id
stringlengths 20
32
| text
stringlengths 100
121k
| image_path
stringlengths 15
235
|
|---|---|---|
arxivcap_2212.13138_7 | Title: Large Language Models Encode Clinical Knowledge | Caption: Selective prediction analysis Analysis of deferral behavior of Flan-PaLM 540B model with self-consistency. We observe that if we defer more often using an uncertainty threshold based on self-consistency, the model becomes increasingly accurate on questions it does not defer | 2212.13138_7.jpg |
arxivcap_1109.6297_3 | Title: Low-rank data modeling via the Minimum Description Length principle | Caption: Results for the "ShoppingMall" sequence (see text for a description of the above pictures and graphs). In this case, the rank of the approximation decomposition is $\rank=7$. Here, the first left-eigenvector models the background, whereas the rest tend to capture people that stood still for a while. Here we see the "phantom" of two such persons in the second left-eigenvector (top-right). | 1109.6297_3.jpg |
arxivcap_2303.14662_1 | Title: OTAvatar: One-shot Talking Face Avatar with Controllable Tri-plane Rendering | Caption: OTAvatar animation results. The source subjects in HDTF<cit.> dataset are animated by OTAvatar using a single portrait as the reference. We use the pose and expression coefficients of 3DMM to represent motion and drive the avatar. Note that these subjects are not included in the training data of OTAvatar. | 2303.14662_1.jpg |
arxivcap_1902.04067_2 | Title: Multi-tier Caching Analysis in CDN-based Over-the-top Video Streaming Systems | Caption: Weighted SDTP versus the percentage bound on the number of video files (i.e., maximum capacity) in the edge cache $\epsilon$. The percentage of the capacity bound is changed from $0.05$ to $0.1$ for a cache capacity of $0.20\times C_{tot}$. | 1902.04067_2.jpg |
arxivcap_1906.05326_20 | Title: Magnetism and superconductivity in Fe$_{1+y}$Te$_{1-x}$Se$_x$ | Caption: From <cit.>. Inelastic magnetic neutron scattering from the SC70 sample measured on HYSPEC at energy transfers ??? = 13 meV (a), (b), (c); 10 meV (d), (e), (f); and 7 meV (g), (h), (i). The sample temperatures are 8 K (a), (d), (g); 100 K (b), (e), (h); and 300 K (c), (f), (i). All slices were taken with an energy width of 2 meV. Measurements, covering approximately two quadrants, have been symmetrized to be 4-fold symmetric, consistent with sample symmetry. Intensity scale is the same in all panels, but 13-meV data have been multiplied by 1.5 to improve visibility. Black regions at the center of each panel are outside of the detector range. Panels (j), (k), (l) are model calculations simulating the 7-meV data, as described in the text, based on weakly correlated slanted UDUD spin plaquettes [see Figs. 5(a) and 5(b)]. The wave vectors for the AFM interplaquette correlations used in the calculation are (j) 100% $Q_{\rm SAF}$, (k) 50% $Q_{\rm SAF}$ and 50% $Q_{\rm DSAF}$, and (l) 100% $Q_{\rm DSAF}$. 2016 American Physical Society. | 1906.05326_20.jpg |
arxivcap_2005.06243_6 | Title: Detecting and analyzing collusive entities on YouTube | Caption: Exploratory analysis of features used to characterise videos submitted for collusive likes. | 2005.06243_6.jpg |
arxivcap_1503.07205_2 | Title: The Observer Strikes Back | Caption: Fine and coarse-grained histories of a box model discussed in the text. The boxes model Hubble volumes. Their color models an observable like the CMB in this case either yellow or green. An `E' means that an observer is in the box observing its color. The top history is fine grained with a color and E or not E in every box. The bottom history is coarse grained. The possibilities have been summed over for every box except one - - our box. That enables a straightforward calculation of the first person probability that we see one color or the other. The details outside our box are irrelevant for this. | 1503.07205_2.jpg |
arxivcap_2303.06845_4 | Title: Transformer Encoder with Multiscale Deep Learning for Pain Classification Using Physiological Signals | Caption: Outline framework of our proposed PainAttnNet. Left bottom: Multiscale Convolutional Network (MSCN). Left top: Squeeze-and-Excitation Residual Network (SEResNet). Right: Transformer Encoder. | 2303.06845_4.jpg |
arxivcap_2005.06188_1 | Title: Star clusters near and far; tracing star formation across cosmic time | Caption: The Galactic young massive star cluster Westerlund 2 in the center of the HII region RCW 49 as it is seen by HST in the optical and near-infrared. This image shows the cluster's very massive O-star population with up to 80 $M_\odot$ stars but also very low-mass stars with only 0.1 $M_\odot$ or even below. Credit: NASA, ESA, the Hubble Heritage Team (STScI/AURA), A. Nota (ESA/STScI), and the Westerlund 2 Science Team. | 2005.06188_1.jpg |
arxivcap_astro-ph_0104294_1 | Title: The Large Area Lyman Alpha Survey | Caption: Keck spectrum of a confirmed $z=4.52$ Lyman-$\alpha$ source. This object has a line flux of $1.7 \times 10^{-17} {\rm\,erg\,cm^{-2}\,s^{-1}}$ and a rest frame equivalent width of $15$ÅṪhe line is asymmetric (see inset) and has a strong continuum decrement from the red to the blue side, both of which are expected for high-redshift Lyman-$\alpha$ emitters (Stern & Spinrad 1999). In order to accentuate the well-sampled line and suppress noise, the spectrum of this source has been smoothed with a boxcar filter of width 9 pixels $= 16.6$Å [main panel] or 3 pixels $= 5.5$Å [inset panel]. | astro-ph_0104294_1.jpg |
arxivcap_2305.05295_1 | Title: Boosting Zero-shot Cross-lingual Retrieval by Training on Artificially Code-Switched Data | Caption: Retrieval performance for different translation probabilities, averaged across all language pairs. | 2305.05295_1.jpg |
arxivcap_2108.09713_3 | Title: Robustness-via-Synthesis: Robust Training with Generative Adversarial Perturbations | Caption: We randomly sample a data point from each category in the test set of CIFAR10. Then, 100 different adversarial samples corresponding to the natural images are generated. the latent representations of the natural and their adversarial counterparts are plotted using t-SNE. In the early stages of the robust training, the domain gap between the natural and adversarial samples is visible in the latent space. Later in the robust training, the adversarial samples start grouping around the natural sample such that it is harder for the generator to find adversarial directions. | 2108.09713_3.jpg |
