Search

Real-time supercomputing performance analysis is a critical aspect of evaluating and optimizing computational systems in a dynamic user environment. The operation of supercomputers produce vast quantities of analytic data from multiple sources and of varying types so compiling this data in an efficient matter is critical to the process. MIT Lincoln Laboratory Supercomputing Center has been utilizing the Unity 3D game engine to create a Digital Twin of our supercomputing systems for several years to perform system monitoring. Unity offers robust visualization capabilities making it ideal for creating a sophisticated representation of the computational processes. As we scale the systems to include a diversity of resources such as accelerators and the addition of more users, we need to implement new analysis tools for the monitoring system. The workloads in research continuously change, as does the capability of Unity, and this allows us to adapt our monitoring tools to scale and incorporate features enabling efficient replay of system wide events, user isolation, and machine level granularity. Our system fully takes advantage of the modern capabilities of the Unity Engine in a way that intuitively represents the real time workload performed on a supercomputer. It allows HPC system engineers to quickly diagnose usage related errors with its responsive user interface which scales efficiently with large data sets.

This paper discusses the use case of energy saving and traffic steering in O-RAN, the mechanism of multi-vendor interoperability to make it work and depict its test methodology., Comment: 6 pages, 8 figures

Hypersparse traffic matrices constructed from network packet source and destination addresses is a powerful tool for gaining insights into network traffic. SuiteSparse: GraphBLAS, an open source package or building, manipulating, and analyzing large hypersparse matrices, is one approach to constructing these traffic matrices. Suricata is a widely used open source network intrusion detection software package. This work demonstrates how Suricata network flow records can be used to efficiently construct hypersparse matrices using GraphBLAS.

HPC systems used for research run a wide variety of software and workflows. This software is often written or modified by users to meet the needs of their research projects, and rarely is built with security in mind. In this paper we explore several of the key techniques that MIT Lincoln Laboratory Supercomputing Center has deployed on its systems to manage the security implications of these workflows by providing enforced separation for processes, filesystem access, network traffic, and accelerators to make every user feel like they are running on a personal HPC.

Aims. To investigate the influence of distance to filaments and dark matter halos on galaxy cold gas content in the empirical model NeutralUniverseMachine (NUM) and the hydrodynamical simulation IllustrisTNG. Methods. We use DisPerSE to identify cosmic web structures and calculate the distance of galaxies to filaments for both observations and models. We show the results of the HI and H2 mass functions, HI- and H2-halo mass relations, HI- and H2-stellar mass relations for galaxies in the NUM model and IllustrisTNG with different distances to filaments and compare them with observational measurements. We also show the evolution of HI, H2 mass densities in different distance to filament bins. Results. We find that the role of filaments in affecting the HI gas is generally less significant compared to the halo environment. There is a weak trend in the observations at z = 0 that low-mass halos lying closer to filaments tend to have reduced HI masses. However, this trend reverses for massive halos with log(Mvir/Msun) > 12.5. This behavior is accurately reproduced in the NUM model due to the dependence of HI gas on the halo formation time, but it does not appear in IllustrisTNG. The influence of filaments on the HI gas becomes slightly weaker at higher redshifts and is only significant for galaxies residing in massive halos in the NUM model. Filaments have almost no impact on the H2-stellar mass relation in both models, confirming that H2 is primarily determined by the galaxy stellar mass and star formation rate., Comment: Submitted to A&A, 10 pages, 7 figures

The MIT/IEEE/Amazon GraphChallenge encourages community approaches to developing new solutions for analyzing graphs and sparse data derived from social media, sensor feeds, and scientific data to discover relationships between events as they unfold in the field. The anonymized network sensing Graph Challenge seeks to enable large, open, community-based approaches to protecting networks. Many large-scale networking problems can only be solved with community access to very broad data sets with the highest regard for privacy and strong community buy-in. Such approaches often require community-based data sharing. In the broader networking community (commercial, federal, and academia) anonymized source-to-destination traffic matrices with standard data sharing agreements have emerged as a data product that can meet many of these requirements. This challenge provides an opportunity to highlight novel approaches for optimizing the construction and analysis of anonymized traffic matrices using over 100 billion network packets derived from the largest Internet telescope in the world (CAIDA). This challenge specifies the anonymization, construction, and analysis of these traffic matrices. A GraphBLAS reference implementation is provided, but the use of GraphBLAS is not required in this Graph Challenge. As with prior Graph Challenges the goal is to provide a well-defined context for demonstrating innovation. Graph Challenge participants are free to select (with accompanying explanation) the Graph Challenge elements that are appropriate for highlighting their innovations., Comment: Accepted to IEEE HPEC 2024

Understanding what is normal is a key aspect of protecting a domain. Other domains invest heavily in observational science to develop models of normal behavior to better detect anomalies. Recent advances in high performance graph libraries, such as the GraphBLAS, coupled with supercomputers enables processing of the trillions of observations required. We leverage this approach to synthesize low-parameter observational models of anonymized Internet traffic with a high regard for privacy., Comment: Accepted to IEEE HPEC, 7 pages, 6 figures, 1 table, 41 references

