Searcharxiv⌕ Search

arXiv subjects

Andres Meza

Publications and source records attributed to Andres Meza.

At least 19 recordsLinked to original sources
arXiv2024

AlphaChimp: Tracking and Behavior Recognition of Chimpanzees

Understanding non-human primate behavior is crucial for improving animal welfare, modeling social behavior, and gaining insights into both distinctly human and shared behaviors. Despite recent advances in computer vision, automated analysis of primate behavior remains challenging due to the complexity of their social interactions and the lack of specialized algorithms. Existing methods often struggle with the nuanced behaviors and frequent occlusions characteristic of primate social dynamics. This study aims to develop an effective method for automated detection, tracking, and recognition of chimpanzee behaviors in video footage. Here we show that our proposed method, AlphaChimp, an end-to-end approach that simultaneously detects chimpanzee positions and estimates behavior categories from videos, significantly outperforms existing methods in behavior recognition. AlphaChimp achieves approximately 10% higher tracking accuracy and a 20% improvement in behavior recognition compared to state-of-the-art methods, particularly excelling in the recognition of social behaviors. This superior performance stems from AlphaChimp's innovative architecture, which integrates temporal feature fusion with a Transformer-based self-attention mechanism, enabling more effective capture and interpretation of complex social interactions among chimpanzees. Our approach bridges the gap between computer vision and primatology, enhancing technical capabilities and deepening our understanding of primate communication and sociality. We release our code and models and hope this will facilitate future research in animal social dynamics. This work contributes to ethology, cognitive science, and artificial intelligence, offering new perspectives on social intelligence.

cs.CV↗
arXiv2024

Reliable edge machine learning hardware for scientific applications

Extreme data rate scientific experiments create massive amounts of data that require efficient ML edge processing. This leads to unique validation challenges for VLSI implementations of ML algorithms: enabling bit-accurate functional simulations for performance validation in experimental software frameworks, verifying those ML models are robust under extreme quantization and pruning, and enabling ultra-fine-grained model inspection for efficient fault tolerance. We discuss approaches to developing and validating reliable algorithms at the scientific edge under such strict latency, resource, power, and area requirements in extreme experimental environments. We study metrics for developing robust algorithms, present preliminary results and mitigation strategies, and conclude with an outlook of these and future directions of research towards the longer-term goal of developing autonomous scientific experimentation methods for accelerated scientific discovery.

cs.LG↗
arXiv2023

ChimpACT: A Longitudinal Dataset for Understanding Chimpanzee Behaviors

Understanding the behavior of non-human primates is crucial for improving animal welfare, modeling social behavior, and gaining insights into distinctively human and phylogenetically shared behaviors. However, the lack of datasets on non-human primate behavior hinders in-depth exploration of primate social interactions, posing challenges to research on our closest living relatives. To address these limitations, we present ChimpACT, a comprehensive dataset for quantifying the longitudinal behavior and social relations of chimpanzees within a social group. Spanning from 2015 to 2018, ChimpACT features videos of a group of over 20 chimpanzees residing at the Leipzig Zoo, Germany, with a particular focus on documenting the developmental trajectory of one young male, Azibo. ChimpACT is both comprehensive and challenging, consisting of 163 videos with a cumulative 160,500 frames, each richly annotated with detection, identification, pose estimation, and fine-grained spatiotemporal behavior labels. We benchmark representative methods of three tracks on ChimpACT: (i) tracking and identification, (ii) pose estimation, and (iii) spatiotemporal action detection of the chimpanzees. Our experiments reveal that ChimpACT offers ample opportunities for both devising new methods and adapting existing ones to solve fundamental computer vision tasks applied to chimpanzee groups, such as detection, pose estimation, and behavior analysis, ultimately deepening our comprehension of communication and sociality in non-human primates.

