Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Bayesian Inference”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,351 records · Page 75

SYSTEMATIC UNCERTAINTIES IN THE SPECTROSCOPIC MEASUREMENTS OF NEUTRON-STAR MASSES AND RADII FROM THERMONUCLEAR X-RAY BURSTS. I. APPARENT RADII

The masses and radii of low-magnetic field neutron stars can be measured by combining different observable quantities obtained from their X-ray spectra during thermonuclear X-ray bursts. One of these quantities is the apparent radius of each neutron star as inferred from the X-ray flux and spectral temperature measured during the cooling tails of bursts, when the thermonuclear flash is believed to have engulfed the entire star. In this paper, we analyze 13,095 X-ray spectra of 446 X-ray bursts observed from 12 sources in order to assess possible systematic effects in the measurements of the apparent radii of neutron stars. We first show that the vast majority of the observed X-ray spectra are consistent with blackbody functions to within a few percent. We find that most X-ray bursts follow a very well determined correlation between X-ray flux and temperature, which is consistent with the whole neutron-star surface emitting uniformly during the cooling tails. We develop a Bayesian Gaussian mixture algorithm to measure the apparent radii of the neutron stars in these sources, while detecting and excluding a small number of X-ray bursts that show irregular cooling behavior. This algorithm also provides us with a quantitative measure of the systematic uncertainties in the measurements. We find that those errors in the spectroscopic determination of neutron-star radii that are introduced by systematic effects in the cooling tails of X-ray bursts are in the range {approx_equal} 3%-8%. Such small errors are adequate to distinguish between different equations of state provided that uncertainties in the distance to each source and the absolute calibration of X-ray detectors do not dominate the error budget.

79 ASTRONOMY AND ASTROPHYSICS↗

Modeling the Swift BAT trigger algorithm with machine learning

To draw inferences about gamma-ray burst (GRB) source populations based on Swift observations, it is essential to understand the detection efficiency of the Swift burst alert telescope (BAT). This study considers the problem of modeling the Swift/BAT triggering algorithm for long GRBs, a computationally expensive procedure, and models it using machine learning algorithms. A large sample of simulated GRBs from Lien et al. is used to train various models: random forests, boosted decision trees (with AdaBoost), support vector machines, and artificial neural networks. The best models have accuracies of ≳97% (≲3% error), which is a significant improvement on a cut in GRB flux, which has an accuracy of 89.6% (10.4% error). These models are then used to measure the detection efficiency of Swift as a function of redshift z, which is used to perform Bayesian parameter estimation on the GRB rate distribution. We find a local GRB rate density of n{sub 0}∼0.48{sub −0.23}{sup +0.41} Gpc{sup −3} yr{sup −1} with power-law indices of n{sub 1}∼1.7{sub −0.5}{sup +0.6} and n{sub 2}∼−5.9{sub −0.1}{sup +5.7} for GRBs above and below a break point of z{sub 1}∼6.8{sub −3.2}{sup +2.8}. This methodology is able to improve upon earlier studies by more accurately modeling Swift detection and using this for fully Bayesian model fitting.

79 ASTRONOMY AND ASTROPHYSICS↗

In Situ Inference for Earth System Predictability

Focal Area: Focal Area 3: Insight gleaned from complex simulated data using AI, big data analytics, and other advanced methods, including explainable AI and physics- or knowledge-guided AI. Science Challenge: An understanding of future evolution in precipitation extremes is critical to numerous DOE mission questions. Extreme events are by nature short time-scale events that are difficult to diagnose in available model data. Accurate modeling of extreme events necessarily requires high spatial resolution at the storm scale locally. However, the environment in which storms grow is dependent on global, remote, processes. These complex spatiotemporal relationships are impossible to diagnose at resolutions required to accurately model storms responsible for extreme precipitation. At exascale, climate simulations will produce results at fine enough resolution to investigate these relationships. However, the resulting data from these simulations will be far too large to save for post-simulation analysis. We advocate for fitting statistical models inside the simulations as they run, a context known as in situ, which will facilitate scientific investigations using the full fine-scale data stream. Figure 1 shows an example of the type of model we could consider, a Bayesian hierarchical spatial regression model. Precipitation extremes at each grid cell are modeled using extreme value distributions. Since extremes are rare, fitting models to individual grid cells can result in high variance and poor estimates. Instead, the model can be made more robust by smoothing the parameters of the extreme value model across space. Additionally, the parameters themselves can be functionally linked to other variables elsewhere in the simulation. Thus, we can use the fine-scale data to build more robust models for extremes that link extreme behavior to other climate patterns.

