arXiv subjects

James Garland

Publications and source records attributed to James Garland.

8 recordsLinked to original sources

arXiv2023

Introducing the Condor Array Telescope. II. Deep imaging observations of the edge-on spiral galaxy NGC 5907 and the NGC 5866 Group: yet another view of the iconic stellar stream

We used the Condor Array Telescope to obtain deep imaging observations through the luminance filter of the entirety of the NGC 5866 Group, including a very extended region surrounding the galaxy NGC 5907 and its stellar stream. We find that the stellar stream consists of a single curved structure that stretches $220$ kpc from a brighter eastern stream to a fainter western stream that bends to the north and then curls back toward the galaxy. This result runs contrary to a previous claim of a second loop of the stellar stream but is consistent with another previous description of the overall morphology of the stream. We further find that: (1) an extension of the western stream appears to bifurcate near its apex, (2) there is an apparent gap of $\approx 6$ kpc in the western stream due east of the galaxy, (3) contrary to a previous claim, there is no evidence of the remnant of a progenitor galaxy within the eastern stream, although (4) there are many other possible progenitor galaxies, (5) there is another structure that, if it is at the distance of the galaxy, stretches 240 kpc and contains two very large, very low-surface-brightness "patches" of emission, one of which was noted previously and another of which was not. We note the number and variety of stellar streams in the vicinity of NGC 5907 and the apparent gap in the western stream, which may be indicative of a dark subhalo or satellite in the vicinity of the galaxy.

astro-ph.GA↗

arXiv2023

HI Rich but Low Star Formation galaxies in MaNGA: Physical Properties and Comparison to Control Samples

Gas rich galaxies are typically star-forming. We make use of HI-MaNGA, a program of HI follow-up for the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey of the Sloan Digital Sky Surveys to construct a sample of unusual neutral hydrogen (HI, 21cm) rich galaxies which have low Star Formation Rates (SFRs); using infra-red color from the Wide-field Infrared Survey Explorer (WISE) as a proxy for specific SFR. Out of a set of 1575 MaNGA galaxies with HI-MaNGA detections, we find 83 (5%) meet our selection criteria to be HI rich with low SFR. We construct two stellar mass-matched control samples: HI rich galaxies with typical SFR (High SF Control) and HI poor galaxies with low SFR (Low HI Control). We investigate the properties of each of these samples, comparing physical parameters such as ionization state maps, stellar and ionized gas velocity and dispersion, environment measures, metallicity, and morphology to search for the reasons why these unusual HI rich galaxies are not forming stars. We find evidence for recent external accretion of gas in some galaxies (via high counter-rotating fractions), along with some evidence for AGN feedback (from a high cLIER and/or red geyser fraction), and bar quenching (via an enhanced strong bar fraction). Some galaxies in the sample are consistent with simply having their HI in a high angular momentum, large radius, low density disc. We conclude that no single physical process can explain all HI rich, low SFR galaxies.

astro-ph.GA↗

arXiv2023

Feature Representation in Deep Metric Embeddings

In deep metric learning (DML), high-level input data are represented in a lower-level representation (embedding) space, such that samples from the same class are mapped close together, while samples from disparate classes are mapped further apart. In this lower-level representation, only a single inference sample from each known class is required to discriminate between classes accurately. The features a DML model uses to discriminate between classes and the importance of each feature in the training process are unknown. To investigate this, this study takes embeddings trained to discriminate faces (identities) and uses unsupervised clustering to identify the features involved in facial identity discrimination by examining their representation within the embedded space. This study is split into two cases; intra class sub-discrimination, where attributes that differ between a single identity are considered; such as beards and emotions; and extra class sub-discrimination, where attributes which differ between different identities/people, are considered; such as gender, skin tone and age. In the intra class scenario, the inference process distinguishes common attributes between single identities, achieving 90.0\% and 76.0\% accuracy for beards and glasses, respectively. The system can also perform extra class sub-discrimination with a high accuracy rate, notably 99.3\%, 99.3\% and 94.1\% for gender, skin tone, and age, respectively.

cs.CV↗

arXiv2021

HOBFLOPS CNNs: Hardware Optimized Bitslice-Parallel Floating-Point Operations for Convolutional Neural Networks

Convolutional neural networks (CNNs) are typically trained using 16- or 32-bit floating-point (FP) and researchers show that low-precision floating-point (FP) can be highly effective for inference. Low-precision FP can be implemented in field programmable gate array (FPGA) and application-specific integrated circuit (ASIC) accelerators, but existing processors do not generally support custom precision FP. We propose hardware optimized bitslice-parallel floating-point operators (HOBFLOPS), a method of generating efficient custom-precision emulated bitslice-parallel software FP arithmetic. We generate custom-precision FP routines optimized using a hardware synthesis design flow to create circuits. We provide standard cell libraries matching the bitwise operations on the target microprocessor architecture, and a code-generator to translate the hardware circuits to bitslice software equivalents. We exploit bitslice parallelism to create a very wide (32-512 element) vectorized convolutional neural network (CNN) convolution. Hardware optimized bitslice-parallel floating-point operators (HOBFLOPS) multiply-accumulate (MAC) performance in CNN convolution on Arm and Intel processors are compared to Berkeley's SoftFP16 equivalent MAC. HOBFLOPS16 outperforms SoftFP16 by 8x on Intel AVX512. HOBFLOPS offers arbitrary-precision FP with custom range and precision e.g., HOBFLOPS9 performs at 6x the performance of HOBFLOPS16 on Arm Neon. HOBFLOPS allows researchers to prototype different levels of custom FP precision in the arithmetic of software CNN accelerators. Furthermore, HOBFLOPS fast custom-precision FP CNNs may be valuable in cases where memory bandwidth is limited.