cs.CV↗
arXiv2023

Tailor: Altering Skip Connections for Resource-Efficient Inference

Deep neural networks use skip connections to improve training convergence. However, these skip connections are costly in hardware, requiring extra buffers and increasing on- and off-chip memory utilization and bandwidth requirements. In this paper, we show that skip connections can be optimized for hardware when tackled with a hardware-software codesign approach. We argue that while a network's skip connections are needed for the network to learn, they can later be removed or shortened to provide a more hardware efficient implementation with minimal to no accuracy loss. We introduce Tailor, a codesign tool whose hardware-aware training algorithm gradually removes or shortens a fully trained network's skip connections to lower their hardware cost. Tailor improves resource utilization by up to 34% for BRAMs, 13% for FFs, and 16% for LUTs for on-chip, dataflow-style architectures. Tailor increases performance by 30% and reduces memory bandwidth by 45% for a 2D processing element array architecture.

cs.CV↗
arXiv2023

Information Flow Coverage Metrics for Hardware Security Verification

Security graphs model attacks, defenses, mitigations, and vulnerabilities on computer networks and systems. With proper attributes, they provide security metrics using standard graph algorithms. A hyperflow graph is a register-transfer level (RTL) hardware security graph that facilitates security verification. A hyperflow graph models information flows and is annotated with attributes that allow security metrics to measure flow paths, flow conditions, and flow rates. Hyperflow graphs enable the understanding of hardware vulnerabilities related to confidentiality, integrity, and availability, as shown on the OpenTitan hardware root of trust under several threat models.

cs.CR↗
arXiv2023

Pentimento: Data Remanence in Cloud FPGAs

Cloud FPGAs strike an alluring balance between computational efficiency, energy efficiency, and cost. It is the flexibility of the FPGA architecture that enables these benefits, but that very same flexibility that exposes new security vulnerabilities. We show that a remote attacker can recover "FPGA pentimenti" - long-removed secret data belonging to a prior user of a cloud FPGA. The sensitive data constituting an FPGA pentimento is an analog imprint from bias temperature instability (BTI) effects on the underlying transistors. We demonstrate how this slight degradation can be measured using a time-to-digital (TDC) converter when an adversary programs one into the target cloud FPGA. This technique allows an attacker to ascertain previously safe information on cloud FPGAs, even after it is no longer explicitly present. Notably, it can allow an attacker who knows a non-secret "skeleton" (the physical structure, but not the contents) of the victim's design to (1) extract proprietary details from an encrypted FPGA design image available on the AWS marketplace and (2) recover data loaded at runtime by a previous user of a cloud FPGA using a known design. Our experiments show that BTI degradation (burn-in) and recovery are measurable and constitute a security threat to commercial cloud FPGAs.

cs.CR↗
arXiv2022

Open-source FPGA-ML codesign for the MLPerf Tiny Benchmark

We present our development experience and recent results for the MLPerf Tiny Inference Benchmark on field-programmable gate array (FPGA) platforms. We use the open-source hls4ml and FINN workflows, which aim to democratize AI-hardware codesign of optimized neural networks on FPGAs. We present the design and implementation process for the keyword spotting, anomaly detection, and image classification benchmark tasks. The resulting hardware implementations are quantized, configurable, spatial dataflow architectures tailored for speed and efficiency and introduce new generic optimizations and common workflows developed as a part of this work. The full workflow is presented from quantization-aware training to FPGA implementation. The solutions are deployed on system-on-chip (Pynq-Z2) and pure FPGA (Arty A7-100T) platforms. The resulting submissions achieve latencies as low as 20 $μ$s and energy consumption as low as 30 $μ$J per inference. We demonstrate how emerging ML benchmarks on heterogeneous hardware platforms can catalyze collaboration and the development of new techniques and more accessible tools.

cs.LG↗
arXiv2021

Applications and Techniques for Fast Machine Learning in Science

In this community review report, we discuss applications and techniques for fast machine learning (ML) in science -- the concept of integrating power ML methods into the real-time experimental data processing loop to accelerate scientific discovery. The material for the report builds on two workshops held by the Fast ML for Science community and covers three main areas: applications for fast ML across a number of scientific domains; techniques for training and implementing performant and resource-efficient ML algorithms; and computing architectures, platforms, and technologies for deploying these algorithms. We also present overlapping challenges across the multiple scientific domains where common solutions can be found. This community report is intended to give plenty of examples and inspiration for scientific discovery through integrated and accelerated ML solutions. This is followed by a high-level overview and organization of technical advances, including an abundance of pointers to source material, which can enable these breakthroughs.

