Search

StyleGAN has shown remarkable performance in unconditional image generation. However, its high computational cost poses a significant challenge for practical applications. Although recent efforts have been made to compress StyleGAN while preserving its performance, existing compressed models still lag behind the original model, particularly in terms of sample diversity. To overcome this, we propose a novel channel pruning method that leverages varying sensitivities of channels to latent vectors, which is a key factor in sample diversity. Specifically, by assessing channel importance based on their sensitivities to latent vector perturbations, our method enhances the diversity of samples in the compressed model. Since our method solely focuses on the channel pruning stage, it has complementary benefits with prior training schemes without additional training cost. Extensive experiments demonstrate that our method significantly enhances sample diversity across various datasets. Moreover, in terms of FID scores, our method not only surpasses state-of-the-art by a large margin but also achieves comparable scores with only half training iterations., Comment: Accepted to CVPR 2024. Project page: https://jiwoogit.github.io/DCP-GAN_site

Many sub-Neptune exoplanets have been believed to be composed of a thick hydrogen-dominated atmosphere and a high-temperature heavier-element-dominant core. From an assumption that there is no chemical reaction between hydrogen and silicates/metals at the atmosphere-interior boundary, the cores of sub-Neptunes have been modeled with molten silicates and metals (magma) in previous studies. In large sub-Neptunes, pressure at the atmosphere-magma boundary can reach tens of gigapascals where hydrogen is a dense liquid. A recent experiment showed that hydrogen can induce the reduction of Fe$^{2+}$ in (Mg,Fe)O to Fe$^0$ metal at the pressure-temperature conditions relevant to the atmosphere-interior boundary. However, it is unclear if Mg, one of the abundant heavy elements in the planetary interiors, remains oxidized or can be reduced by H. Our experiments in the laser-heated diamond-anvil cell found that heating of MgO + Fe to 3500-4900 K (close to or above their melting temperatures) in a H medium leads to the formation of Mg$_2$FeH$_6$ and H$_2$O at 8-13 GPa. At 26-29 GPa, the behavior of the system changes, and Mg-H in an H fluid and H$_2$O were detected with separate FeH$_x$. The observations indicate the dissociation of the Mg-O bond by H and subsequent production of hydride and water. Therefore, the atmosphere-magma interaction can lead to a fundamentally different mineralogy for sub-Neptune exoplanets compared with rocky planets. The change in the chemical reaction at the higher pressures can also affect the size demographics (i.e., "radius cliff") and the atmosphere chemistry of sub-Neptune exoplanets.

In this paper, we first investigate a visual quality degradation problem observed in recent high-resolution virtual try-on approach. The tendency is empirically found that the textures of clothes are squeezed at the sleeve, as visualized in the upper row of Fig.1(a). A main reason for the issue arises from a gradient conflict between two popular losses, the Total Variation (TV) and adversarial losses. Specifically, the TV loss aims to disconnect boundaries between the sleeve and torso in a warped clothing mask, whereas the adversarial loss aims to combine between them. Such contrary objectives feedback the misaligned gradients to a cascaded appearance flow estimation, resulting in undesirable squeezing artifacts. To reduce this, we propose a Sequential Deformation (SD-VITON) that disentangles the appearance flow prediction layers into TV objective-dominant (TVOB) layers and a task-coexistence (TACO) layer. Specifically, we coarsely fit the clothes onto a human body via the TVOB layers, and then keep on refining via the TACO layer. In addition, the bottom row of Fig.1(a) shows a different type of squeezing artifacts around the waist. To address it, we further propose that we first warp the clothes into a tucked-out shirts style, and then partially erase the texture from the warped clothes without hurting the smoothness of the appearance flows. Experimental results show that our SD-VITON successfully resolves both types of artifacts and outperforms the baseline methods. Source code will be available at https://github.com/SHShim0513/SD-VITON., Comment: Accepted to AAAI 2024