54 ENVIRONMENTAL SCIENCES↗

Do LIGO/Virgo Black Hole Mergers Produce AGN Flares? The Case of GW190521 and Prospects for Reaching a Confident Association

The recent report of an association of the gravitational-wave (GW) binary black hole (BBH) merger GW190521 with a flare in the active galactic nuclei (AGNs) J124942.3 + 344929 has generated tremendous excitement. However, GW190521 has one of the largest localization volumes among all of the GW events detected so far. The 90% localization volume likely contains 7400 unobscured AGNs brighter than g ≤ 20.5 AB mag, and it results in a ≳70% probability of chance coincidence for an AGN flare consistent with the GW event. We present a Bayesian formalism to estimate the confidence of an AGN association by analyzing a population of BBH events with dedicated follow-up observations. Depending on the fraction of BBHs arising from AGNs, counterpart searches of $\mathcal{O}(1) - \mathcal{O}(100)$ GW events are needed to establish a confident association, and more than an order of magnitude more for searches without follow-up (i.e., using only the locations of AGN and GW events). Follow-up campaigns of the top ~5% (based on volume localization and binary mass) of BBH events with total rest-frame mass ≥50 M ⊙ are expected to establish a confident association during the next LIGO/Virgo/KAGRA observing run (O4), as long as the true value of the fraction of BBHs giving rise to AGN flares is >0.1. Our formalism allows us to jointly infer cosmological parameters from a sample of BBH events that include chance coincidence flares. Until the confidence of AGN associations is established, the probability of chance coincidence must be taken into account to avoid biasing astrophysical and cosmological constraints.

79 ASTRONOMY AND ASTROPHYSICS↗

Galaxy cluster matter profiles - I. Self-similarity, mass calibration, and observable-mass relation validation employing cluster mass posteriors

We present a study of the weak lensing inferred matter profiles ΔΣ(R) of 698 South Pole Telescope (SPT) thermal Sunyaev-Zel’dovich effect (tSZE) selected and MCMF optically confirmed galaxy clusters in the redshift range 0.25 < z < 0.94 that have associated weak gravitational lensing shear profiles from the Dark Energy Survey (DES). Rescaling these profiles to account for the mass dependent size and the redshift dependent density produces average rescaled matter profiles ΔΣ(R/R200c)/(ρcritR200c) with a lower dispersion than the unscaled ΔΣ(R) versions, indicating a significant degree of self-similarity. Galaxy clusters from hydrodynamical simulations also exhibit matter profiles that suggest a high degree of self-similarity, with RMS variation among the average rescaled matter profiles with redshift and mass falling by a factor of approximately six and 23, respectively, compared to the unscaled average matter profiles. We employed this regularity in a new Bayesian method for weak lensing mass calibration that employs the so-called cluster mass posterior P(M200|ζ̂, λ̂, z), which describes the individual cluster masses given their tSZE (ζ̂) and optical (λ̂, z) observables. This method enables simultaneous constraints on richness λ-mass and tSZE detection significance ζ-mass relations using average rescaled cluster matter profiles. We validated the method using realistic mock datasets and present observable-mass relation constraints for the SPT×DES sample, where we constrained the amplitude, mass trend, redshift trend, and intrinsic scatter. Our observable-mass relation results are in agreement with the mass calibration derived from the recent cosmological analysis of the SPT×DES data based on a cluster-by-cluster lensing calibration. Our new mass calibration technique offers a higher efficiency when compared to the single cluster calibration technique. We present new validation tests of the observable-mass relation that indicate the underlying power-law form and scatter are adequate to describe the real cluster sample but that also suggest a redshift variation in the intrinsic scatter of the λ-mass relation may offer a better description. In addition, the average rescaled matter profiles offer high signal-to-noise ratio (S/N) constraints on the shape of real cluster matter profiles, which are in good agreement with available hydrodynamical ΛCDM simulations. This high S/N profile contains information about baryon feedback, the collisional nature of dark matter, and potential deviations from general relativity.Key words: gravitational lensing: weak / galaxies: clusters: general / large-scale structure of Universe