We report the discovery of Corvus A, a low-mass, gas-rich galaxy at a distance of approximately 3.5 Mpc, identified in DR10 of the Dark Energy Camera Legacy Imaging Survey during the initial phase of our ongoing SEmi-Automated Machine LEarning Search for Semi-resolved galaxies (SEAMLESS). Jansky Very Large Array observations of Corvus A detect HI line emission at a radial velocity of $523\pm2$ km/s. Magellan/Megacam imaging reveals an irregular and complex stellar population with both young and old stars. We detect UV emission in Neil Gehrels Swift observations, indicative of recent star formation. However, there are no signs of HII regions in H$\alpha$ imaging from Steward Observatory's Kuiper telescope. Based on the Megacam color magnitude diagram we measure the distance to Corvus A via the tip-of-the-red-giant-branch standard candle as $3.48\pm0.24$ Mpc. This makes Corvus A remarkably isolated, with no known galaxy within $\sim$1 Mpc. Based on this distance, we estimate the HI and stellar mass of Corvus A to be $\log M_\mathrm{HI}/\mathrm{M_\odot} = 6.59$ and $\log M_\ast/\mathrm{M_\odot} = 6.0$. Although there are some signs of rotation, the HI distribution of Corvus A appears to be close to face-on, analogous to that of Leo T, and we therefore do not attempt to infer a dynamical mass from its HI line width. Higher resolution synthesis imaging is required to confirm this morphology and to draw robust conclusions from its gas kinematics., Comment: Accepted to ApJL

One of the more complex tasks for researchers using HPC systems is performance monitoring and tuning of their applications. Developing a practice of continuous performance improvement, both for speed-up and efficient use of resources is essential to the long term success of both the HPC practitioner and the research project. Profiling tools provide a nice view of the performance of an application but often have a steep learning curve and rarely provide an easy to interpret view of resource utilization. Lower level tools such as top and htop provide a view of resource utilization for those familiar and comfortable with Linux but a barrier for newer HPC practitioners. To expand the existing profiling and job monitoring options, the MIT Lincoln Laboratory Supercomputing Center created LLoad, a tool that captures a snapshot of the resources being used by a job on a per user basis. LLload is a tool built from standard HPC tools that provides an easy way for a researcher to track resource usage of active jobs. We explain how the tool was designed and implemented and provide insight into how it is used to aid new researchers in developing their performance monitoring skills as well as guide researchers in their resource requests.

In this paper we present the European Low Frequency Survey (ELFS), a project that will enable foregrounds-free measurements of the primordial $B$-mode polarization and a detection of the tensor-to-scalar ratio, $r$, to a level $\sigma(r) = 0.001$ by measuring the Galactic and extra-galactic emissions in the 5--120\,GHz frequency window. Indeed, the main difficulty in measuring the B-mode polarization comes from the fact that many other processes in the Universe also emit polarized microwaves, which obscure the faint Cosmic Microwave Background (CMB) signal. The first stage of this project is being carried out in synergy with the Simons Array (SA) collaboration, installing a 5.5--11\,GHz (X-band) coherent receiver at the focus of one of the three 3.5\,m SA telescopes in Atacama, Chile, followed by the installation of the QUIJOTE-MFI2 in the 10--20 GHz range. We designate this initial iteration of the ELFS program as ELFS-SA. The receivers are equipped with a fully digital back-end that will provide a frequency resolution of 1\,MHz across the band, allowing us to clean the scientific signal from unwanted radio frequency interference, particularly from low-Earth orbit satellite mega constellations. This paper reviews the scientific motivation for ELFS and its instrumental characteristics, and provides an update on the development of ELFS-SA., Comment: SPIE conference on Astronomical Telescopes + Instrumentation, Yokohama, 16-22 June 2024. New version with correction in Eq. (3) arXiv admin note: text overlap with arXiv:2310.16509

The Internet has become a critical domain for modern society that requires ongoing efforts for its improvement and protection. Network traffic matrices are a powerful tool for understanding and analyzing networks and are broadly taught in online graph theory educational resources. Network traffic matrix concepts are rarely available in online computer network and cybersecurity educational resources. To fill this gap, an interactive game environment has been developed to teach the foundations of traffic matrices to the computer networking community. The game environment provides a convenient, broadly accessible, delivery mechanism that enables making material available rapidly to a wide audience. The core architecture of the game is a facility to add new network traffic matrix training modules via an easily editable JSON file. Using this facility an initial set of modules were rapidly created covering: basic traffic matrices, traffic patterns, security/defense/deterrence, a notional cyber attack, a distributed denial-of-service (DDoS) attack, and a variety of graph theory concepts. The game environment enables delivery in a wide range of contexts to enable rapid feedback and improvement. The game can be used as a core unit as part of a formal course or as a simple interactive introduction in a presentation., Comment: 9 pages, 10 figures, 52 references; accepted to IEEE GrAPL

LiDAR datasets for autonomous driving exhibit biases in properties such as point cloud density, range, and object dimensions. As a result, object detection networks trained and evaluated in different environments often experience performance degradation. Domain adaptation approaches assume access to unannotated samples from the test distribution to address this problem. However, in the real world, the exact conditions of deployment and access to samples representative of the test dataset may be unavailable while training. We argue that the more realistic and challenging formulation is to require robustness in performance to unseen target domains. We propose to address this problem in a two-pronged manner. First, we leverage paired LiDAR-image data present in most autonomous driving datasets to perform multimodal object detection. We suggest that working with multimodal features by leveraging both images and LiDAR point clouds for scene understanding tasks results in object detectors more robust to unseen domain shifts. Second, we train a 3D object detector to learn multimodal object features across different distributions and promote feature invariance across these source domains to improve generalizability to unseen target domains. To this end, we propose CLIX$^\text{3D}$, a multimodal fusion and supervised contrastive learning framework for 3D object detection that performs alignment of object features from same-class samples of different domains while pushing the features from different classes apart. We show that CLIX$^\text{3D}$ yields state-of-the-art domain generalization performance under multiple dataset shifts., Comment: technical report