arxivcap_1104.3706_7 | Title: Detection of a large massive circumstellar disk around a high-mass young stellar object in the Carina Nebula | Caption: The disk object and its surroundings seen at different wavelengths. The upper left panel shows the optical HST image taken through the filter F555; the position of the (optically invisible) disk object is marked by the white circle. The next three panels show the HAWK-I images, and the lower row shows the Spitzer IRAC images. | 1104.3706_7.jpg |
arxivcap_2210.09436_1 | Title: Analytic Neutrino Oscillation Probabilities | Caption: Violation of unitary relations (<ref>) in matter (solid lines) and in vacuum (dashed lines). | 2210.09436_1.jpg |
arxivcap_2206.01820_1 | Title: A Robust Backpropagation-Free Framework for Images | Caption: t-SNE visualization depicting the learned representations of EKDAA, shown for Epoch $0$ (initial weights) and the final training epoch: (Left:) FMNIST and (Right) CIFAR-10. | 2206.01820_1.jpg |
arxivcap_1802.01257_3 | Title: Promoting cooperation by punishing minority | Caption: (Color online) The fraction of cooperators $\rho_{c}$ as a function of the punishment fine $\alpha$ for different values of the multiplication factor $r$. For each value of $r$, $\rho_{c}$ increases with $\alpha$, indicating that severe punishment is in favor of cooperation. | 1802.01257_3.jpg |
arxivcap_1401.8257_4 | Title: Online Clustering of Bandits | Caption: Results on synthetic datasets. Each plot displays the behavior of the ratio of the current cumulative regret of the algorithm ("Alg") to the current cumulative regret of RAN, where "Alg" is either "CLUB" or "LinUCB-IND" or "LinUCB-ONE" or "GOBLIN"or "CLAIRVOYANT". In the top two plots cluster sizes are balanced ($z=0$), while in the bottom two they are unbalanced ($z=2$). | 1401.8257_4.jpg |
arxivcap_2305.05282_4 | Title: Fooling State-of-the-Art Deepfake Detection with High-Quality Deepfakes | Caption: Left: Forward pass of our model. The loss is computed separately for each identity (A and B) and finally summed up. Right: Detailed architecture of the proposed decoder. | 2305.05282_4.jpg |
arxivcap_2202.13870_2 | Title: Simulating Network Paths with Recurrent Buffering Units | Caption: $MMD^2$ (with $\zeta= 0.1$) vs starting packet number (of chunks) on the dataset for (left) the training protocol Cubic and (right) the test protocol Vegas. | 2202.13870_2.jpg |
arxivcap_1109.6218_1 | Title: Dynamics and efficiency of a self-propelled, diffusiophoretic swimmer | Caption: Schematic representation of the model for an active swimmer. Molecules of type $a$ and $b$ are dissolved in the fluid surrounding the swimmer with radius $R$. The molecules are transformed into each other through chemical surface reactions and the overall concentrations are maintained constant. An interaction between the swimmer and the type $b$ molecules occurs through the potential $\tPsi$, whose lengthscale is given by $L$. An inhomogeneous distribution of solutes causes the swimmer to move with the speed $\tU$. $\mathbf{\tr}$ is the position vector and $\theta$ is the polar angle in a spherical coordinate system used throughout. | 1109.6218_1.jpg |
arxivcap_1503.07205_3 | Title: The Observer Strikes Back | Caption: Third (left) and first person (right) probabilities for the number of efolds $N_e$ of matter driven inflation in the early universe in the no-boundary quantum state $\Psi$. The first person probabilities favor universes with a large amount of inflation because in the larger universes that result from an extended period of early inflation there are more Hubble volumes for us to be. | 1503.07205_3.jpg |
arxivcap_2212.13138_8 | Title: Large Language Models Encode Clinical Knowledge | Caption: Clinician evaluation of answers agreement with scientific and clinical consensus Clinicians were asked to rate answers provided to questions in the HealthSearchQA, Live QA and Medication question answering datasets. Clinicians were asked to identify whether the answer is aligned with the prevailing medical/scientific consensus; whether the answer was in opposition to consensus; or whether there is no medical/scientific consensus for how to answer that particular question (or whether it was not possible to answer this question). We observe that while the Flan-PaLM model answers are only found to be in agreement with the scientific consensus 61.9% of the time, this improves to 92.9% for Med-PaLM matching expert answers. | 2212.13138_8.jpg |
arxivcap_1709.03144_26 | Title: The 12C(a,g)16O reaction and its implications for stellar helium burning | Caption: (Color online) Ratio of $^{12}$C and $^{16}$O mass fractions at the stellar center at the end of core He burning (left axis; red circles) and number ratios at the surface (right axis; grey and blue circles) during the 10$^{\rm th}$ and 15$^{\rm th}$ thermal pulses (He shell-flashes) in the 3 M$_\odot$ models. The points connected by lines are differences arising in the models when using the $^{12}$C$(\alpha,\gamma)^{16}$O rate of this work and that of <cit.>. The red arrow along the left y-axis spans the range of $X_\mathrm{C}(^{12}\mathrm{C}/^{16}\mathrm{O})$ resulting from the uncertainty in the rate of this work. The blue arrow along the right y-axis is the equivalent range of $Y_\mathrm{surf}(^{12}\mathrm{C}/^{16}\mathrm{O})$ during the 15$^\mathrm{th}$ thermal pulse. | 1709.03144_26.jpg |
arxivcap_1806.08917_5 | Title: Halide-Perovskite Resonant Nanophotonics | Caption: {Nanoimprinted perovskite metasurfaces for enhanced luminescence. (a) SEM images and respective schemes of the employed nanoimprint lithographic methods. (b) Photoluminescence enhancement of different metasurfaces (nanostripes, nanoholes) over the unstructured film. An increase over 70 times can be seen between the PL of the nanostripe film over the unstructured film, see Ref. <cit.>} | 1806.08917_5.jpg |
arxivcap_2210.09436_2 | Title: Analytic Neutrino Oscillation Probabilities | Caption: NHS identity (<ref>) is only satisfied in the case where the matter potential is diagonal (black solid line). Otherwise, it is described by eq. (<ref>) which depends on the matter potential (blue dashed and magenta dotted lines). | 2210.09436_2.jpg |