79 ASTRONOMY AND ASTROPHYSICS↗

A dendritic strontium river isoscape for fisheries applications in the Sacramento River basin, California, USA

Objective Understanding the origins and movements of fish is fundamental to effective conservation and fisheries management. Strontium isotope ratios ( 87 Sr/ 86 Sr) in otoliths provide a powerful tracer of natal origin and migratory pathways. However, existing 87 Sr/ 86 Sr isoscapes for the Sacramento River basin, an ecosystem that supports ecologically and economically important salmon populations, rely on discrete classification approaches that overlook unsampled habitats and do not incorporate spatial uncertainty. Our objective was to develop a continuous, network-explicit 87 Sr/ 86 Sr isoscape with quantified uncertainty to fill in data gaps and enable probabilistic assignments of fish origin and movement. Methods We used river water 87 Sr/ 86 Sr data from 106 sites (1997–2021) to develop spatial stream network models that use dendritic connectivity and watershed characteristics (lithology, bedrock age, and land cover) to predict river water 87 Sr/ 86 Sr throughout the basin. Models were fitted using maximum and restricted likelihood and were evaluated via Akaike’s information criterion and leave-one-out cross validation. We produced both historical (pre-dam) and present-day (below-dam) isoscapes, delineated uncertainty-informed isotopic ranges using k -means clustering, and applied a proof-of-concept Bayesian assignment to estimate natal origins and early rearing habitats for two endangered winter-run Chinook Salmon Oncorhynchus tshawytscha. Results Cross validation indicated strong performance of the 87 Sr/ 86 Sr model (leave-one-out cross validation: R 2 = 0.91; root mean square error = 0.0005). Uncertainty-informed clustering identified 19 isotopic “suites” (reaches with indistinguishable 87 Sr/ 86 Sr values) in present-day anadromous habitats and 25 suites in the historical network. Example natal and early rearing assignments included predictions that challenged expectations for juvenile salmon migration based on predicted river 87 Sr/ 86 Sr compositions. Conclusions This study developed a continuous, network-explicit 87 Sr/ 86 Sr isoscape that integrates existing river data to predict 87 Sr/ 86 Sr in unsampled reaches and the likely achievable range and resolution of otolith-based origin and life history inference. The resulting river isoscape provides a valuable tool to predict salmon movements and identify habitats supporting their survival and growth that otherwise might remain undetected. Coupling these predictions with complementary approaches that ground-truth juvenile presence (e.g., targeted fish surveys) represents an important step toward science-informed restoration and management of critical habitats throughout the Sacramento River basin.

Environmental sciences↗

(Lack of) Cosmological evidence for dark radiation after Planck

We use Bayesian model comparison to determine whether extensions to Standard-Model neutrino physics — primarily additional effective numbers of neutrinos and/or massive neutrinos — are merited by the latest cosmological data. Given the significant advances in cosmic microwave background (CMB) observations represented by the Planck data, we examine whether Planck temperature and CMB lensing data, in combination with lower redshift data, have strengthened (or weakened) the previous findings. We conclude that the state-of-the-art cosmological data do not show evidence for deviations from the standard (ΛCDM) cosmological model (which has three massless neutrino families). This does not mean that the model is necessarily correct — in fact we know it is incomplete as neutrinos are not massless — but it does imply that deviations from the standard model (e.g., non-zero neutrino mass) are too small compared to the current experimental uncertainties to be inferred from cosmological data alone.