Popular representation learning methods encourage feature invariance under transformations applied at the input. However, in 3D perception tasks like object localization and segmentation, outputs are naturally equivariant to some transformations, such as rotation. Using pre-training loss functions that encourage equivariance of features under certain transformations provides a strong self-supervision signal while also retaining information of geometric relationships between transformed feature representations. This can enable improved performance in downstream tasks that are equivariant to such transformations. In this paper, we propose a spatio-temporal equivariant learning framework by considering both spatial and temporal augmentations jointly. Our experiments show that the best performance arises with a pre-training approach that encourages equivariance to translation, scaling, and flip, rotation and scene flow. For spatial augmentations, we find that depending on the transformation, either a contrastive objective or an equivariance-by-classification objective yields best results. To leverage real-world object deformations and motion, we consider sequential LiDAR scene pairs and develop a novel 3D scene flow-based equivariance objective that leads to improved performance overall. We show our pre-training method for 3D object detection which outperforms existing equivariant and invariant approaches in many settings., Comment: technical report

Spin qubits in silicon are strong contenders for realizing a practical quantum computer. This technology has made remarkable progress with the demonstration of single and two-qubit gates above the fault-tolerant threshold and entanglement of up to three qubits. However, maintaining high fidelity operations while executing multi-qubit algorithms has remained elusive, only being achieved for two spin qubits to date due to the small qubit size, which makes it difficult to control qubits without creating crosstalk errors. Here, we use a four-qubit silicon processor with every operation above the fault tolerant limit and demonstrate Grover's algorithm with a ~95% probability of finding the marked state, one of the most successful implementations to date. Our four-qubit processor is made of three phosphorus atoms and one electron spin precision-patterned into 1.5 nm${}^2$ isotopically pure silicon. The strong resulting confinement potential, without additional confinement gates that can increase cross-talk, leverages the benefits of having both electron and phosphorus nuclear spins. Significantly, the all-to-all connectivity of the nuclear spins provided by the hyperfine interaction not only allows for efficient multi-qubit operations, but also provides individual qubit addressability. Together with the long coherence times of the nuclear and electron spins, this results in all four single qubit fidelities above 99.9% and controlled-Z gates between all pairs of nuclear spins above 99% fidelity. The high control fidelities, combined with >99% fidelity readout of all nuclear spins, allows for the creation of a three-qubit Greenberger-Horne-Zeilinger (GHZ) state with 96.2% fidelity, the highest reported for semiconductor spin qubits so far. Such nuclear spin registers can be coupled via electron exchange, establishing a path for larger scale fault-tolerant quantum processors., Comment: 16 pages, 9 figures, 3 tables

As research and deployment of AI grows, the computational burden to support and sustain its progress inevitably does too. To train or fine-tune state-of-the-art models in NLP, computer vision, etc., some form of AI hardware acceleration is virtually a requirement. Recent large language models require considerable resources to train and deploy, resulting in significant energy usage, potential carbon emissions, and massive demand for GPUs and other hardware accelerators. However, this surge carries large implications for energy sustainability at the HPC/datacenter level. In this paper, we study the aggregate effect of power-capping GPUs on GPU temperature and power draw at a research supercomputing center. With the right amount of power-capping, we show significant decreases in both temperature and power draw, reducing power consumption and potentially improving hardware life-span with minimal impact on job performance. While power-capping reduces power draw by design, the aggregate system-wide effect on overall energy consumption is less clear; for instance, if users notice job performance degradation from GPU power-caps, they may request additional GPU-jobs to compensate, negating any energy savings or even worsening energy consumption. To our knowledge, our work is the first to conduct and make available a detailed analysis of the effects of GPU power-capping at the supercomputing scale. We hope our work will inspire HPCs/datacenters to further explore, evaluate, and communicate the impact of power-capping AI hardware accelerators for more sustainable AI.

We have discovered the stellar counterpart to the ALFALFA Virgo 7 cloud complex, which has been thought to be optically dark and nearly star-free since its discovery in 2007. This ~190 kpc long chain of enormous atomic gas clouds ($M_\mathrm{HI} \sim 10^9 \; \mathrm{M_\odot}$) is embedded in the hot intracluster medium of the Virgo galaxy cluster but is isolated from any galaxy. Its faint, blue stellar counterpart, BC6, was identified in a visual search of archival optical and UV imaging. Follow-up observations with the Green Bank Telescope, Hobby-Eberly Telescope, and Hubble Space Telescope demonstrate that this faint counterpart is at the same velocity as the atomic gas, is actively forming stars, and is metal-rich ($12 + \mathrm{(O/H)} = 8.58 \pm 0.25$). We estimate its stellar mass to be only $\log ( M_\ast/\mathrm{M_\odot}) \sim 4.4$, making it one of the most gas-rich stellar systems known. Aside from its extraordinary gas content, the properties of BC6 are entirely consistent with those of a recently identified class of young, low-mass, isolated, and star-forming clouds in Virgo, that appear to have formed via extreme ram pressure stripping events. We expand the existing discussion of the origin of this structure and suggest NGC 4522 as a likely candidate, however, the current evidence is not fully consistent with any of our proposed progenitor galaxies. We anticipate that other "dark" gas clouds in Virgo may have similarly faint, star-forming counterparts. We aim to identify these through the help of a citizen science search of the entire cluster., Comment: Accepted to ApJL

We have conducted a systematic search around the Milky Way (MW) analog NGC 253 (D=3.5 Mpc), as a part of the Panoramic Imaging Survey of Centaurus and Sculptor (PISCeS) - a Magellan+Megacam survey to identify dwarfs and other substructures in resolved stellar light around MW-mass galaxies outside of the Local Group. In total, NGC 253 has five satellites identified by PISCeS within 100 kpc with an absolute V-band magnitude $M_V<-7$. We have additionally obtained deep Hubble Space Telescope imaging of four reported candidates beyond the survey footprint: Do III, Do IV, and dw0036m2828 are confirmed to be satellites of NGC 253, while SculptorSR is found to be a background galaxy. We find no convincing evidence for the presence of a plane of satellites surrounding NGC 253. We construct its satellite luminosity function, which is complete down to $M_V$$\lesssim$$-8$ out to 100 kpc and $M_V$$\lesssim$$-9$ out to 300 kpc, and compare it to those calculated for other Local Volume galaxies. Exploring trends in satellite counts and star-forming fractions among satellite systems, we find relationships with host stellar mass, environment, and morphology, pointing to a complex picture of satellite formation, and a successful model has to reproduce all of these trends., Comment: Submitted to AAS Journal. Comments are welcome

