Search NASA⌕ Search

Engineering topics

Liu, Jiaming

Publications and source records attributed to Liu, Jiaming.

DOLCE: A Model-Based Probabilistic Diffusion Framework for Limited-Angle CT Reconstruction

Limited-Angle Computed Tomography (LACT) is a non-destructive evaluation technique used in a variety of applications ranging from security to medicine. The limited angle coverage in LACT is often a dominant source of severe artifacts in the reconstructed images, making it a challenging inverse problem. We present DOLCE, a new deep model-based framework for LACT that uses a conditional diffusion model as an image prior. Diffusion models are a recent class of deep generative models that are relatively easy to train due to their implementation as image denoisers. DOLCE can form high-quality images from severely under-sampled data by integrating data-consistency updates with the sampling updates of a diffusion model, which is conditioned on the transformed limited-angle data. We show through extensive experimentation on several challenging real LACT datasets that, the same pre-trained DOLCE model achieves the SOTA performance on drastically different types of images. Additionally, we show that, unlike standard LACT reconstruction methods, DOLCE naturally enables the quantification of the reconstruction uncertainty by generating multiple samples consistent with the measured data.

Kim, Hyojin↗

SGD-Net: Efficient Model-Based Deep Learning with Theoretical Guarantees

Deep unfolding networks have recently gained popularity for solving imaging inverse problems. However, the computational and memory complexity of data-consistency layers within traditional deep unfolding networks scales with the number of measurements, limiting their applicability to large-scale imaging inverse problems. We propose SGD-Net as a new methodology for improving the efficiency of deep unfolding through stochastic approximations of the data-consistency layers. Our theoretical analysis shows that SGD-Net can be trained to approximate batch deep unfolding networks to an arbitrary precision. Our simulations on intensity diffraction tomography and sparse-view computed tomography show that SGD-Net can match the performance of the traditional batch network at a fraction of training and testing complexity.Deep unfolding networks have recently gained popularity for solving imaging inverse problems. However, the computational and memory complexity of data-consistency layers within traditional deep unfolding networks scales with the number of measurements, limiting their applicability to large-scale imaging inverse problems. We propose SGD-Net as a new methodology for improving the efficiency of deep unfolding through stochastic approximations of the data-consistency layers. Our theoretical analysis shows that SGD-Net can be trained to approximate batch deep unfolding networks to an arbitrary precision. Our simulations on intensity diffraction tomography and sparse-view computed tomography show that SGD-Net can match the performance of the traditional batch network at a fraction of training and testing complexity.

97 MATHEMATICS AND COMPUTING↗

Diagnosing Open Cluster Stock 2: Member Candidates and Mass Distribution with Gaia DR2 and LAMOST

We identify 1325 member candidates of the open cluster (OC) Stock 2 using data from Gaia DR2. We use the algorithms Clusterix 2.0 and HDBSCAN to select cluster candidates and further refine the final cluster membership by defining neighbors in 5D phase space (X {sub cp}, Y {sub cp}, Z{sub cp},κ⋅μ{sub α}{sup ∗}/ϖ, κ · μ {sub δ}/ϖ). Among these candidates, less than half have G, G {sub BP}, and G {sub RP} extinctions from Gaia. When Gaia extinctions are unavailable, we compute extiction using empirical formulas and E(B − V) = 0.350. We analyze the spatial distribution and mass profile of Stock 2. Our results reveal Stock 2 is still a bound OC and we find evidence of mass segregation. By comparing initial mass functions, the present-day mass function indicates that Stock 2 is a massive stellar cluster with a mass of 4000 M {sub ⊙}. The core radius and tidal radius, calculated via the radial density profile and total mass, are 3.97 pc and 22.65 pc, respectively. Common stars between our selected member candidates and the Large Sky Area Multi-Object Fiber Spectroscopic Telescope DR7 medium-resolution catalog give a metalliclity of [Fe/H] = −0.040 ± 0.147.

47 OTHER INSTRUMENTATION↗

Boosting the Performance of Plug-and-Play Priors via Denoiser Scaling

Plug-and-play priors (PnP) is an image reconstruction framework that utilizes an image denoiser as an imaging prior. Unlike traditional regularized inversion, PnP does not require the prior to be expressible in the form of a regularization function. This flexibility enables PnP algorithms to exploit the most effective image denoisers, leading to their state-of-the-art performance in various imaging tasks. However, many powerful denoisers, such as the ones based on convolutional neural networks (CNNs), do not have tunable parameters that would allow controlling their influence within PnP. To address this issue, in this paper, we introduce a scaling parameter that adjusts the magnitude of the denoiser input and output. We theoretical justify the denoiser scaling from the perspectives of proximal optimization, statistical estimation, and consensus equilibrium. Finally, we provide numerical experiments demonstrating the ability of denoiser scaling to systematically improve the performance of PnP for denoising CNN priors that do not have explicitly tunable parameters.