arxivcap_1912.05084_15 | Title: Bayesian Copula Density Deconvolution for Zero-Inflated Data in Nutritional Epidemiology | Caption: =10pt Results for the EATS data sets with sample size $n = 965$, $q=2$ episodic components, milk and whole grains, and $p=2$ regular components, sodium and energy, each subject having $m_{i}=4$ replicates. These results correspond to the final MCMC iteration but are representative of other iterations in steady state. The left panel shows the mixture component specific parameters $\{\mu_{X,k}, \sigma_{X,k}^{2}\}_{k=1}^{K_{X}=10}$ used to model the marginal densities $f_{X,\ell}(X_{\ell})$ of the two regular dietary components. The middle left panel shows the associated `empirical' mixture probabilities $\wh\pi_{X,\ell,k} = \sum_{i=1}^{n}1\{C_{X,\ell,i}=k\}/n$. Only the mixture components $2,3$ and $9$ were actually used to model the densities and these mixture components were shared between the two dietary components, the other mixture components were redundant. The middle right panel shows the mixture component specific parameters $\{p_{\epsilon,k}, \mu_{\epsilon,k}, \sigma_{\epsilon,k,1}^{2}, \sigma_{\epsilon,k,2}^{2}\}_{k=1}^{K_{\epsilon}=10}$ used to model the marginal densities $f_{\epsilon,\ell}(\epsilon_{\ell})$ of the scaled errors of all dietary components. The right panel shows the associated `empirical' mixture probabilities $\wh\pi_{\epsilon,\ell,k} = \sum_{i=1}^{n}\sum_{j=1}^{m_{i}}1\{C_{\epsilon,\ell,i,j}=k\}/\sum_{i=1}^{n}m_{i}$. Only the mixture components $1,2,3,4,5$ and $6$ were actually used to model the densities and these mixture components were shared between the four dietary components, the other mixture components were redundant. | 1912.05084_15.jpg |
arxivcap_1111.0261_5 | Title: Extended Lagrangian free energy molecular dynamics | Caption: (Color online) The fluctuations in the total free energy, $E^{\rm tot}_F = E_{\rm K} + U - T_e{\cal S}$, for a small Li cluster at $T_e = 2,000$ K simulated using FreeON within the density matrix extension. The Pulay force has been calculated either with the original zero-temperature expression, Eq. (<ref>), or its finite temperature generalization, Eq. (<ref>). | 1111.0261_5.jpg |
arxivcap_1906.05326_21 | Title: Magnetism and superconductivity in Fe$_{1+y}$Te$_{1-x}$Se$_x$ | Caption: From <cit.>. Thermal evolution of the magnetic scattering at $\hbar\omega= 5$ meV. The data are measured through QAF along the transverse direction for the Ni02 sample at (a) 100 K, (b) 40 K, (c) 15 K, (d) 2.8 K, and (e) for the Ni04 sample plotted as an intensity contour map in temperatureÐwave- vector space. The data have been smoothed. The yellow and black symbols in (e) denote the corresponding peak positions for the Ni02 sample (yellow squares) and for a superconducting Fe$_{1+y}$Te$_{0.35}$Se$_{0.65}$ sample (black circles) [17]. 2012 American Physical Society. | 1906.05326_21.jpg |
arxivcap_2202.01477_3 | Title: Unsourced Random Access with a Massive MIMO Receiver Using Multiple Stages of Orthogonal Pilots | Caption: {The required $E_b/N_0$ as a function of the number of active users in the proposed scheme and the results in <cit.> for $M=50$.} | 2202.01477_3.jpg |
arxivcap_1902.04067_5 | Title: Multi-tier Caching Analysis in CDN-based Over-the-top Video Streaming Systems | Caption: Edge-cache miss rate versus edge-cache capacity ratio. | 1902.04067_5.jpg |
arxivcap_2202.00203_12 | Title: Constraining low redshift [CII] Emission by Cross-Correlating FIRAS and BOSS Data | Caption: Comparison of measured values of $b_{\rm \cii}I_{\rm \cii}$ to various models from the literature. The gray boxes show the values allowed by the and 95% upper limits, with their widths representing the redshift ranges used. The black circle shows the measurement from <cit.>. The horizontal error bars show the full redshift range of that measurement, and the vertical error bars show the 95% confidence intervals on their measurement of $b_{\rm \cii}I_{\rm \cii}$. The black square shows the value inferred from the $z\sim0$ luminosity function <cit.>, assuming $b_{\rm \cii}=1$. The error bars come from their $1\sigma$ errors on luminosity function parameters. The purple hatched region shows the range of brightnesses allowed by the scaling-relation models from <cit.>, the yellow dotted region shows the range of brightnesses of the collisional excitation models from <cit.>, and the green dashed curve shows the semi-empirical model from <cit.>. The blue dashed curve shows the <cit.> model linearly rescaled to match the measurement at the appropriate redshift. The red region below 100 Jy/sr in the redshift region shows the projected sensitivity achievable with a cross-correlation between PIXIE and CMASS galaxies in the same redshift range used for the analysis. This would easily detect even the most pessimistic of the plotted models at high sensitivity. | 2202.00203_12.jpg |
arxivcap_2012.05453_3 | Title: Causal BERT : Language models for causality detection between events expressed in text | Caption: Masked Event C-BERT Pre-training on Masked Event C-BERT and Fine-Tuning on Event Aware C-BERT | 2012.05453_3.jpg |
arxivcap_2202.13870_5 | Title: Simulating Network Paths with Recurrent Buffering Units | Caption: Traces produced by (middle) and the (right) models along with Ground-Truth traces (left) for the test protocol TCP Cubic. The blue curves represent sending rates decided by the test protocol and the red curves represent delays generated by the models. For every pair of rows, the first row shows the entire trace and the second row focuses on the first few seconds. These are randomly picked traces, and there is no one-to-one correspondence between columns. | 2202.13870_5.jpg |
arxivcap_1503.07205_4 | Title: The Observer Strikes Back | Caption: A contour plot of the fraction $f(\Lambda, Q, \Dl)$ of baryons in the form of galaxies by the time $t_0 \sim 11 {\rm Gyr}$, adapted from the calculation of <cit.>. As discussed in the text, the fraction is negligible in the dark (blue) region either because gravitationally bound systems don't collapse or because they collapse to black holes. | 1503.07205_4.jpg |
arxivcap_2108.09522_1 | Title: A Technical Survey and Evaluation of Traditional Point Cloud Clustering Methods for LiDAR Panoptic Segmentation | Caption: The demonstration of the panoptic segmentation task and how our pipeline solves it. After the semantic segmentation, the clustering algorithm works on object points to further segment each instance. | 2108.09522_1.jpg |