We present new follow-up observations of two ultra-diffuse galaxies (UDGs), part of a total sample of five chosen for their distorted morphologies, suggestive of tidal influence. Using Hubble Space Telescope Advanced Camera for Surveys F555W and F814W imaging, we identify 8+/-2 globular clusters (GCs) in KUG 0203-Dw1 and 6+/-2 in KDG 013, abundances that are fairly typical for normal dwarf galaxies of similar stellar mass. Jansky Very Large Array data reveal a clear HI detection of KUG 0203-Dw1 with a gas mass estimate of log(MHI/Msun) < 7.4 and evidence of active stripping by the host. HI gas is found near the location of KDG~013 but is likely unrelated to the UDG itself due to the morphology and the numerous gas tails within the host group. Given that these UDGs have GC abundances typical for galaxies at their luminosity, these findings suggest that they likely originated as normal dwarf galaxies that have been subjected to significant stripping and tidal heating, causing them to become more diffuse. These two UDGs complete a sample of five exhibiting tidal features in the Canada-France-Hawaii Telescope Legacy Survey area (CFHTLS; ~150 sq deg), including UDGs with and without UV emission, indicative of recent star formation. Four UDGs in this sample, consistent with dwarfs `puffed-up' by tidal interactions, contrast with an outlier, suggesting a dwarf merger origin. These findings indicate that tidal heating of dwarfs is a viable formation pathway for UDGs., Comment: 18 pages, 6 figures, submitted to AJ

We analyze a little-known property of apportionment methods that captures how allocations scale with the size of the house: specifically, if, for a fixed population distribution, the house size and allocation can be scaled down within the set of integers, then the apportionment should be correspondingly scaled down. Balinski and Young (2001) include this property among the minimal requirements for a "reasonable" apportionment method. We argue that this property is better understood as a consistency requirement since quota-based apportionments that are "less proportional" meet this requirement while others that are "more proportional" do not. We also show that the family of quotatone methods based on stationary divisors (including the quota method) do not satisfy this property.

The accurate computation of properties of large molecular systems is classically infeasible and is one of the applications in which it is hoped that quantum computers will demonstrate an advantage over classical devices. However, due to the limitations of present-day quantum hardware, variational-hybrid algorithms introduced to tackle these problems struggle to meet the accuracy and precision requirements of chemical applications. Here, we apply the Quantum Computed Moments (QCM) approach combined with a variety of noise-mitigation techniques to an 8 qubit/spin-orbital representation of the water molecule (H$_2$O). A noise-stable improvement on the variational result for a 4-excitation trial-state (circuit depth 25, 22 CNOTs) was obtained, with the ground-state energy computed to be within $1.4\pm1.2$ mHa of exact diagonalisation in the 14 spin-orbital basis. Thus, the QCM approach, despite an increased number of measurements and noisy quantum hardware (CNOT error rates c.1% corresponding to expected error rates on the trial-state circuit of order 20%), is able to determine the ground-state energy of a non-trivial molecular system at the required accuracy (c.0.1%). To the best of our knowledge, these results are the largest calculations performed on a physical quantum computer to date in terms of encoding individual spin-orbitals producing chemically relevant accuracy, and a promising indicator of how such hybrid approaches might scale to problems of interest in the low-error/fault-tolerant regimes as quantum computers develop., Comment: 20 pages, 8 figures (incl. SI)

We have imaged the entirety of eight (plus one partial) Milky Way-like satellite systems, a total of 42 (45) satellites, from the Satellites Around Galactic Analogs (SAGA) II catalog in both H$\alpha$ and HI with the Canada-France-Hawaii Telescope and the Jansky Very Large Array. In these eight systems we have identified four cases where a satellite appears to be currently undergoing ram pressure stripping (RPS) as its HI gas collides with the circumgalactic medium (CGM) of its host. We also see a clear suppression of gas fraction ($M_\mathrm{HI}/M_\ast$) with decreasing (projected) satellite--host separation; to our knowledge, the first time this has been observed in a sample of Milky Way-like systems. Comparisons to the Auriga, APOSTLE, and TNG50 cosmological zoom-in simulations show consistent global behavior, but they systematically under-predict gas fractions across all satellites by roughly 0.5 dex. Using a simplistic RPS model we estimate the average peak CGM density that satellites in these systems have encountered to be $\log \rho_\mathrm{cgm}/\mathrm{g\,cm^{-3}} \approx -27.3$. Furthermore, we see tentative evidence that these satellites are following a specific star formation rate-to-gas fraction relation that is distinct from field galaxies. Finally, we detect one new gas-rich satellite in the UGC903 system with an optical size and surface brightness meeting the standard criteria to be considered an ultra-diffuse galaxy., Comment: Accepted to ApJ

In this paper we present the European Low Frequency Survey (ELFS), a project that will enable foregrounds-free measurements of primordial $B$-mode polarization to a level 10$^{-3}$ by measuring the Galactic and extra-Galactic emissions in the 5--120\,GHz frequency window. Indeed, the main difficulty in measuring the B-mode polarization comes not just from its sheer faintness, but from the fact that many other objects in the Universe also emit polarized microwaves, which mask the faint CMB signal. The first stage of this project will be carried out in synergy with the Simons Array (SA) collaboration, installing a 5.5--11 GHz coherent receiver at the focus of one of the three 3.5\,m SA telescopes in Atacama, Chile ("ELFS on SA"). The receiver will be equipped with a fully digital back-end based on the latest Xilinx RF System-on-Chip devices that will provide frequency resolution of 1\,MHz across the whole observing band, allowing us to clean the scientific signal from unwanted radio frequency interference, particularly from low-Earth orbit satellite mega-constellations. This paper reviews the scientific motivation for ELFS and its instrumental characteristics, and provides an update on the development of ELFS on SA., Comment: to appear in Proc. of the mm Universe 2023 conference, Grenoble (France), June 2023, published by F. Mayet et al. (Eds), EPJ Web of conferences, EDP Sciences