Boosting↗

Provable Convergence of Plug-and-Play Priors With MMSE Denoisers

Plug-and-play priors (PnP) is a methodology for regularized image reconstruction that specifies the prior through an image denoiser. While PnP algorithms are well understood for denoisers performing maximum a posteriori probability (MAP) estimation, they have not been analyzed for the minimum mean squared error (MMSE) denoisers. Here we address this gap by establishing the first theoretical convergence result for the iterative shrinkage/thresholding algorithm (ISTA) variant of PnP for MMSE denoisers. We show that the iterates produced by PnP-ISTA with an MMSE denoiser converge to a stationary point of some global cost function. We validate our analysis on sparse signal recovery in compressive sensing by comparing two types of denoisers, namely the exact MMSE denoiser and the approximate MMSE denoiser obtained by training a deep neural net.

97 MATHEMATICS AND COMPUTING↗

On the Relation of Silicates and SiO Maser in Evolved Stars

The SiO molecule is one of the candidates for the seed of silicate dust in the circumstellar envelope of evolved stars, but this opinion is challenged. In this work we investigate the relation of the SiO maser emission power and the silicate dust emission power. With both our own observation by using the PMO/Delingha 13.7 m telescope and archive data, a sample is assembled of 21 SiO v = 1, J = 2 − 1 sources and 28 SiO v = 1, J = 1 − 0 sources that exhibit silicate emission features in the ISO /SWS spectrum as well. The analysis of their SiO maser and silicate emission power indicates a clear correlation, which is not against the hypothesis that the SiO molecules are the seed nuclei of silicate dust. On the other hand, no correlation is found between SiO maser and silicate crystallinity, which may imply that silicate crystallinity does not correlate with mass-loss rate.

79 ASTRONOMY AND ASTROPHYSICS↗

A SYSTEMATIC SEARCH FOR THE SPECTRA WITH FEATURES OF CRYSTALLINE SILICATES IN THE SPITZER IRS ENHANCED PRODUCTS

The crystalline silicate features are mainly reflected in infrared bands. The Spitzer Infrared Spectrograph (IRS) collected numerous spectra of various objects and provided a big database to investigate crystalline silicates in a wide range of astronomical environments. We apply the manifold ranking algorithm to perform a systematic search for the spectra with crystalline silicate features in the Spitzer IRS Enhanced Products available. In total, 868 spectra of 790 sources are found to show the features of crystalline silicates. These objects are cross-matched with the SIMBAD database as well as with the Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST)/DR2. The average spectrum of young stellar objects shows a variety of features dominated either by forsterite or enstatite or neither, while the average spectrum of evolved objects consistently present dominant features of forsterite in AGB, OH/IR, post-AGB, and planetary nebulae. They are identified optically as early-type stars, evolved stars, galaxies and so on. In addition, the strength of spectral features in typical silicate complexes is calculated. The results are available through CDS for the astronomical community to further study crystalline silicates.

79 ASTRONOMY AND ASTROPHYSICS↗

A PRECISE determination of the mid-infrared interstellar extinction law based on the Apogee spectroscopic survey

A precise measure of the mid-infrared interstellar extinction law is crucial for investigating the properties of interstellar dust, especially larger-sized grains. Based on the stellar parameters derived from the SDSS-III/Apache Point Observatory Galaxy Evolution Experiment (APOGEE) spectroscopic survey, we select a large sample of G-type and K-type giants as the tracers of the Galactic mid-infrared extinction. We calculate the intrinsic stellar color excesses from the stellar effective temperatures and use them to determine the mid-infrared extinction for a given line of sight. For the entire sky of the Milky Way surveyed by APOGEE, we derive the extinctions (relative to A{sub K{sub S}}, the K {sub S}-band extinction at wavelength λ = 2.16 μm) for the four Wide-field Infrared Survey Explorer (WISE) bands at 3.4, 4.6, 12, and 22 μm, the four Spitzer/Infrared Array Camera bands at 3.6, 4.5, 5.8, and 8 μm, the Spitzer/MIPS24 band at 23.7 μm, and, for the first time, the AKARI/S9W band at 8.23 μm. Our results agree with previous works in that the extinction curve is flat in the ∼3–8 μm wavelength range and is generally consistent with the R{sub V} = 5.5 model curve, except our determination exceeds the model prediction in the WISE/W4 band. Although some previous works found that the mid-IR extinction law appears to vary with the extinction depth A{sub K{sub S}}, no noticeable variation has been found in this work. The uncertainties are analyzed in terms of the bootstrap resampling method and Monte-Carlo simulation and are found to be rather small.