79 ASTRONOMY AND ASTROPHYSICS↗

Modeling the Swift BAT Trigger Algorithm with Machine Learning

To draw inferences about gamma-ray burst (GRB) source populations based on Swift observations, it is essential to understand the detection efficiency of the Swift burst alert telescope (BAT). This study considers the problem of modeling the Swift BAT triggering algorithm for long GRBs, a computationally expensive procedure, and models it using machine learning algorithms. A large sample of simulated GRBs from Lien et al. (2014) is used to train various models: random forests, boosted decision trees (with AdaBoost), support vector machines, and artificial neural networks. The best models have accuracies of approximately greater than 97% (approximately less than 3% error), which is a significant improvement on a cut in GRB flux which has an accuracy of 89:6% (10:4% error). These models are then used to measure the detection efficiency of Swift as a function of redshift z, which is used to perform Bayesian parameter estimation on the GRB rate distribution. We find a local GRB rate density of eta(sub 0) approximately 0.48(+0.41/-0.23) Gpc(exp -3) yr(exp -1) with power-law indices of eta(sub 1) approximately 1.7(+0.6/-0.5) and eta(sub 2) approximately -5.9(+5.7/-0.1) for GRBs above and below a break point of z(sub 1) approximately 6.8(+2.8/-3.2). This methodology is able to improve upon earlier studies by more accurately modeling Swift detection and using this for fully Bayesian model fitting. The code used in this is analysis is publicly available online.

gamma rays: general↗

Characterizing the Gamma-Ray Variability of the Brightest Flat Spectrum Radio Quasars Observed with the Fermi LAT

Almost 10 yr of γ-ray observations with the Fermi Large Area Telescope have revealed extreme γ-ray outbursts from flat spectrum radio quasars (FSRQs), temporarily making these objects the brightest γ-ray emitters in the sky. Yet, the location and mechanisms of the γ-ray emission remain elusive. We characterize long-term γ-ray variability and the brightest γ-ray flares of six FSRQs. Consecutively zooming in on the brightest flares, which we identify in an objective way through Bayesian blocks and a hill-climbing algorithm, we find variability on subhour timescales and as short as minutes for two sources in our sample (3C 279 and CTA 102) and weak evidence for variability at timescales less than the Fermi satellite's orbit of 95 minutes for PKS 1510–089 and 3C 454.3. This indicates extremely compact emission regions in the jet. We do not find any signs of γ-ray absorption in the broad-line region (BLR), which indicates that γ-rays are produced at distances greater than hundreds of gravitational radii from the central black hole. This is further supported by a cross-correlation analysis between γ-ray and radio/millimeter light curves, which is consistent with γ-ray production at the same location as the millimeter core for 3C 273, CTA 102, and 3C 454.3. The inferred locations of the γ-ray production zones are still consistent with the observed decay times of the brightest flares if the decay is caused by external Compton scattering with BLR photons. Yet, the minute-scale variability is challenging to explain in such scenarios.

79 ASTRONOMY AND ASTROPHYSICS↗

Toward Accelerated Nuclear-physics Parameter Estimation from Binary Neutron Star Mergers: Emulators for the Tolman–Oppenheimer–Volkoff Equations

Abstract Gravitational-wave observations of binary neutron-star (BNS) mergers have the potential to revolutionize our understanding of the nuclear equation of state (EOS) and the fundamental interactions that determine its properties. However, Bayesian parameter estimation frameworks do not typically sample over microscopic nuclear-physics parameters that determine the EOS. One of the major hurdles in doing so is the computational cost involved in solving the neutron-star structure equations, known as the Tolman–Oppenheimer–Volkoff (TOV) equations. In this paper, we explore approaches to emulating solutions for the TOV equations: multilayer perceptrons (MLPs), Gaussian processes, and a data-driven variant of the reduced basis method (RBM). We implement these emulators for three different parameterizations of the nuclear EOS, each with a different degree of complexity represented by the number of model parameters. We find that our MLP-based emulators are generally more accurate than the other two algorithms, whereas the RBM results in the largest speedup with respect to the full high-fidelity TOV solver. We employ these emulators for a simple parameter inference using a potentially loud BNS observation and show that the posteriors predicted by our emulators are in excellent agreement with those obtained from the full TOV solver.