In the next decades, the astrobiological community will debate whether the first observations of oxygen in an exoplanet$'$s atmosphere signifies life, so it is critical to establish procedures now for collection and interpretation of such data. We present a step-by-step observational strategy for using oxygen as a robust biosignature, to prioritize exoplanet targets and design future observations. It is premised on avoiding planets lacking subaerial weathering of continents, which would imply geochemical cycles drastically different from Earth$'$s, precluding use of oxygen as a biosignature. The strategy starts with the most readily obtained data: semi-major axis and stellar luminosity to ensure residence in the habitable zone; stellar XUV flux, to ensure an exoplanet can retain a secondary (outgassed) atmosphere. Next, high-precision mass and radius information should be combined with high-precision stellar abundance data, to constrain the exoplanet$'$s water content; those incompatible with less than 0.1 wt % H$_{2}$O can be deprioritized. Then, reflectance photometry or low-resolution transmission spectroscopy should confirm an optically thin atmosphere. Subsequent long-duration, high-resolution transmission spectroscopy should search for oxygen and ensure that water vapor and CO$_{2}$ are present only at low (10$^{2}$-10$^{4}$ ppm levels). Assuming oxygen is found, attribution to life requires the difficult acquisition of a detailed, multispectral light curve of the exoplanet to ensure both surface land and water. Exoplanets failing some of these steps might be habitable, even have observable biogenic oxygen, but should be deprioritized because oxygen could not be attributed unambiguously to life. We show how this is the case for the Solar System, the 55 Cnc System, and the TRAPPIST-1 System, in which only the Earth and TRAPPIST-1e successfully pass through our procedure., Comment: 27 Pages, 1 Figure, 0 Tables (accepted 09 June 2020, in press)

Here we advocate an observational strategy to help prioritize exoplanet observations. It starts with more easily obtained observational data, and ranks exoplanets for more difficult follow-up observations based on the likelihood of avoiding planets for which oxygen is a false positives or even an inconclusive signature of life. We find that for oxygen to be a robust biosignature, both land and surface water must be present. Landless exoplanets have much slower biogeochemical cycles, so while oxygenic photosynthesizing life could exist on such planets, it could not produce oxygen at a rate competitive with abiotic rates such as photolysis. These habitable planets, whose life would not be detectable, should be avoided., Comment: 6 pages, 1 figure, white paper submitted in response to the solicitation of feedback for the "Astrobiology Science Strategy for the Search for Life in the Universe" by the National Academy of Sciences

We have studied the ferrous (Fe2+) and ferric (Fe3+) iron concentrations in Al-free Fe containing Mg-silicate perovskite (Mg-Pv) at pressure (P), temperature (T), and oxygen fugacity (fO2) conditions related to the lower mantle using a thermodynamic model based on ab-initio calculations. We consider the oxidation reaction and the charge disproportionation reaction, both of which can produce Fe3+ in Mg-Pv. The model shows qualitatively good agreement with available experimental data on Fe3+/{\Sigma}Fe ({\Sigma}Fe = total Fe in system), spin transitions, and equations of state. We predict that under lower-mantle conditions Fe3+/{\Sigma}Fe determined by the charge disproportionation is estimated to be 0.01-0.07 in Al-free Mg-Pv, suggesting that low Al Mg-Pv in the uppermost pyrolitic mantle (where majoritic garnet contains most of the Al) and in the harzburgitic heterogeneities throughout the lower mantle contains very little Fe3+. We find that the volume reduction by the spin transition of the B-site Fe3+ leads to a minimum Fe3+/{\Sigma}Fe in Mg-Pv at mid-mantle pressures. The model shows that configurational entropy is a key driving force to create Fe3+ and therefore Fe3+ content is highly temperature sensitive. The temperature sensitivity may lead to a maximum Fe3+/{\Sigma}Fe in Mg-Pv in warm regions at the core-mantle boundary region, such as Large Low Shear Velocity Provinces (LLSVPs), potentially altering the physical (e.g., bulk modulus) and transport (e.g., thermal and electrical conductivities) properties of the heterogeneities.

{\tau} Ceti (HD10700), a G8 dwarf with solar mass of 0.78, is a close (3.65 pc) sun-like star where 5 possibly terrestrial planet candidates (minimum masses of 2, 3.1, 3.5, 4.3, and 6.7 Earth masses) have recently been discovered. We report abundances of 23 elements using spectra from the MIKE spectrograph on Magellan. Using stellar models with the abundances determined here, we calculate the position of the classical habitable zone with time. At the current best fit age, 7.63 Gy, up to two planets (e and f) may be in the habitable zone, depending on atmospheric properties. The Mg/Si ratio of the star is found to be 1.78, which is much greater than for Earth (about 1.2). With a system that has such an excess of Mg to Si ratio it is possible that the mineralogical make-up of planets around {\tau} Ceti could be significantly different from that of Earth, with possible oversaturation of MgO, resulting in an increase in the content of olivine and ferropericlase compared with Earth. The increase in MgO would have a drastic impact on the rheology of the mantles of the planets around {\tau} Ceti., Comment: 17 pages, 1 figure, supplemental table, Astrophysical Journal