79 ASTRONOMY AND ASTROPHYSICS↗

Substrate Specificity within a Family of Outer Membrane Carboxylate Channels

Many Gram-negative bacteria, including human pathogens such as Pseudomonas aeruginosa, do not have large-channel porins. This results in an outer membrane (OM) that is highly impermeable to small polar molecules, making the bacteria intrinsically resistant towards many antibiotics. In such microorganisms, the majority of small molecules are taken up by members of the OprD outer membrane protein family. Here we show that OprD channels require a carboxyl group in the substrate for efficient transport, and based on this we have renamed the family Occ, for outer membrane carboxylate channels. We further show that Occ channels can be divided into two subfamilies, based on their very different substrate specificities. Our results rationalize how certain bacteria can efficiently take up a variety of substrates under nutrient-poor conditions without compromising membrane permeability. In addition, they explain how channel inactivation in response to antibiotics can cause resistance but does not lead to decreased fitness.

59 BASIC BIOLOGICAL SCIENCES↗

{sup 149}Sm(n,{alpha}){sup 146}Nd Cross Sections in the MeV Region

We have measured the {sup 149}Sm(n,{alpha}){sup 146}Nd cross section at 4.5, 5.0, 5.5, 6.0, and 6.5 MeV. Measurements were performed at the 4.5 MV Van de Graaff accelerator of Peking University with monoenergetic neutrons produced via the {sup 2}H(d,n){sup 3}He reaction using a deuterium gas target. Alpha particles were detected with a double-section gridded ionization chamber having two back-to-back {sup 149}Sm{sub 2}O{sub 3} samples attached to the common cathode. Absolute neutron flux was measured using a small {sup 238}U fission chamber and monitored by a BF{sub 3} long counter. These are the first reported cross sections for this reaction at these energies, except at 6.0 eV, where our new data are in good agreement with our earlier result. The present results help to much better constrain the {sup 149}Sm(n,{alpha}){sup 146}Nd cross section in a region where its energy dependence is changing fairly rapidly and there are large differences between evaluated nuclear data libraries.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Cross Section Measurement for the 95Mo(n, {alpha})92Zr Reaction at 4.0, 5.0 and 6.0 MeV

For the {sup 95}Mo(n, {alpha}){sup 92}Zr reaction cross section, there is only one experimental datum in the MeV neutron energy region with large uncertainty. As a result, very large deviations exist in different evaluated nuclear data libraries. This paper report the measurement of cross sections of the {sup 95}Mo(n, {alpha}){sup 92}Zr reaction at En = 4.0, 5.0 and 6.0 MeV. Experiments were performed at the 4.5 MV Van de Graaff of Peking University, China. A twin gridded ionization chamber was used as alpha particle detector and two large area {sup 95}Mo samples placed back to back were adopted. Fast neutrons were produced through the D(d, n){sup 3}He reaction by using a deuterium gas target. A small {sup 238}U fission chamber was adopted for absolute neutron flux determination and a BF{sub 3} long counter was used for neutron flux monitor. Present experimental data are compared with existing evaluations and measurement.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Cross-section measurement and analysis for the {sup 149}Sm(n,alpha){sup 146}Nd reaction at 6.0 MeV

We have measured the {sup 149}Sm(n,alpha){sup 146}Nd cross section at 6.0 MeV to be 0.12+-0.018 mb. This is the first reported result for this cross section in the MeV region and so should be helpful for constraining nuclear data evaluations (which differ by a factor of 40 for this reaction at this energy) and for testing and improving nuclear models. The experiment was performed at the at the 4.5-MV Van de Graaff accelerator of Peking University. Neutrons were produced via the {sup 2}H(d,n){sup 3}He reaction using a deuterium gas target, and absolute neutron flux was determined with a small {sup 238}U fission chamber. Alpha particles were detected using a two-section gridded ionization chamber in which two large-area {sup 149}Sm{sub 2}O{sub 3} samples were placed back-to-back so that the cross section and forward/backward ratio were measured for nearly the entire 4pi solid angle. The data were compared to statistical-model predictions using the code talys. Good agreement between the measured and theoretical cross sections could be obtained with a small modification of the alpha-optical potential parameters from their default values in talys. However, we were not able to reproduce the measured forward/backward ratio. In addition, using the previously reported {sup 147}Sm(n,alpha){sup 144}Nd cross section at 6.0 MeV together with our new result, good agreement was found between the measured and predicted cross-section ratio for these two isotopes. Finally, none of the existing evaluated data libraries are in agreement with our new data to within the experimental uncertainties.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Cross Section Measurement for the 95Mo(n,alpha)92Zr Reaction at 4.0, 5.0 and 6.0 MeV

Measurements of cross sections of the {sup 95}Mo(n, {alpha}){sup 92}Zr reaction at E{sub n} = 4.0, 5.0 and 6.0 MeV were carried out at the 4.5 MV Van de Graaff of Peking University, China. A twin gridded ionization chamber and two large-area {sup 95}Mo samples were adopted. Fast neutrons were produced through the D(d, n){sup 3}He reaction by using a deuterium gas target. A small {sup 238}U fission chamber was employed for absolute neutron flux determination. Present data are compared with existing evaluations and measurement.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