79 ASTRONOMY AND ASTROPHYSICS↗

We Know Where We Don't Know: 3D Bayesian CNNs for Credible Geometric Uncertainty

Deep learning has been successfully applied to the segmentation of 3D Computed Tomography (CT) scans. Establishing the credibility of these segmentations requires uncertainty quantification (UQ) to identify untrustworthy predictions. Recent UQ architectures include Monte Carlo dropout networks (MCDNs), which approximate deep Gaussian processes, and Bayesian neural networks (BNNs), which learn the distribution of the weight space. BNNs are advantageous over MCDNs for UQ but are thought to be computationally infeasible in high dimension, and neither architecture has produced interpretable geometric uncertainty maps. We propose a novel 3D Bayesian convolutional neural network (BCNN), the first deep learning method which generates statistically credible geometric uncertainty maps and scales for application to 3D data. We present experimental results on CT scans of graphite electrodes and laser-welded metals and show that our BCNN outperforms an MCDN in recent uncertainty metrics. The geometric uncertainty maps generated by our BCNN capture distributions of sigmoid values that are interpretable as confidence intervals, critical for applications that rely on deep learning for high-consequence decisions.

97 MATHEMATICS AND COMPUTING↗

Mathematical algorithms for approximate reasoning

Most state of the art expert system environments contain a single and often ad hoc strategy for approximate reasoning. Some environments provide facilities to program the approximate reasoning algorithms. However, the next generation of expert systems should have an environment which contain a choice of several mathematical algorithms for approximate reasoning. To meet the need for validatable and verifiable coding, the expert system environment must no longer depend upon ad hoc reasoning techniques but instead must include mathematically rigorous techniques for approximate reasoning. Popular approximate reasoning techniques are reviewed, including: certainty factors, belief measures, Bayesian probabilities, fuzzy logic, and Shafer-Dempster techniques for reasoning. A group of mathematically rigorous algorithms for approximate reasoning are focused on that could form the basis of a next generation expert system environment. These algorithms are based upon the axioms of set theory and probability theory. To separate these algorithms for approximate reasoning various conditions of mutual exclusivity and independence are imposed upon the assertions. Approximate reasoning algorithms presented include: reasoning with statistically independent assertions, reasoning with mutually exclusive assertions, reasoning with assertions that exhibit minimum overlay within the state space, reasoning with assertions that exhibit maximum overlay within the state space (i.e. fuzzy logic), pessimistic reasoning (i.e. worst case analysis), optimistic reasoning (i.e. best case analysis), and reasoning with assertions with absolutely no knowledge of the possible dependency among the assertions. A robust environment for expert system construction should include the two modes of inference: modus ponens and modus tollens. Modus ponens inference is based upon reasoning towards the conclusion in a statement of logical implication, whereas modus tollens inference is based upon reasoning away from the conclusion. These algorithms allow one to reason accurately with uncertain data. The above environment can replicate state-f-the-art expert system environments which provides a continuity between the current expert systems which cannot be validated or verified and future expert systems which should be both validated and verified

Murphy, John H.↗

THE MASS DEPENDENCE BETWEEN PROTOPLANETARY DISKS AND THEIR STELLAR HOSTS

We present a substantial extension of the millimeter (mm) wave continuum photometry catalog for circumstellar dust disks in the Taurus star-forming region, based on a new ''snapshot'' {lambda} = 1.3 mm survey with the Submillimeter Array. Combining these new data with measurements in the literature, we construct a mm-wave luminosity distribution, f(L{sub mm}), for Class II disks that is statistically complete for stellar hosts with spectral types earlier than M8.5 and has a 3{sigma} depth of roughly 3 mJy. The resulting census eliminates a longstanding selection bias against disks with late-type hosts, and thereby demonstrates that there is a strong correlation between L{sub mm} and the host spectral type. By translating the locations of individual stars in the Hertzsprung-Russell diagram into masses and ages, and adopting a simple conversion between L{sub mm} and the disk mass, M{sub d} , we confirm that this correlation corresponds to a statistically robust relationship between the masses of dust disks and the stars that host them. A Bayesian regression technique is used to characterize these relationships in the presence of measurement errors, data censoring, and significant intrinsic scatter: the best-fit results indicate a typical 1.3 mm flux density of {approx}25 mJy for 1 M{sub Sun} hosts and a power-law scaling L{sub mm}{proportional_to}M{sub *}{sup 1.5-2.0}. We suggest that a reasonable treatment of dust temperature in the conversion from L{sub mm} to M{sub d} favors an inherently linear M{sub d} {proportional_to}M{sub *} scaling, with a typical disk-to-star mass ratio of {approx}0.2%-0.6%. The measured rms dispersion around this regression curve is {+-}0.7 dex, suggesting that the combined effects of diverse evolutionary states, dust opacities, and temperatures in these disks imprint a full width at half-maximum range of a factor of {approx}40 on the inferred M{sub d} (or L{sub mm}) at any given host mass. We argue that this relationship between M{sub d} and M{sub *} likely represents the origin of the inferred correlation between giant planet frequency and host star mass in the exoplanet population, and provides some basic support for the core accretion model for planet formation. Moreover, we caution that the effects of incompleteness and selection bias must be considered in comparative studies of disk evolution, and illustrate that fact with statistical comparisons of f(L{sub mm}) between the Taurus catalog presented here and incomplete subsamples in the Ophiuchus, IC 348, and Upper Sco young clusters.