cs.LG↗
arXiv2021

Isadora: Automated Information Flow Property Generation for Hardware Designs

Isadora is a methodology for creating information flow specifications of hardware designs. The methodology combines information flow tracking and specification mining to produce a set of information flow properties that are suitable for use during the security validation process, and which support a better understanding of the security posture of the design. Isadora is fully automated; the user provides only the design under consideration and a testbench and need not supply a threat model nor security specifications. We evaluate Isadora on a RISC-V processor plus two designs related to SoC access control. Isadora generates security properties that align with those suggested by the Common Weakness Enumerations (CWEs), and in the case of the SoC designs, align with the properties written manually by security experts.

cs.CR↗
arXiv2021

AKER: A Design and Verification Framework for Safe andSecure SoC Access Control

Modern systems on a chip (SoCs) utilize heterogeneous architectures where multiple IP cores have concurrent access to on-chip shared resources. In security-critical applications, IP cores have different privilege levels for accessing shared resources, which must be regulated by an access control system. AKER is a design and verification framework for SoC access control. AKER builds upon the Access Control Wrapper (ACW) -- a high performance and easy-to-integrate hardware module that dynamically manages access to shared resources. To build an SoC access control system, AKER distributes the ACWs throughout the SoC, wrapping controller IP cores, and configuring the ACWs to perform local access control. To ensure the access control system is functioning correctly and securely, AKER provides a property-driven security verification using MITRE common weakness enumerations. AKER verifies the SoC access control at the IP level to ensure the absence of bugs in the functionalities of the ACW module, at the firmware level to confirm the secure operation of the ACW when integrated with a hardware root-of-trust (HRoT), and at the system level to evaluate security threats due to the interactions among shared resources. The performance, resource usage, and security of access control systems implemented through AKER is experimentally evaluated on a Xilinx UltraScale+ programmable SoC, it is integrated with the OpenTitan hardware root-of-trust, and it is used to design an access control system for the OpenPULP multicore architecture.

cs.CR↗
arXiv2019

In Pursuit of Galactic Archaeology: Astro2020 Science White Paper

The next decade affords tremendous opportunity to achieve the goals of Galactic archaeology. That is, to reconstruct the evolutionary narrative of the Milky Way, based on the empirical data that describes its current morphological, dynamical, temporal and chemical structures. Here, we describe a path to achieving this goal. The critical observational objective is a Galaxy-scale, contiguous, comprehensive mapping of the disk's phase space, tracing where the majority of the stellar mass resides. An ensemble of recent, ongoing, and imminent surveys are working to deliver such a transformative stellar map. Once this empirical description of the dust-obscured disk is assembled, we will no longer be operationally limited by the observational data. The primary and significant challenge within stellar astronomy and Galactic archaeology will then be in fully utilizing these data. We outline the next-decade framework for obtaining and then realizing the potential of the data to chart the Galactic disk via its stars. One way to support the investment in the massive data assemblage will be to establish a Galactic Archaeology Consortium across the ensemble of stellar missions. This would reflect a long-term commitment to build and support a network of personnel in a dedicated effort to aggregate, engineer, and transform stellar measurements into a comprehensive perspective of our Galaxy.

astro-ph.GA↗
arXiv2017

The central spheroids of Milky Way mass-sized galaxies

We study the properties of the central spheroids located within 10 kpc of the centre of mass of Milky Way mass-sized galaxies simulated in a cosmological context. The simulated central regions are dominated by stars older than 10 Gyr, mostly formed in situ, with a contribution of ~30 per cent from accreted stars. These stars formed in well-defined starbursts, although accreted stars exhibit sharper and earlier ones. The fraction of accreted stars increases with galactocentric distance, so that at a radius of ~8-10 kpc a fraction of ~40 per cent, on average, are detected. Accreted stars are slightly younger, lower metallicity, and more $α$-enhanced than in situ stars. A significant fraction of old stars in the central regions come from a few ($2-3$) massive satellites ($\sim 10^{10}{\rm M}_\odot$). The bulge components receive larger contributions of accreted stars formed in dwarfs smaller than $\sim 10^{9.5}{\rm M}_\odot$. The difference between the distributions of ages and metallicities of old stars is thus linked to the accretion histories -- those central regions with a larger fraction of accreted stars are those with contributions from more massive satellites. The kinematical properties of in situ and accreted stars are consistent with the latter being supported by their velocity dispersions, while the former exhibit clear signatures of rotational support. Our simulations demonstrate a range of characteristics, with some systems exhibiting a co-existing bar and spheroid in their central regions, resembling in some respect the central region of the Milky Way.