Low-power small form factor data processing units (DPUs) enable offloading and acceleration of a broad range of networking and security services. DPUs have accelerated the transition to programmable networking by enabling the replacement of FPGAs/ASICs in a wide range of network oriented devices. The GraphBLAS sparse matrix graph open standard math library is well-suited for constructing anonymized hypersparse traffic matrices of network traffic which can enable a wide range of network analytics. This paper measures the performance of the GraphBLAS on an ARM based NVIDIA DPU (BlueField 2) and, to the best of our knowledge, represents the first reported GraphBLAS results on a DPU and/or ARM based system. Anonymized hypersparse traffic matrices were constructed at a rate of over 18 million packets per second.

This paper is an update of the survey of AI accelerators and processors from past four years, which is now called the Lincoln AI Computing Survey - LAICS (pronounced "lace"). As in past years, this paper collects and summarizes the current commercial accelerators that have been publicly announced with peak performance and peak power consumption numbers. The performance and power values are plotted on a scatter graph, and a number of dimensions and observations from the trends on this plot are again discussed and analyzed. Market segments are highlighted on the scatter plot, and zoomed plots of each segment are also included. Finally, a brief description of each of the new accelerators that have been added in the survey this year is included., Comment: 7 pages, 6 figures, 2023 IEEE High Performance Extreme Computing (HPEC) conference, September 2023

Large language models (LLMs) have exploded in popularity due to their new generative capabilities that go far beyond prior state-of-the-art. These technologies are increasingly being leveraged in various domains such as law, finance, and medicine. However, these models carry significant computational challenges, especially the compute and energy costs required for inference. Inference energy costs already receive less attention than the energy costs of training LLMs -- despite how often these large models are called on to conduct inference in reality (e.g., ChatGPT). As these state-of-the-art LLMs see increasing usage and deployment in various domains, a better understanding of their resource utilization is crucial for cost-savings, scaling performance, efficient hardware usage, and optimal inference strategies. In this paper, we describe experiments conducted to study the computational and energy utilization of inference with LLMs. We benchmark and conduct a preliminary analysis of the inference performance and inference energy costs of different sizes of LLaMA -- a recent state-of-the-art LLM -- developed by Meta AI on two generations of popular GPUs (NVIDIA V100 \& A100) and two datasets (Alpaca and GSM8K) to reflect the diverse set of tasks/benchmarks for LLMs in research and practice. We present the results of multi-node, multi-GPU inference using model sharding across up to 32 GPUs. To our knowledge, our work is the one of the first to study LLM inference performance from the perspective of computational and energy resources at this scale.

We report the discovery of Pavo, a faint ($M_V = -10.0$), star-forming, irregular, and extremely isolated dwarf galaxy at $D\approx2$ Mpc. Pavo was identified in Dark Energy Camera Legacy Survey imaging via a novel approach that combines low surface brightness galaxy search algorithms and machine learning candidate classifications. Follow-up imaging with the Inamori-Magellan Areal Camera & Spectrograph on the 6.5 m Magellan Baade telescope revealed a color--magnitude diagram (CMD) with an old stellar population, in addition to the young population that dominates the integrated light, and a tip-of-the-red-giant-branch distance estimate of $1.99^{+0.20}_{-0.22}$ Mpc. The blue population of stars in the CMD is consistent with the youngest stars having formed no later than 150 Myr ago. We also detected no H$\alpha$ emission with SOAR telescope imaging, suggesting we may be witnessing a temporary low in Pavo's star formation. We estimate the total stellar mass of Pavo to be $\log M_\ast/\mathrm{M_\odot} = 5.6 \pm 0.2$ and measure an upper limit on its HI gas mass of $1.0 \times 10^6\,\mathrm{M_\odot}$ based on the HIPASS survey. Given these properties, Pavo's closest analog is Leo P ($D=1.6$ Mpc), previously the only known isolated, star-forming, Local Volume dwarf galaxy in this mass range. However, Pavo appears to be even more isolated, with no other known galaxy residing within over 600 kpc. As surveys and search techniques continue to improve, we anticipate an entire population of analogous objects being detected just outside the Local Group., Comment: Accepted to ApJL

Expanding the scientific tools available to protect computer networks can be aided by a deeper understanding of the underlying statistical distributions of network traffic and their potential geometric interpretations. Analyses of large scale network observations provide a unique window into studying those underlying statistics. Newly developed GraphBLAS hypersparse matrices and D4M associative array technologies enable the efficient anonymized analysis of network traffic on the scale of trillions of events. This work analyzes over 100,000,000,000 anonymized packets from the largest Internet telescope (CAIDA) and over 10,000,000 anonymized sources from the largest commercial honeyfarm (GreyNoise). Neither CAIDA nor GreyNoise actively emit Internet traffic and provide distinct observations of unsolicited Internet traffic (primarily botnets and scanners). Analysis of these observations confirms the previously observed Cauchy-like distributions describing temporal correlations between Internet sources. The Gull lighthouse problem is a well-known geometric characterization of the standard Cauchy distribution and motivates a potential geometric interpretation for Internet observations. This work generalizes the Gull lighthouse problem to accommodate larger classes of coastlines, deriving a closed-form solution for the resulting probability distributions, stating and examining the inverse problem of identifying an appropriate coastline given a continuous probability distribution, identifying a geometric heuristic for solving this problem computationally, and applying that heuristic to examine the temporal geometry of different subsets of network observations. Application of this method to the CAIDA and GreyNoise data reveals a several orders of magnitude difference between known benign and other traffic which can lead to potentially novel ways to protect networks., Comment: 9 pages, 7 figures, IEEE HPEC 2023 (accepted)