79 ASTRONOMY AND ASTROPHYSICS↗

Modeling the Swift Bat Trigger Algorithm with Machine Learning

To draw inferences about gamma-ray burst (GRB) source populations based on Swift observations, it is essential to understand the detection efficiency of the Swift burst alert telescope (BAT). This study considers the problem of modeling the Swift / BAT triggering algorithm for long GRBs, a computationally expensive procedure, and models it using machine learning algorithms. A large sample of simulated GRBs from Lien et al. is used to train various models: random forests, boosted decision trees (with AdaBoost), support vector machines, and artificial neural networks. The best models have accuracies of greater than or equal to 97 percent (less than or equal to 3 percent error), which is a significant improvement on a cut in GRB flux, which has an accuracy of 89.6 percent (10.4 percent error). These models are then used to measure the detection efficiency of Swift as a function of redshift z, which is used to perform Bayesian parameter estimation on the GRB rate distribution. We find a local GRB rate density of n (sub 0) approaching 0.48 (sup plus 0.41) (sub minus 0.23) per cubic gigaparsecs per year with power-law indices of n (sub 1) approaching 1.7 (sup plus 0.6) (sub minus 0.5) and n (sub 2) approaching minus 5.9 (sup plus 5.7) (sub minus 0.1) for GRBs above and below a break point of z (redshift) (sub 1) approaching 6.8 (sup plus 2.8) (sub minus 3.2). This methodology is able to improve upon earlier studies by more accurately modeling Swift detection and using this for fully Bayesian model fitting.

gamma-ray burst: general – gamma-rays: general â↗

TOWARD CHARACTERIZATION OF THE TYPE IIP SUPERNOVA PROGENITOR POPULATION: A STATISTICAL SAMPLE OF LIGHT CURVES FROM Pan-STARRS1

In recent years, wide-field sky surveys providing deep multiband imaging have presented a new path for indirectly characterizing the progenitor populations of core-collapse supernovae (SNe): systematic light-curve studies. We assemble a set of 76 grizy-band Type IIP SN light curves from Pan-STARRS1, obtained over a constant survey program of 4 yr and classified using both spectroscopy and machine-learning-based photometric techniques. We develop and apply a new Bayesian model for the full multiband evolution of each light curve in the sample. We find no evidence of a subpopulation of fast-declining explosions (historically referred to as ''Type IIL'' SNe). However, we identify a highly significant relation between the plateau phase decay rate and peak luminosity among our SNe IIP. These results argue in favor of a single parameter, likely determined by initial stellar mass, predominantly controlling the explosions of red supergiants. This relation could also be applied for SN cosmology, offering a standardizable candle good to an intrinsic scatter of ≲ 0.2 mag. We compare each light curve to physical models from hydrodynamic simulations to estimate progenitor initial masses and other properties of the Pan-STARRS1 Type IIP SN sample. We show that correction of systematic discrepancies between modeled and observed SN IIP light-curve properties and an expanded grid of progenitor properties are needed to enable robust progenitor inferences from multiband light-curve samples of this kind. This work will serve as a pathfinder for photometric studies of core-collapse SNe to be conducted through future wide-field transient searches.