arxivcap_2303.06845_5 | Title: Transformer Encoder with Multiscale Deep Learning for Pain Classification Using Physiological Signals | Caption: The heat stimuli, with a break in between interval and window segmentations. (a) demonstrates the original experiment settings of BioVid, with a duration of 4 seconds for each heat stimulus and an interval of 8 to 12 seconds between each stimulus. The yellow segmentation displays the 5-second timeframe for each collected signal. (b) Thiam introduces a different segmentation in red-strip rectangle which takes 4.5 seconds as opposed to 5.5 seconds. | 2303.06845_5.jpg |
arxivcap_2303.14662_2 | Title: OTAvatar: One-shot Talking Face Avatar with Controllable Tri-plane Rendering | Caption: Qualitative result for cross-identity reenactment. Examples are sampled from the HDTF dataset <cit.>. Both source and driving subjects are not included in the training set. | 2303.14662_2.jpg |
arxivcap_2212.13138_9 | Title: Large Language Models Encode Clinical Knowledge | Caption: Clinician evaluation of comprehension, retrieval and reasoning capabilities (a) Evaluation of correctness and (b) evaluation of incorrectness of reading comprehension, recall of knowledge and reasoning step. The results indicate Med-PaLM shows any evidence of incorrect comprehension only 5.0% of the time. With regard to evidence of correct retrieval of medical knowledge, we found that clinician answers scored 97.8% while Flan-PaLM only scored 76.3%. However, the instruction prompt-tuned Med-PaLM model scored 95.4%, reducing the inferiority of the model compared to clinicians. | 2212.13138_9.jpg |
arxivcap_2210.09436_3 | Title: Analytic Neutrino Oscillation Probabilities | Caption: Probability of $\nu_\mu \to \nu_e$ using the simplified PREM model (<ref>) for the nonunitary, NSI and quasi-Dirac neutrino scenarios. | 2210.09436_3.jpg |
arxivcap_1802.01257_4 | Title: Promoting cooperation by punishing minority | Caption: (Color online) Schematic presentation of the interface that separates cooperators (blue) and defectors (red). The two leading elementary processes that contribute the most to the change of the interface are marked by arrows. The average payoff difference between the two strategies is $P_{C}-P_{D}=3r/2+17\alpha/24-5$, which increases as the punishment fine $\alpha$ increases. | 1802.01257_4.jpg |
arxivcap_1906.05247_2 | Title: Bootstrapping Upper Confidence Bound | Caption: Cumulative regrets for truncated-normal bandit and Gaussian bandit. Sigma is the upper bound on the standard deviation of the noise. The results are averaged over 200 realizations. | 1906.05247_2.jpg |
arxivcap_1503.07275_1 | Title: X-Ray Detection of Transient Magnetic Moments Induced by a Spin Current in Cu | Caption: (a) Timing of current pulses, x-ray pulses and data collection periods, as discussed in the text. (b) Schematic of the x-ray microscopy measurements. The x-ray spot size at the sample was 35 nm and the transmitted x-rays were detected by an avalanche photo diode. Images were recorded by raster scanning of the sample. (c) The sample consisted of a nanopillar of 240 nm diameter containing a ferromagnetic multilayer with perpendicular magnetization direction, as discussed in the text. Current to the pillar was supplied by Au and Ru contact leads, as shown. The current is defined as positive when flowing from Cu to the ferromagnet, corresponding to electron flow in the opposite direction. At the bottom right, we show a representative STXM contrast image revealing the nanopillar, taken at the Cu L$_3$ resonance energy of 932.7 eV. | 1503.07275_1.jpg |
arxivcap_1104.3722_6 | Title: Investigating the Distribution of Password Choices | Caption: Guesswork statistics for Uniform model, Real data and Zipf model. | 1104.3722_6.jpg |
arxivcap_astro-ph_0104294_2 | Title: The Large Area Lyman Alpha Survey | Caption: Keck spectrum of another confirmed Lyman-$\alpha$ source, this time at $z=4.37$. This object has a line flux of $\sim 4 \times 10^{-17} {\rm\,erg\,cm^{-2}\,s^{-1}}$ and a rest frame equivalent width $> 100 \AA$ at the $2 \sigma$ level. The continuum emission from this source is undetected in our deep broadband images and at best marginally detected in the spectrum. In order to accentuate the well-sampled line and suppress noise, the spectrum of this source has been smoothed with a Gaussian filter with a 4 pixel $=19.4$Å full width at half maximum. The lower panel shows $1\sigma$ photon counting errors, suitably adjusted for the smoothing. The weak "line" to the left of Lyman-$\alpha$ is a residual from the subtraction of the $6300$Å sky line. | astro-ph_0104294_2.jpg |
arxivcap_2206.01820_2 | Title: A Robust Backpropagation-Free Framework for Images | Caption: EKDAA forward and backward updates. Forward propagation convolves the current pathway with applicable filters (in red) and then continues on to the next layer (in orange). To propagate error from layer $\ell$ to $\ell-1$, backprop computes a circular convolution of $\mathbf{z}^n$ on $\mathbf{f}^n$ to translate the signal to $\ell-1$. EKDAA uses error map $\mathbf{e}^n$ and computes a circular convolution on $\mathbf{E}^n$ to create the signal pathway for $\ell$. In a layer-wise update, tensors used to continue the signal pathway backwards are shown in blue and their resulting layers are shown in green. | 2206.01820_2.jpg |
arxivcap_1109.6218_6 | Title: Dynamics and efficiency of a self-propelled, diffusiophoretic swimmer | Caption: Rate constant dependence of the swimming speed $U$ for $\Psi(r)=\exp\left[-\left(r-1\right)/\lambda \right]$. See Eq. (<ref>) for the definition of $k_+$. (Dashed line) Reaction induced concentration distortion neglected as explained in the text, leading to $U = K_1/3$ with $K_1 \simeq -0.89$. (Dotted line) Approximation $U_0$ (Eq. (<ref>)) for $\lambda \ll 1$ which takes the concentration distortion into account. Full symbols are numerical results for $k_{\rm ba} =0$ and open symbols are results for $k_{\rm ab} =0$. ($\blacklozenge$) $\lambda=0.005$, $\Pe/\delta =0$. ($\blacksquare$,$\square$) $\lambda=0.1$, $\Pe/\delta =0$. ($\bullet$,$\circ$) $\lambda=0.1$, $\Pe / \delta =10$. ($\blacktriangle$,$\vartriangle$) $\lambda=1$, $\Pe / \delta =0$. | 1109.6218_6.jpg |