The primary bottleneck towards obtaining good recognition performance in IR images is the lack of sufficient labeled training data, owing to the cost of acquiring such data. Realizing that object detection methods for the RGB modality are quite robust (at least for some commonplace classes, like person, car, etc.), thanks to the giant training sets that exist, in this work we seek to leverage cues from the RGB modality to scale object detectors to the IR modality, while preserving model performance in the RGB modality. At the core of our method, is a novel tensor decomposition method called TensorFact which splits the convolution kernels of a layer of a Convolutional Neural Network (CNN) into low-rank factor matrices, with fewer parameters than the original CNN. We first pretrain these factor matrices on the RGB modality, for which plenty of training data are assumed to exist and then augment only a few trainable parameters for training on the IR modality to avoid over-fitting, while encouraging them to capture complementary cues from those trained only on the RGB modality. We validate our approach empirically by first assessing how well our TensorFact decomposed network performs at the task of detecting objects in RGB images vis-a-vis the original network and then look at how well it adapts to IR images of the FLIR ADAS v1 dataset. For the latter, we train models under scenarios that pose challenges stemming from data paucity. From the experiments, we observe that: (i) TensorFact shows performance gains on RGB images; (ii) further, this pre-trained model, when fine-tuned, outperforms a standard state-of-the-art object detector on the FLIR ADAS v1 dataset by about 4% in terms of mAP 50 score., Comment: Accepted to ICCV 2023, LIMIT Workshop. The first two authors contributed equally

Prototypical part neural networks (ProtoPartNNs), namely PROTOPNET and its derivatives, are an intrinsically interpretable approach to machine learning. Their prototype learning scheme enables intuitive explanations of the form, this (prototype) looks like that (testing image patch). But, does this actually look like that? In this work, we delve into why object part localization and associated heat maps in past work are misleading. Rather than localizing to object parts, existing ProtoPartNNs localize to the entire image, contrary to generated explanatory visualizations. We argue that detraction from these underlying issues is due to the alluring nature of visualizations and an over-reliance on intuition. To alleviate these issues, we devise new receptive field-based architectural constraints for meaningful localization and a principled pixel space mapping for ProtoPartNNs. To improve interpretability, we propose additional architectural improvements, including a simplified classification head. We also make additional corrections to PROTOPNET and its derivatives, such as the use of a validation set, rather than a test set, to evaluate generalization during training. Our approach, PIXPNET (Pixel-grounded Prototypical part Network), is the only ProtoPartNN that truly learns and localizes to prototypical object parts. We demonstrate that PIXPNET achieves quantifiably improved interpretability without sacrificing accuracy., Comment: 21 pages

We present an approach to estimating camera rotation in crowded, real-world scenes from handheld monocular video. While camera rotation estimation is a well-studied problem, no previous methods exhibit both high accuracy and acceptable speed in this setting. Because the setting is not addressed well by other datasets, we provide a new dataset and benchmark, with high-accuracy, rigorously verified ground truth, on 17 video sequences. Methods developed for wide baseline stereo (e.g., 5-point methods) perform poorly on monocular video. On the other hand, methods used in autonomous driving (e.g., SLAM) leverage specific sensor setups, specific motion models, or local optimization strategies (lagging batch processing) and do not generalize well to handheld video. Finally, for dynamic scenes, commonly used robustification techniques like RANSAC require large numbers of iterations, and become prohibitively slow. We introduce a novel generalization of the Hough transform on SO(3) to efficiently and robustly find the camera rotation most compatible with optical flow. Among comparably fast methods, ours reduces error by almost 50\% over the next best, and is more accurate than any method, irrespective of speed. This represents a strong new performance point for crowded scenes, an important setting for computer vision. The code and the dataset are available at https://fabiendelattre.com/robust-rotation-estimation., Comment: Published at ICCV 2023

Reducing unexpected urban traffic congestion caused by en-route events (e.g., road closures, car crashes, etc.) often requires fast and accurate reactions to choose the best-fit traffic signals. Traditional traffic light control systems, such as SCATS and SCOOT, are not efficient as their traffic data provided by induction loops has a low update frequency (i.e., longer than 1 minute). Moreover, the traffic light signal plans used by these systems are selected from a limited set of candidate plans pre-programmed prior to unexpected events' occurrence. Recent research demonstrates that camera-based traffic light systems controlled by deep reinforcement learning (DRL) algorithms are more effective in reducing traffic congestion, in which the cameras can provide high-frequency high-resolution traffic data. However, these systems are costly to deploy in big cities due to the excessive potential upgrades required to road infrastructure. In this paper, we argue that Unmanned Aerial Vehicles (UAVs) can play a crucial role in dealing with unexpected traffic congestion because UAVs with onboard cameras can be economically deployed when and where unexpected congestion occurs. Then, we propose a system called "AVARS" that explores the potential of using UAVs to reduce unexpected urban traffic congestion using DRL-based traffic light signal control. This approach is validated on a widely used open-source traffic simulator with practical UAV settings, including its traffic monitoring ranges and battery lifetime. Our simulation results show that AVARS can effectively recover the unexpected traffic congestion in Dublin, Ireland, back to its original un-congested level within the typical battery life duration of a UAV.