79 ASTRONOMY AND ASTROPHYSICS↗

The hunt for exomoons with Kepler (HEK). IV. A search for moons around eight M dwarfs

With their smaller radii and high cosmic abundance, transiting planets around cool stars hold a unique appeal. As part of our ongoing project to measure the occurrence rate of extrasolar moons, in this work we present results from a survey focusing on eight Kepler planetary candidates associated with M dwarfs. Using photodynamical modeling and Bayesian multimodal nested sampling, we find no compelling evidence for an exomoon in these eight systems. Upper limits on the presence of such bodies probe down to masses of ∼0.4 M {sub ⊕} in the best case. For KOI-314, we are able to confirm the planetary nature of two out of the three known transiting candidates using transit timing variations. Of particular interest is KOI-314c, which is found to have a mass of 1.0{sub −0.3}{sup +0.4} M {sub ⊕}, making it the lowest mass transiting planet discovered to date. With a radius of 1.61{sub −0.15}{sup +0.16} R {sub ⊕}, this Earth-mass world is likely enveloped by a significant gaseous envelope comprising ≥17{sub −13}{sup +12}% of the planet by radius. We also find evidence to support the planetary nature of KOI-784 via transit timing, but we advocate further observations to verify the signals. In both systems, we infer that the inner planet has a higher density than the outer world, which may be indicative of photo-evaporation. These results highlight both the ability of Kepler to search for sub-Earth-mass moons and the exciting ancillary science that often results from such efforts.

79 ASTRONOMY AND ASTROPHYSICS↗

Ps and Qs: Quantization-Aware Pruning for Efficient Low Latency Neural Network Inference

Efficient machine learning implementations optimized for inference in hardware have wide-ranging benefits, depending on the application, from lower inference latency to higher data throughput and reduced energy consumption. Two popular techniques for reducing computation in neural networks are pruning, removing insignificant synapses, and quantization, reducing the precision of the calculations. In this work, we explore the interplay between pruning and quantization during the training of neural networks for ultra low latency applications targeting high energy physics use cases. Techniques developed for this study have potential applications across many other domains. We study various configurations of pruning during quantization-aware training, which we term quantization-aware pruning, and the effect of techniques like regularization, batch normalization, and different pruning schemes on performance, computational complexity, and information content metrics. We find that quantization-aware pruning yields more computationally efficient models than either pruning or quantization alone for our task. Further, quantization-aware pruning typically performs similar to or better in terms of computational efficiency compared to other neural architecture search techniques like Bayesian optimization. Surprisingly, while networks with different training configurations can have similar performance for the benchmark application, the information content in the network can vary significantly, affecting its generalizability.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Measuring the Properties of Dark Energy with Photometrically Classified Pan-STARRS Supernovae. I. Systematic Uncertainty from Core-collapse Supernova Contamination

The Pan-STARRS (PS1) Medium Deep Survey discovered over 5000 likely supernovae (SNe) but obtained spectral classifications for just 10% of its SN candidates. We measured spectroscopic host galaxy redshifts for 3147 of these likely SNe and estimate that ∼1000 are Type Ia SNe (SNe Ia) with light-curve quality sufficient for a cosmological analysis. We use these data with simulations to determine the impact of core-collapse SN (CC SN) contamination on measurements of the dark energy equation of state parameter, w. Using the method of Bayesian Estimation Applied to Multiple Species (BEAMS), distances to SNe Ia and the contaminating CC SN distribution are simultaneously determined. We test light-curve-based SN classification priors for BEAMS as well as a new classification method that relies upon host galaxy spectra and the association of SN type with host type. By testing several SN classification methods and CC SN parameterizations on large SN simulations, we estimate that CC SN contamination gives a systematic error on w (σ{sub w}{sup CC}) of 0.014, 29% of the statistical uncertainty. Our best method gives σ{sub w}{sup CC}=0.004, just 8% of the statistical uncertainty, but could be affected by incomplete knowledge of the CC SN distribution. This method determines the SALT2 color and shape coefficients, α and β, with ∼3% bias. However, we find that some variants require α and β to be fixed to known values for BEAMS to yield accurate measurements of w. Finally, the inferred abundance of bright CC SNe in our sample is greater than expected based on measured CC SN rates and luminosity functions.

79 ASTRONOMY AND ASTROPHYSICS↗