arxivcap_1806.08917_6 | Title: Halide-Perovskite Resonant Nanophotonics | Caption: Basic designs of resonant nanostructures made of halide perovskites. Single nanoparticles supporting the Mie resonances (Section 3.1). Nanowires supporting the Fabry-Perot resonances and waveguide modes (Section 3.2). Nanoplates supporting whispering-gallery modes (Section 3.2). Two-dimensioanl arrays of resonant nanoparticles creating either photonic crystals or metasurfaces (Section 3.3). Layer of halide perovskite can be deposited on resonant substrates (Sections 3.3 and 3.4). Resonant non-perovskite nanoparticles combined with a perovskite layer or incorporated into a perovskite slab, respectively (Section 3.4 and Section 4). | 1806.08917_6.jpg |
arxivcap_1906.05247_3 | Title: Bootstrapping Upper Confidence Bound | Caption: Cumulative regrets for Bernoulli bandit and Beta bandit. The results are averaged over 200 realizations. | 1906.05247_3.jpg |
arxivcap_2303.06773_1 | Title: Protecting Classical-Quantum Signals in Free Space Optical Channels | Caption: Transmissivities obtained from the model of the free-space channel following the parameters in Table I. The shaded area represents the standard deviation of the curves. | 2303.06773_1.jpg |
arxivcap_2202.13870_6 | Title: Simulating Network Paths with Recurrent Buffering Units | Caption: Traces produced by (middle) and the (right) models along with Ground-Truth traces (left) for the test protocol TCP Vegas. The blue curves represent sending rates decided by the test protocol and the red curves represent delays generated by the models. For every pair of rows, the first row shows the entire trace and the second row focuses on the first few seconds. These are randomly picked traces, and there is no one-to-one correspondence between columns. | 2202.13870_6.jpg |
arxivcap_1902.04067_6 | Title: Multi-tier Caching Analysis in CDN-based Over-the-top Video Streaming Systems | Caption: A schematic showing an example of how a time-line of a file $i$ can change according to the request pattern of a video file $i$. | 1902.04067_6.jpg |
arxivcap_2305.05344_1 | Title: Trustworthy Multi-phase Liver Tumor Segmentation via Evidence-based Uncertainty | Caption: The insight of trustworthy multi-phase liver tumor segmentation (TMPLiTS). Black box model might only provide the final segmentation result, such naïve result could not interpret the reliability of model explicitly for multi-phase liver segmentation, resulting in the clinicians would make a diagnosis with slight hesitation. Inspired by <cit.>, we argue that uncertainty estimation could reveal the reliability of black box model. Thus, a novel TMPLiTS is proposed, jointly achieving the segmentation and uncertainty estimation. Evidence-based uncertainty is introduced to quantify the uncertainty of prediction comprehensively, while trustworthy results could assist the clinicians to make a reliable diagnosis. | 2305.05344_1.jpg |
arxivcap_2202.01477_4 | Title: Unsourced Random Access with a Massive MIMO Receiver Using Multiple Stages of Orthogonal Pilots | Caption: {The required $E_b/N_0$ as a function of the number of active users in the proposed scheme for $M=50$, $J = 2$, and different values of $n_p$.} | 2202.01477_4.jpg |
arxivcap_1906.05253_1 | Title: Search on the Replay Buffer: Bridging Planning and Reinforcement Learning | Caption: Search on the Replay Buffer: (a) Goal-conditioned RL often fails to reach distant goals, but can successfully reach the goal if starting nearby (inside the green region). (b) Our goal is to use observations in our replay buffer (yellow squares) as waypoints leading to the goal. (c) We automatically find these waypoints by using the agent's value function to predict when two states are nearby, and building the corresponding graph. (d) We run graph search to find the sequence of waypoints (blue arrows), and then use our goal-conditioned policy to reach each waypoint. | 1906.05253_1.jpg |
arxivcap_2005.06188_2 | Title: Star clusters near and far; tracing star formation across cosmic time | Caption: The smoothed projected stellar surface density from the Massive Young Star-Forming Complex Study in Infrared and X-ray <cit.> is shown with a color bar in units of observed stars pc$^{-2}$. The brown contours show the increase in surface density by factors of 1.5. The ellipses mark the core regions of the isothermal ellipsoids. To identify the subclusters and to estimate the model parameters of the young stars a finite mixture model with a maximum likelihood estimation was used. The left panel shows NGC 2264 while the right panel shows the Carina star-forming region. This figure was published as part of Fig. 2 in <cit.>. | 2005.06188_2.jpg |
arxivcap_1709.03144_27 | Title: The 12C(a,g)16O reaction and its implications for stellar helium burning | Caption: (Color online) Percent difference in the carbon-oxygen core mass, $\Delta M_{\rm{CO}}$ (left axis; red circles), and the central carbon/oxygen ratio, $\Delta X_{\rm{c}}$($^{12}$C/$^{16}$O) ratio (right axis; blue circles), between using the $^{12}$C($\alpha$,$\gamma$)$^{16}$O rate of this work and that of <cit.> for 15 M$_{\odot}$ (top panel) and 25 M$_\odot$ (bottom panel) models at core He depletion (defined as when the central He mass fraction falls below $10^{-5}$). The x-axis gives the low, adopted, and high $^{12}$C($\alpha$,$\gamma$)$^{16}$O rate for the respective rate selection. The red arrow along the left axis is the range of $M_\mathrm{CO}$ resulting from the uncertainty in the rate of this work. The blue arrow along the right y-axis is the equivalent range for the central carbon mass fraction. | 1709.03144_27.jpg |
arxivcap_2004.03425_6 | Title: Elemental abundances in M31: [Fe/H] and [{\alpha}/Fe] in M31 Dwarf Galaxies Using Coadded Spectra | Caption: Top: Comparison between the weighted-average abundances with those measured from coadded spectra, for six Milky Way globular clusters. Bottom left: residuals of the difference between the weighted-average abundances and those measured from coadded spectra. The dashed grey lines indicate the mean and standard deviation of the difference. The weighted average and values are determined as described in Section <ref>. Values for the mean and dispersion of the residuals are given in the text. As the offset and dispersion are smaller than the average total uncertainty, we conclude that the measurement from our coaddition method accurately recovers the weighted average of abundance measurements from individual stars in a given coadd. | 2004.03425_6.jpg |