pPython seeks to provide a parallel capability that provides good speed-up without sacrificing the ease of programming in Python by implementing partitioned global array semantics (PGAS) on top of a simple file-based messaging library (PythonMPI) in pure Python. pPython follows a SPMD (single program multiple data) model of computation. pPython runs on a single-node (e.g., a laptop) running Windows, Linux, or MacOS operating systems or on any combination of heterogeneous systems that support Python, including on a cluster through a Slurm scheduler interface so that pPython can be executed in a massively parallel computing environment. It is interesting to see what performance pPython can achieve compared to the traditional socket-based MPI communication because of its unique file-based messaging implementation. In this paper, we present the point-to-point and collective communication performances of pPython and compare them with those obtained by using mpi4py with OpenMPI. For large messages, pPython demonstrates comparable performance as compared to mpi4py., Comment: arXiv admin note: substantial text overlap with arXiv:2208.14908

Matrix/array analysis of networks can provide significant insight into their behavior and aid in their operation and protection. Prior work has demonstrated the analytic, performance, and compression capabilities of GraphBLAS (graphblas.org) hypersparse matrices and D4M (d4m.mit.edu) associative arrays (a mathematical superset of matrices). Obtaining the benefits of these capabilities requires integrating them into operational systems, which comes with its own unique challenges. This paper describes two examples of real-time operational implementations. First, is an operational GraphBLAS implementation that constructs anonymized hypersparse matrices on a high-bandwidth network tap. Second, is an operational D4M implementation that analyzes daily cloud gateway logs. The architectures of these implementations are presented. Detailed measurements of the resources and the performance are collected and analyzed. The implementations are capable of meeting their operational requirements using modest computational resources (a couple of processing cores). GraphBLAS is well-suited for low-level analysis of high-bandwidth connections with relatively structured network data. D4M is well-suited for higher-level analysis of more unstructured data. This work demonstrates that these technologies can be implemented in operational settings., Comment: Accepted to IEEE HPEC, 8 pages, 8 figures, 1 table, 69 references. arXiv admin note: text overlap with arXiv:2203.13934. text overlap with arXiv:2309.01806

Defending community-owned cyber space requires community-based efforts. Large-scale network observations that uphold the highest regard for privacy are key to protecting our shared cyberspace. Deployment of the necessary network sensors requires careful sensor placement, focusing, and calibration with significant volumes of network observations. This paper demonstrates novel focusing and calibration procedures on a multi-billion packet dataset using high-performance GraphBLAS anonymized hypersparse matrices. The run-time performance on a real-world data set confirms previously observed real-time processing rates for high-bandwidth links while achieving significant data compression. The output of the analysis demonstrates the effectiveness of these procedures at focusing the traffic matrix and revealing the underlying stable heavy-tail statistical distributions that are necessary for anomaly detection. A simple model of the corresponding probability of detection ($p_{\rm d}$) and probability of false alarm ($p_{\rm fa}$) for these distributions highlights the criticality of network sensor focusing and calibration. Once a sensor is properly focused and calibrated it is then in a position to carry out two of the central tenets of good cybersecurity: (1) continuous observation of the network and (2) minimizing unbrokered network connections., Comment: Accepted to IEEE HPEC, 9 pages, 12 figures, 1 table, 63 references, 2 appendices

Coarse-grained molecular models of proteins permit access to length and time scales unattainable by all-atom models and the simulation of processes that occur on long-time scales such as aggregation and folding. The reduced resolution realizes computational accelerations but an atomistic representation can be vital for a complete understanding of mechanistic details. Backmapping is the process of restoring all-atom resolution to coarse-grained molecular models. In this work, we report DiAMoNDBack (Diffusion-denoising Autoregressive Model for Non-Deterministic Backmapping) as an autoregressive denoising diffusion probability model to restore all-atom details to coarse-grained protein representations retaining only C{\alpha} coordinates. The autoregressive generation process proceeds from the protein N-terminus to C-terminus in a residue-by-residue fashion conditioned on the C{\alpha} trace and previously backmapped backbone and side chain atoms within the local neighborhood. The local and autoregressive nature of our model makes it transferable between proteins. The stochastic nature of the denoising diffusion process means that the model generates a realistic ensemble of backbone and side chain all-atom configurations consistent with the coarse-grained C{\alpha} trace. We train DiAMoNDBack over 65k+ structures from Protein Data Bank (PDB) and validate it in applications to a hold-out PDB test set, intrinsically-disordered protein structures from the Protein Ensemble Database (PED), molecular dynamics simulations of fast-folding mini-proteins from DE Shaw Research, and coarse-grained simulation data. We achieve state-of-the-art reconstruction performance in terms of correct bond formation, avoidance of side chain clashes, and diversity of the generated side chain configurational states. We make DiAMoNDBack model publicly available as a free and open source Python package.

Accurately modeling the cold gas content in the universe is challenging for current theoretical models. We propose a new empirical model NeutralUniverseMachine for the evolution of HI and H$_2$ gas along with dark matter halos based on the UniverseMachine catalog. It is able to accurately describe the observed HI and H$_2$ mass functions, molecular-to-atomic ratio, HI-halo mass relation, HI/H$_2$-stellar mass relations at $z\sim0$, as well as the evolution of cosmic gas densities $\rho_{\rm HI}$ and $\rho_{\rm H_2}$ in $0