cs.AR↗

arXiv2021

Low precision logarithmic number systems: Beyond base-2

Logarithmic number systems (LNS) are used to represent real numbers in many applications using a constant base raised to a fixed-point exponent making its distribution exponential. This greatly simplifies hardware multiply, divide and square root. LNS with base-2 is most common, but in this paper we show that for low-precision LNS the choice of base has a significant impact. We make four main contributions. First, LNS is not closed under addition and subtraction, so the result is approximate. We show that choosing a suitable base can manipulate the distribution to reduce the average error. Second, we show that low-precision LNS addition and subtraction can be implemented efficiently in logic rather than commonly used ROM lookup tables, the complexity of which can be reduced by an appropriate choice of base. A similar effect is shown where the result of arithmetic has greater precision than the input. Third, where input data from external sources is not expected to be in LNS, we can reduce the conversion error by selecting a LNS base to match the expected distribution of the input. Thus, there is no one base which gives the global optimum, and base selection is a trade-off between different factors. Fourth, we show that circuits realized in LNS require lower area and power consumption for short word lengths.

math.NA↗

arXiv2018

Low Complexity Multiply-Accumulate Units for Convolutional Neural Networks with Weight-Sharing

Convolutional neural networks (CNNs) are one of the most successful machine learning techniques for image, voice and video processing. CNNs require large amounts of processing capacity and memory bandwidth. Hardware accelerators have been proposed for CNNs which typically contain large numbers of multiply-accumulate (MAC) units, the multipliers of which are large in an integrated circuit (IC) gate count and power consumption. "Weight sharing" accelerators have been proposed where the full range of weight values in a trained CNN are compressed and put into bins and the bin index used to access the weight-shared value. We reduce power and area of the CNN by implementing parallel accumulate shared MAC (PASM) in a weight shared CNN. PASM re-architects the MAC to instead count the frequency of each weight and place it in a bin. The accumulated value is computed in a subsequent multiply phase, significantly reducing gate count and power consumption of the CNN. In this paper, we implement PASM in a weight-shared CNN convolution hardware accelerator and analyze its effectiveness. Experiments show that for a clock speed 1GHz implemented on a 45nm ASIC process our approach results in fewer gates, smaller logic, and reduced power with only a slight increase in latency. We also show that the same weight-shared-with-PASM CNN accelerator can be implemented in resource-constrained FPGAs, where the FPGA has limited numbers of digital signal processor (DSP) units to accelerate the MAC operations.

cs.AR↗

arXiv2017

Small-N collisional dynamics III: The battle for the realm of not-so-small-N

In this paper, the third in the series, we continue our study of combinatorics in chaotic Newtonian dynamics. We study the chaotic four-body problem in Newtonian gravity assuming finite-sized particles, and we focus on interactions that produce direct collisions between any two stars. Our long-term goal is to construct an equation that gives the probability of a given collision event occurring over the course of the interaction, as a function of the total encounter energy and angular momentum as well as the numbers and properties of the particles. In previous papers, we varied the number of interacting particles and the distribution of particle radii, for all equal mass particles. Here, we focus on the effects of different combinations of particle masses. We develop an analytic formalism for calculating the time-scales for different collision scenarios to occur. Our analytic time-scales reproduce the simulated time-scales when gravitational focusing is included. We present a method for calculating the relative rates for different types of collisions to occur, assuming two different limits for the particle orbits; radial and tangential. These limits yield relative collision probabilities that bracket the probabilities we obtain directly from numerical scattering experiments, and are designed to reveal important information about the (time-averaged) trajectories of the particles as a function of the interaction parameters. Finally, we present a Collision Rate Diagram (CRD), which directly compares the predictions of our analytic rates to the simulations and quantifies the quality of the agreement. The CRD will facilitate refining our analytic collision rates in future work, as we expand in to the remaining parameter space.

astro-ph.SR↗

arXiv2017

Low Complexity Multiply Accumulate Unit for Weight-Sharing Convolutional Neural Networks

Convolutional Neural Networks (CNNs) are one of the most successful deep machine learning technologies for processing image, voice and video data. CNNs require large amounts of processing capacity and memory, which can exceed the resources of low power mobile and embedded systems. Several designs for hardware accelerators have been proposed for CNNs which typically contain large numbers of Multiply Accumulate (MAC) units. One approach to reducing data sizes and memory traffic in CNN accelerators is "weight sharing", where the full range of values in a trained CNN are put in bins and the bin index is stored instead of the original weight value. In this paper we propose a novel MAC circuit that exploits binning in weight-sharing CNNs. Rather than computing the MAC directly we instead count the frequency of each weight and place it in a bin. We then compute the accumulated value in a subsequent multiply phase. This allows hardware multipliers in the MAC circuit to be replaced with adders and selection logic. Experiments show that for the same clock speed our approach results in fewer gates, smaller logic, and reduced power.

cs.NE↗