astro-ph.GA↗
arXiv2016

Two Groups of Red Giants with Distinct Chemical Abundances in the Bulge Globular Cluster NGC 6553 Through the Eyes of APOGEE

Multiple populations revealed in globular clusters (GCs) are important windows to the formation and evolution of these stellar systems. The metal-rich GCs in the Galactic bulge are an indispensable part of this picture, but the high optical extinction in this region has prevented extensive research. In this work, we use the high resolution near-infrared (NIR) spectroscopic data from APOGEE to study the chemical abundances of NGC 6553, which is one of the most metal-rich bulge GCs. We identify ten red giants as cluster members using their positions, radial velocities, iron abundances, and NIR photometry. Our sample stars show a mean radial velocity of $-0.14\pm5.47$ km s$^{-1}$, and a mean [Fe/H] of $-0.15\pm 0.05$. We clearly separate two populations of stars in C and N in this GC for the first time. NGC 6553 is the most metal-rich GC where the multiple stellar population phenomenon is found until now. Substantial chemical variations are also found in Na, O, and Al. However, the two populations show similar Si, Ca, and iron-peak element abundances. Therefore, we infer that the CNO, NeNa, and MgAl cycles have been activated, but the MgAl cycle is too weak to show its effect on Mg. Type Ia and Type II supernovae do not seem to have significantly polluted the second generation stars. Comparing with other GC studies, NGC 6553 shows similar chemical variations as other relatively metal-rich GCs. We also confront current GC formation theories with our results, and suggest possible avenues for improvement in the models.

astro-ph.GA↗
arXiv2016

The SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Overview and Early Data

The Extended Baryon Oscillation Spectroscopic Survey (eBOSS) will conduct novel cosmological observations using the BOSS spectrograph at Apache Point Observatory. Observations will be simultaneous with the Time Domain Spectroscopic Survey (TDSS) designed for variability studies and the Spectroscopic Identification of eROSITA Sources (SPIDERS) program designed for studies of X-ray sources. eBOSS will use four different tracers to measure the distance-redshift relation with baryon acoustic oscillations (BAO). Using more than 250,000 new, spectroscopically confirmed luminous red galaxies at a median redshift z=0.72, we project that eBOSS will yield measurements of $d_A(z)$ to an accuracy of 1.2% and measurements of H(z) to 2.1% when combined with the z>0.6 sample of BOSS galaxies. With ~195,000 new emission line galaxy redshifts, we expect BAO measurements of $d_A(z)$ to an accuracy of 3.1% and H(z) to 4.7% at an effective redshift of z= 0.87. A sample of more than 500,000 spectroscopically-confirmed quasars will provide the first BAO distance measurements over the redshift range 0.9 2.1; these new data will enhance the precision of $d_A(z)$ and H(z) by a factor of 1.44 relative to BOSS. Furthermore, eBOSS will provide improved tests of General Relativity on cosmological scales through redshift-space distortion measurements, improved tests for non-Gaussianity in the primordial density field, and new constraints on the summed mass of all neutrino species. Here, we provide an overview of the cosmological goals, spectroscopic target sample, demonstration of spectral quality from early data, and projected cosmological constraints from eBOSS.