The Nancy Grace Roman Space Telescope is capable of delivering an unprecedented all-sky, high-spatial resolution, multi-epoch infrared map to the astronomical community. This opportunity arises in the midst of numerous ground- and space-based surveys that will provide extensive spectroscopy and imaging together covering the entire sky (such as Rubin/LSST, Euclid, UNIONS, SPHEREx, DESI, SDSS-V, GALAH, 4MOST, WEAVE, MOONS, PFS, UVEX, NEO Surveyor, etc.). Roman can uniquely provide uniform high-spatial-resolution (~0.1 arcsec) imaging over the entire sky, vastly expanding the science reach and precision of all of these near-term and future surveys. This imaging will not only enhance other surveys, but also facilitate completely new science. By imaging the full sky over two epochs, Roman can measure the proper motions for stars across the entire Milky Way, probing 100 times fainter than Gaia out to the very edge of the Galaxy. Here, we propose NANCY: a completely public, all-sky survey that will create a high-value legacy dataset benefiting innumerable ongoing and forthcoming studies of the universe. NANCY is a pure expression of Roman's potential: it images the entire sky, at high spatial resolution, in a broad infrared bandpass that collects as many photons as possible. The majority of all ongoing astronomical surveys would benefit from incorporating observations of NANCY into their analyses, whether these surveys focus on nearby stars, the Milky Way, near-field cosmology, or the broader universe., Comment: Submitted to the call for white papers for the Roman Core Community Survey (June 16th, 2023), and to the Bulletin of the AAS

We investigate the ultra-diffuse galaxy (UDG) UGC 9050-Dw1, which was selected because of its disturbed morphology as part of a larger sample of UDGs that display evidence for significant interactions. We use the Hubble Space Telescope's Advanced Camera for Surveys to identify globular clusters (GCs) associated with UGC 9050-Dw1, and the Jansky Very Large Array to measure its $\mathrm{H}\mathrm{I}$ content. UGC 9050-Dw1, a neighbor to the low surface brightness spiral UGC 9050, exhibits a unique UV bright central ``clump'' with clearly associated $\mathrm{H}{\mathrm{I}}$ gas and an extended stellar tidal plume to the north. We identify $52^{+4}_{-6}$ GCs, implying a specific frequency $S_\mathrm{N} = 122_{-24}^{+30}$, one of the highest reported for a UDG of this luminosity. Additionally, $\sim 20\%$ of the total light of the galaxy is contributed by GCs. Nearly uniform GC colors suggest they were formed during a single intense episode of star formation. We propose that UGC 9050-Dw1 formed via a rare dwarf merger event where induced, clumpy star formation led to its current observed properties., Comment: 23 pages, 9 figures, published in ApJL

In 2009, Crisp, Godelle and Wiest constructed a linear-time algorithm to solve the conjugacy problem in right-angled Artin groups. This algorithm has now been implemented in Python, and the code is freely available on GitHub. This document provides a summary of how the code works. As well as determining whether two elements $w_{1}, w_{2}$ are conjugate in a RAAG $A_{\Gamma}$, our code also returns a conjugating element $x \in A_{\Gamma}$ such that $w_{1} = x^{-1}w_{2}x$, if $w_{1}$ and $w_{2}$ are conjugate., Comment: Comments welcome!

We present the first measurement of the HI mass function (HIMF) using data from MeerKAT, based on 276 direct detections from the MIGHTEE Survey Early Science data covering a period of approximately a billion years ($0 \leq z \leq 0.084 $). This is the first HIMF measured using interferometric data over non-group or cluster field, i.e. a deep blank field. We constrain the parameters of the Schechter function which describes the HIMF with two different methods: $1/\rm V_{\rm max}$ and Modified Maximum Likelihood (MML). We find a low-mass slope $\alpha=-1.29^{+0.37}_{-0.26}$, `knee' mass $\log_{10}(M_{*}/{\rm M_{\odot}}) = 10.07^{+0.24}_{-0.24}$ and normalisation $\log_{10}(\phi_{*}/\rm Mpc^{-3})=-2.34^{+0.32}_{-0.36}$ ($H_0 = 67.4$ kms$^{-1}$ Mpc$^{-1}$) for $1/\rm V_{\rm max}$ and $\alpha=-1.44^{+0.13}_{-0.10}$, `knee' mass $\log_{10}(M_{*}/{\rm M_{\odot}}) = 10.22^{+0.10}_{-0.13}$ and normalisation $\log_{10}(\phi_{*}/\rm Mpc^{-3})=-2.52^{+0.19}_{-0.14}$ for MML. When using $1/\rm V_{\rm max}$ we find both the low-mass slope and `knee' mass to be consistent within $1\sigma$ with previous studies based on single-dish surveys. The cosmological mass density of HI is found to be slightly larger than previously reported: $\Omega_{\rm HI}=5.46^{+0.94}_{-0.99} \times 10^{-4}h^{-1}_{67.4}$ from $1/\rm V_{\rm max}$ and $\Omega_{\rm HI}=6.31^{+0.31}_{-0.31} \times 10^{-4}h^{-1}_{67.4}$ from MML but consistent within the uncertainties. We find no evidence for evolution of the HIMF over the last billion years., Comment: 13 pages, 9 figures, accepted for publication in MNRAS

The Andromeda galaxy (M31) is our closest neighbouring spiral galaxy, making it an ideal target for studying the physics of the interstellar medium in a galaxy very similar to our own. Using new observations of M31 at 4.76GHz by the C-Band All-Sky Survey (C-BASS), and all available radio data at $1^\circ$ resolution, we produce the integrated spectrum and put new constraints on the synchrotron spectral index and anomalous microwave emission (AME) from M31. We use aperture photometry and spectral modelling to fit for the integrated spectrum of M31, and subtract a comprehensive model of nearby background radio sources. The AME in M31 is detected at $3\sigma$ significance with a peak near 30GHz and flux density $0.27\pm0.09$Jy. The synchrotron spectral index of M31 is flatter than our own Galaxy at $\alpha = -0.66 \pm 0.03$ with no strong evidence of spectral curvature. The emissivity of AME, averaged over the total emission from M31 is lower than typical AME sources in our Galaxy, implying that AME is not uniformly distributed throughout M31 and instead is likely confined to sub-regions -- this will need to be confirmed using future higher resolution observations around 20--30GHz., Comment: 16 pages, 6 figures. Accepted by MNRAS