was calculated by the Omega coefficient, in all the dimensions the coefficients were greater than .70, therefore the re-liability of the instrument is good.

Camera pose estimation for two-view geometry traditionally relies on RANSAC. Normally, a multitude of image correspondences leads to a pool of proposed hypotheses, which are then scored to find a winning model. The inlier count is generally regarded as a reliable indicator of "consensus". We examine this scoring heuristic, and find that it favors disappointing models under certain circumstances. As a remedy, we propose the Fundamental Scoring Network (FSNet), which infers a score for a pair of overlapping images and any proposed fundamental matrix. It does not rely on sparse correspondences, but rather embodies a two-view geometry model through an epipolar attention mechanism that predicts the pose error of the two images. FSNet can be incorporated into traditional RANSAC loops. We evaluate FSNet on fundamental and essential matrix estimation on indoor and outdoor datasets, and establish that FSNet can successfully identify good poses for pairs of images with few or unreliable correspondences. Besides, we show that naively combining FSNet with MAGSAC++ scoring approach achieves state of the art results.

The University of California San Francisco Brain Metastases Stereotactic Radiosurgery (UCSF-BMSR) dataset is a public, clinical, multimodal brain MRI dataset consisting of 560 brain MRIs from 412 patients with expert annotations of 5136 brain metastases. Data consists of registered and skull stripped T1 post-contrast, T1 pre-contrast, FLAIR and subtraction (T1 pre-contrast - T1 post-contrast) images and voxelwise segmentations of enhancing brain metastases in NifTI format. The dataset also includes patient demographics, surgical status and primary cancer types. The UCSF-BSMR has been made publicly available in the hopes that researchers will use these data to push the boundaries of AI applications for brain metastases. The dataset is freely available for non-commercial use at https://imagingdatasets.ucsf.edu/dataset/1, Comment: 15 pages, 2 tables, 2 figures