arxivcap_2108.09522_3 | Title: A Technical Survey and Evaluation of Traditional Point Cloud Clustering Methods for LiDAR Panoptic Segmentation | Caption: The demonstration of the panoptic segmentation task and how our pipeline solves it. After the semantic segmentation, the clustering algorithm works on object points to further segment each instance. | 2108.09522_3.jpg |
arxivcap_2205.02281_5 | Title: Three-Body Problem in Modified Dynamics | Caption: The time evolution ( $300~\mathcal{T}$ ) of the equilateral triangle solution for three equal mass galaxies ( all three galaxies with mass $\mathcal{M}$ ) in two dimensions and according to Newtonian and MOD models with exactly the same initial conditions, i.e. initial positions and velocity. The coordinates $x$ and $y$ are in units of $\mathcal{D}$. Please note that due to the overlap of the Newtonian orbits on the right hand side, the trajectories of $m_{1}$ and $m_{2}$ are almost indistinguishable. | 2205.02281_5.jpg |
arxivcap_2012.05350_2 | Title: Automatic Diagnosis of Malaria from Thin Blood Smear Images using Deep Convolutional Neural Network with Multi-Resolution Feature Fusion | Caption: Convolutions with kernel size of $(3\times 3)$ with different dilation rate covering different receptive areas. (a) dilation rate = 1, (b) dilation rate=2, and (c) dilation rate=3. | 2012.05350_2.jpg |
arxivcap_2210.09459_2 | Title: Extensible Proxy for Efficient NAS | Caption: Illustration of the validation losses of two architectures on downstream task(left). A sophisticated few-shot proxy (right) can reflect the actual performance of architectures. | 2210.09459_2.jpg |
arxivcap_1912.05084_16 | Title: Bayesian Copula Density Deconvolution for Zero-Inflated Data in Nutritional Epidemiology | Caption: =10pt Results for the EATS data sets with sample size $n = 965$, $q=2$ episodic components, milk and whole grains, and $p=2$ regular components, sodium and energy, each subject having $m_{i}=4$ replicates. These results correspond to the final MCMC iteration but are representative of other iterations in steady state. The panels show the scatterplots of scaled `residuals' ($\wh\epsilon_{\ell,i,j-1}$ vs $\wh\epsilon_{\ell,i,j}$, $j=2,3,4$) at adjacent sampling occasions and their estimated correlation coefficients for milk (top left), whole grains (top right), sodium (bottom left) and energy (bottom right). | 1912.05084_16.jpg |
arxivcap_1111.0261_6 | Title: Extended Lagrangian free energy molecular dynamics | Caption: (Color online) The change in total free energy as a function of time for a periodic 8 atom silicon cell for conventional Born-Oppenheimer MD as implemented in VASP and extended Lagrangian free energy MD. Both simulations using a time step of 0.25 fs, and a energy convergence criteria of 0.5 meV. The conventional Born-Oppenheimer MD show a clear unphysical systematic drift in comparison to the extended Lagrangian free energy MD. | 1111.0261_6.jpg |
arxivcap_1802.01257_6 | Title: Promoting cooperation by punishing minority | Caption: (Color online) Stationary density of the four types of groups as a function of the punishment fine $\alpha$ when the multiplication factor $r=3.7$. The stationary density of $G_{cc}$ ($G_{dd}$) groups increases (decreases) as $\alpha$ increases. The stationary density of $G_{cd}$ and $G_{dc}$ groups are maximized at moderate values of $\alpha$. Note that for $\alpha<0.25$, the density of $G_{cd}$ groups is lower than that of $G_{dc}$ groups. Conversely, the density of $G_{cd}$ groups is higher than that of $G_{dc}$ groups when $\alpha>0.25$. | 1802.01257_6.jpg |
arxivcap_2206.01826_1 | Title: The Gamma Generalized Normal Distribution: A Descriptor of SAR Imagery | Caption: Plots of the GGN density function for some parameter values: $\mu=0$, $\sigma=1$, and { $s\in\{16, 8 , 4 , 2, 1, 1/2 \}$ (dash-dotted, solid, dotted, dashed, dash-dotted, solid). } | 2206.01826_1.jpg |
arxivcap_2303.14662_3 | Title: OTAvatar: One-shot Talking Face Avatar with Controllable Tri-plane Rendering | Caption: Qualitative comparison of decoupling-by-inverting training and joint training. Joint training cannot preserve identity information in one-shot avatar construction. | 2303.14662_3.jpg |
arxivcap_2202.13870_7 | Title: Simulating Network Paths with Recurrent Buffering Units | Caption: Traces produced by (middle) and the (right) models along with Ground-Truth traces (left) for the test protocol TCP LEDBAT. The blue curves represent sending rates decided by the test protocol and the red curves represent delays generated by the models. For every pair of rows, the first row shows the entire trace and the second row focuses on the first few seconds. These are randomly picked traces, and there is no one-to-one correspondence between columns. | 2202.13870_7.jpg |
arxivcap_1906.05253_2 | Title: Search on the Replay Buffer: Bridging Planning and Reinforcement Learning | Caption: Simple 2D Navigation: Two simple navigation environments. An agent that combines a goal-conditioned policy with search is substantially more successful at reaching distant goals in these environments than using the goal-conditioned policy alone. A standard goal-conditioned policy (top) fails to reach distant goals. Applying graph search on top of that same policy (bottom) yields a sequence of intermediate waypoints (yellow squares) that enable the agent to successfully reach distant goals. | 1906.05253_2.jpg |
arxivcap_1902.04067_8 | Title: Multi-tier Caching Analysis in CDN-based Over-the-top Video Streaming Systems | Caption: An illustration of our system model for video content delivery, consisting of a datacenter, four cache servers ($m=4$), and $2$ edge routers. $d_j$ and $F_j$ parallel connections are assumed between datacenter and cache server $j$, and datacenter and edge router, respectively. | 1902.04067_8.jpg |
arxivcap_2303.06773_2 | Title: Protecting Classical-Quantum Signals in Free Space Optical Channels | Caption: Fidelity as a function of $g$, each curve corresponds to a specific set of values of $[T_1, T_2, T_3]$. The red dots mark the value of $g=g_\mathrm{opt}$ corresponding to the maximum fidelity, and each label denotes the values of the set $[T_1, T_2, T_3]$ corresponding to the adjacent curve. | 2303.06773_2.jpg |
arxivcap_1109.6218_8 | Title: Dynamics and efficiency of a self-propelled, diffusiophoretic swimmer | Caption: The swimming efficiency $\epsilon$ as a function of $\lambda$ with constant $\Pe/\delta$ for $\Psi(y) = \exp\left[-(r-1)/\lambda \right]$ and $k_{\rm ab}=k_{\rm ba}=1$. (Dashed lines) Asymptotic results calculated from Eq. (<ref>). ($\triangle$) $\Pe/\delta=10$. ($\lozenge$) $\Pe/\delta=1$. ($\square$) $\Pe/\delta=0.1$. ($\circ$) $\Pe/\delta=0.01$. | 1109.6218_8.jpg |