astro-ph.CO↗
arXiv2009

Future Evolution of Bound Superclusters in an Accelerating Universe

The evolution of marginally bound supercluster-like objects in an accelerating LambdaCDM Universe is followed, by means of cosmological simulations, from the present time to an expansion factor a = 100. The objects are identified on the basis of the binding density criterion introduced by Dunner et al. (2006). superclusters are identified with the ones whose mass M > 10^15 M_sun/h, the most massive one with M ~ 8x10^15 M_sun/h, comparable to the Shapley supercluster. The spatial distribution of the superclusters remains essentially the same after the present epoch, reflecting the halting growth of the Cosmic Web as Lambda gets to dominate the expansion of the Universe. The same trend can be seen in the stagnation of the development of the mass function of virialized haloes and bound objects. The situation is considerably different when looking at the internal evolution, quantified in terms of their shape, compactness and density profile, and substructure in terms of their multiplicity function. We find a continuing evolution from a wide range of triaxial shapes at a = 1 to almost perfect spherical shapes at a = 100. We also find a systematic trend towards a higher concentration. Meanwhile, we see their substructure gradually disappearing, as the surrounding subclumps fall in and merge to form one coherent, virialized system.

astro-ph↗
arXiv2005

Accretion relicts in the solar neighbourhood: debris from omegaCen's parent galaxy

We use numerical simulations to investigate the orbital characteristics of tidal debris from satellites whose orbits are dragged into the plane of galactic disks by dynamical friction before disruption. We find that these satellites may deposit a significant fraction of their stars into the disk components of a galaxy, and use our results to motivate the search for accretion relicts in samples of metal-poor disk stars in the vicinity of the Sun. Satellites disrupted on very eccentric orbits coplanar with the disk are expected to shed stars in "trails" of distinct orbital energy and angular momentum during each pericentric passage. To an observer located between the pericenter and apocenter of such orbits, these trails would show as distinct groupings of stars with low vertical velocity and a broad, symmetric, often double-peaked distribution of Galactocentric radial velocities. One group of stars with these characteristics stands out in available compilations of nearby metal-poor stars. These stars have specific angular momenta similar to that of the globular cluster omegaCen, long hypothesized to be the nucleus of a dwarf galaxy disrupted by the Milky Way tidal field. In addition to their kindred kinematics, stars in the omegaCen group share distinct chemical abundance characteristics, and trace a well-defined track in the [alpha/Fe] versus [Fe/H] plane, consistent with simple closed-box enrichment models and a protracted star formation history. The dynamical and chemical coherence of this group suggests that it consists of stars that once belonged to the dwarf that brought omegaCen into the Galaxy. The presence of this and other "tidal relicts" in the solar neighbourhood suggest an extra-Galactic origin for the presence of nearby stars with odd kinematics and chemistry.

astro-ph↗
arXiv2003

Simulations of Galaxy Formation in a Lambda CDM Universe III: The Dissipative Formation of an Elliptical Galaxy

We examine in detail the dynamical structure of an elliptical galaxy simulated in the Lambda CDM scenario. The morphology of the galaxy evolves dramatically over time in response to the mode and timing of mass accretion; smooth deposition of cooled gas leads to the formation of centrifugally supported disks, whilst major mergers disperse stellar disks into spheroids. These two modes of accretion alternate successively until z~0.6, when the galaxy undergoes one last major (1:2) merger that consumes much of the remaining gas into stars. Little gas cools and accretes subsequently and, as a result, most stars at z=0 are in a spheroidal component that resembles present-day elliptical galaxies. Dynamically, the galaxy is well approximated by an E4 oblate rotator. Boxy isophotes are obtained when the galaxy is seen face-on and Vrot/sigma<<1. On the other hand, disky isophotes are found for inclinations which maximize Vrot/sigma. The line-of-sight velocity distribution is nearly Gaussian at all radii. The sign of the Gauss-Hermite skewness parameter h3 is anti-correlated with the apparent sense of rotation, in agreement with observed trends. The simulated galaxy has much higher effective surface brightness than normal ellipticals of similar luminosity, in a way reminiscent of the less common type of compact ellipticals. Our simulation shows that repeated episodes of dissipational collapse, followed by merger events, lead to stellar spheroids that are only mildly triaxial and of relatively simple kinematic structure. This is in better agreement with observation than earlier models based on dissipationless mergers of stellar disks, and a positive step towards reconciling the observed structure of ellipticals with the hierarchical merging scenario.

astro-ph↗