arxivcap_1806.08917_7 | Title: Halide-Perovskite Resonant Nanophotonics | Caption: {Optical properties of halide perovskites.} {(a) Photoluminescence peak energy and FWHM for pure chloride, bromide, and iodide perovskite thin films in the cubic phase.} {(b)} Refractive indices $n$ and {(c)} extinction coefficients $k$ of three commonly used perovskite composites: MAPbCl$_3$, MAPbBr$_3$, and MAPbI$_3$, shown together with the corresponding results for SiO$_2$, Si, and GaAs, for comparison. The data for the perovskites are obtained by Leguy {et al.} through single crystal ellipsometry and from Refs. <cit.> for SiO$_2$, Si, and GaAs. These data are collected from Refs. <cit.>. | 1806.08917_7.jpg |
arxivcap_2202.00203_13 | Title: Constraining low redshift [CII] Emission by Cross-Correlating FIRAS and BOSS Data | Caption: Simulations of fits to the CIB and amplitude for two different types of FIRAS models. The red contours show simulations that use the parametric FIRAS model described in Section <ref>. The blue contours show simulations that use the empirical FIRAS model, where the unbinned FIRAS data are used directly as the FIRAS model. The left and right plots show the and simulations, respectively. | 2202.00203_13.jpg |
arxivcap_2202.01537_2 | Title: Bending Graphs: Hierarchical Shape Matching using Gated Optimal Transport | Caption: The dense matching result of our model on TOSCA dataset. Our method is able to learn robust features that work on an unseen class. a) target shape. b) source shape with the predicted correspondences c) ground-truth correspondences, d) error map. | 2202.01537_2.jpg |
arxivcap_1111.0261_7 | Title: Extended Lagrangian free energy molecular dynamics | Caption: (Color online) The fluctuations for the total nuclear and potential energy, $E_K + U$, in comparison to the total free energy, $E_K +U +T_e S$, of a periodic 8 atom silicon system. The simulation was carried out using extend Lagrangian free energy MD in VASP. Using a time step of 0.25 fs and energy convergence criteria of $5\times 10^{-10}$ eV with ultrasoft pseudopotentials. Both the nuclear and electronic had an average temperature of about 500 K. | 1111.0261_7.jpg |
arxivcap_2212.13138_10 | Title: Large Language Models Encode Clinical Knowledge | Caption: Clinician evaluation of answers contents (a) Presence of incorrect content and (b) presence of missing content in the answers. Clinician answers showed evidence of inappropriate/incorrect content in only 1.4% of the cases, compared to 16.1% for Flan-PaLM. Surprisingly, Med-PaLM seemed to further degrade performance, with 18.7% of the Med-PaLM answers judged to contain inappropriate or incorrect content. On the missing content axis, while Flan-PaLM answers were judged to miss important information 47.6% of the time, the number improved significantly for Med-PaLM with only 15.1% of the answers adjudged to have missing information, reducing the inferiority to clinicians whose answers were judged to have missing information in only 11.1% of the cases. | 2212.13138_10.jpg |
arxivcap_2305.05344_6 | Title: Trustworthy Multi-phase Liver Tumor Segmentation via Evidence-based Uncertainty | Caption: The qualitative comparisons against the three perturbations. The first column illustrates the three examples of the difference perturbations, including $\text{\textit{Re}}_\text{noise}$, $\text{\textit{Re}}_\text{blur}$, and $\text{\textit{Re}}_\text{miss}$. The variance coefficients of these perturbations are $\sigma^{2}_{n} = 0.1$, $(\sigma^{2}_{b}, k) = (10, 13)$, and $\Omega=1$, respectively. The results illustrate the reliability of TMPLiTS qualitatively. | 2305.05344_6.jpg |
arxivcap_2005.06188_3 | Title: Star clusters near and far; tracing star formation across cosmic time | Caption: This figure shows the larger area around the YSC NGC 3603 showing O-star candidates with a high probabilty to be ejected from the cluster core. Objects coloured in red are those with a high ejection probability while the two stars marked in blue have a trajectory consistent with a reduced probability to originate from the star cluster. The lines mark the direction of movement. This figure was published as Fig.7 in <cit.>. | 2005.06188_3.jpg |
arxivcap_astro-ph_0104286_1 | Title: THE NEARBY FIELD GALAXY SURVEY | Caption: Overview of the global properties of the selected galaxy sample. Presented are the number distributions as a function of: {a)} absolute $B$ magnitude; {b)} numerical galaxy type (note that we choose to place the unclassified early-type galaxies at $T=-4$ (cD) rather than $-7$, as we do not have any cD galaxies in our sample, and this placing seems more natural with respect to the compact ellipticals at $T=-6$); {c)} the surface brightness at the effective radius in $B$; and {d)} the effective $B-R$ colour, measured within that effective radius. In panels {c)} and {d)} the open histograms represent the selected sample, and the shaded ones the subsample of galaxies with types later than S0/a. | astro-ph_0104286_1.jpg |
arxivcap_2202.13870_8 | Title: Simulating Network Paths with Recurrent Buffering Units | Caption: Traces produced by (middle) and the (right) models along with Ground-Truth traces (left) for the test protocol TCP NewReno. The blue curves represent sending rates decided by the test protocol and the red curves represent delays generated by the models. For every pair of rows, the first row shows the entire trace and the second row focuses on the first few seconds. These are randomly picked traces, and there is no one-to-one correspondence between columns. | 2202.13870_8.jpg |
arxivcap_2108.09628_1 | Title: DisenKGAT: Knowledge Graph Embedding with Disentangled Graph Attention Network | Caption: An example of the "Kobe Bryant" and "Bill Gates" neighboring distribution, where different color depicts a "topic" or "cluster". It is clear to figure that an entity always plays multiple roles in different scenarios. | 2108.09628_1.jpg |
arxivcap_1709.03144_28 | Title: The 12C(a,g)16O reaction and its implications for stellar helium burning | Caption: (Color online) Top panel: mass-averaged mass fractions of the weak $s$-process elements in the central 2 M$_\odot$ of the 15 M$_\odot$ models at the end of core He-burning using the $^{12}$C($\alpha$,$\gamma$)$^{16}$O rate of this work and that of <cit.>.
Bottom panel: percentage difference in weak $s$-process production at the end of core He-burning as both total elemental masses and mass-averaged mass fractions for the 15 M$_\odot$ models using the rates of this work and <cit.>. | 1709.03144_28.jpg |
arxivcap_2210.09459_3 | Title: Extensible Proxy for Efficient NAS | Caption: a: Comparison with efficient proxies and early stopping methods on NAS-Bench-Trans-Micro Autoencoder task. It shows the effectiveness of DPS compared with early stopping methods on either the target task or a classification task when evaluating 4096 architectures. b, c: On NDS DARTS-ImageNet task, Eproxy and Eproxy+DPS (Searched on DARTS-CIFAR-10, transferred to ImageNet) achieve 0.51, 0.85 $\rho$ respectively. It shows DPS can find a search-space-aware Eproxy. | 2210.09459_3.jpg |
arxivcap_1802.01257_7 | Title: Promoting cooperation by punishing minority | Caption: (Color online) Snapshots of typical distributions of cooperators (blue) and defectors (red) at different time steps. To facilitate observe, we use a smaller system size, that is, $200\times 200$ lattices. The multiplication factor $r=3.7$. The punishment fine $\alpha$ for (a)-(d) and (e)-(f) is 0 and 0.4 respectively. For $\alpha=0$, cooperators gradually die out. For $\alpha=0.4$, initially the number of cooperators greatly reduces and the survival cooperators form some small clusters. With time the cooperator clusters continue to expand and finally occupy the whole system. | 1802.01257_7.jpg |
arxivcap_1109.6218_9 | Title: Dynamics and efficiency of a self-propelled, diffusiophoretic swimmer | Caption: The swimming efficiency $\epsilon$ as a function of $\Pe/\delta$ with constant $\lambda$ for $\Psi(y) = \exp\left[-(r-1)/\lambda \right]$, $k_{\rm ab}=k_{\rm ba}=1$. (Dashed line) Asymptotic result for $\lambda=0.01$ calculated from Eq. (<ref>). ($\blacksquare$) $\lambda=10$. ($\triangle$) $\lambda=1$. ($\circ$) $\lambda=0.1$. ($\blacktriangle$) $\lambda=0.01$. | 1109.6218_9.jpg |
arxivcap_2303.06773_3 | Title: Protecting Classical-Quantum Signals in Free Space Optical Channels | Caption: Total fidelity for different values of erasure rate. Solid lines represent transmission via the protocol, and dashed lines correspond to direct transmission. The ancilla entangled state uses a squeezing value of $r=10$ dB. | 2303.06773_3.jpg |
arxivcap_2004.03425_9 | Title: Elemental abundances in M31: [Fe/H] and [{\alpha}/Fe] in M31 Dwarf Galaxies Using Coadded Spectra | Caption: [Fe/H] distributions for our ten M31 dSphs roughly ordered from most massive to least massive (left column, top to bottom, then right column, top to bottom). The filled grey histogram represents the individual measurements from both shallow and deep (low and high S/N) data. The colored histograms represent the measurements made from coadded spectra, where the solid line denotes the validation coadds, and the dashed line indicates the science coadds. The stellar mass of each dSph is indicated in its respective panel. | 2004.03425_9.jpg |
arxivcap_1806.08917_8 | Title: Halide-Perovskite Resonant Nanophotonics | Caption: Resonant halide-perovskite nanoparticles. (a) {Analytically calculated total scattering efficiency $Q_{scat}$ of a spherical MAPbBr$_3$ nanoparticle in vacuum. Overlaid lines depict approximate dispersion of the electric dipole, magnetic dipole, and magnetic quadrupole resonances of the particle. The arrow indicates position of MAPbBr$_3$ exciton.} (b) Calculated near-field distributions for {single MAPbI$_3$ nanoparticles supporting the magnetic dipolar and quadrupolar resonances at emission wavelength}.{Adapted with permission from <cit.>. Copyright 2018 American Chemical Society.} (c) PL enhancement for a nanoparticle at magnetic quadrupole resonance {(red curve)}, as compared with thin a film {(black curve)} and non-resonant nanoparticle {(green curve)}. {Adapted with permission from <cit.>. Copyright 2018 American Chemical Society.} (d) Lasing modes in CVD-grown CsPbBr$_3$ submicron particles of different sizes.{Reprinted with permission from <cit.>. Copyright 2018 American Chemical Society.} (e) Lasing from perovskite submicron {CsPbX$_3$ spheres with different halogens X: Cl,Br, I and mixed ones}. {Reprinted with permission from <cit.>. Copyright 2018 American Chemical Society.} | 1806.08917_8.jpg |
arxivcap_2303.14662_5 | Title: OTAvatar: One-shot Talking Face Avatar with Controllable Tri-plane Rendering | Caption: Overview of OTAvatar. OTAvatar contains a 3D face animator network and a latent code decoupling strategy, namely decoupling-by-inverting. The 3D face animator is composed of a pre-trained 3D face generator<cit.> and a motion controller module. The decoupling-by-inverting algorithm is an optimization-based image inversion that can decouple the latent code $\mathbf{w}$ into identity code $\mathbf{w}_{id}$ and motion code $\mathbf{w}_x$. When the model is well-trained, the motion-free identity code can be inferred from a single reference image, and an avatar of an unseen subject can be constructed. The identity code can be integrated with any other motion code predicted by the controller to animate the identity with desired motion. | 2303.14662_5.jpg |
arxivcap_2305.05349_3 | Title: Towards the Characterization of Representations Learned via Capsule-based Network Architectures | Caption: Number of top-$k$ units exclusive for each class (left) and shared w.r.t all the other classes (right). | 2305.05349_3.jpg |
arxivcap_1912.05084_17 | Title: Bayesian Copula Density Deconvolution for Zero-Inflated Data in Nutritional Epidemiology | Caption: =10pt Exploratory a data set simulated according to the process detailed in Section <ref> in the Supplementary Material with sample size $n = 1000$, $q=2$ episodic components and $p=1$ regular components, each subject having $m_{i}=3$ replicates. Left panels: histogram of recalls $Y_{\ell,i,j}$ (red) and histogram of strictly positive recalls $Y_{\ell,i,j}(>0)$ (blue) superimposed on each other; middle panels: subject-specific means $\overline{Y}_{\ell,i}$ vs subject-specific variances $S_{Y,\ell,i}^{2}$ when multiple strictly positive recalls are available; right panels: box plots of proportion of zero recalls vs corresponding subject-specific means $\overline{Y}_{\ell,i}$. | 1912.05084_17.jpg |
arxivcap_1503.07275_4 | Title: X-Ray Detection of Transient Magnetic Moments Induced by a Spin Current in Cu | Caption: (a) Schematic density of states (DOS) without current flow ($E_\mathrm{F}\!=\!\mu^\uparrow\!=\!\mu^\downarrow$) for itinerant $s$-$p$ and localized $d$ spins for bulk Co and Cu, and Cu interface atoms. The exchange-split $d$ DOS of the Co and magnetic Cu interface atoms exhibits occupied (blue) and unoccupied minority spin states (yellow), which is absent in bulk Cu. The static XMCD effect arises from transitions to the yellow shaded unoccupied $d$ states. (b) Model of the spin dependent electron chemical potentials in the presence of electron spin flow from Co to Cu across an abrupt interface without interface states. The diagram corresponds to the shown Co magnetization direction and +5mA current as in Fig. <ref> (c). The spin averaged chemical potentials $\bar{\mu}_\mathrm{Co}$ in Co and $\bar{\mu}_\mathrm{Cu}$ in Cu are shown in red. The chemical potentials decay exponentially with distance from their maximum values $\mu^\uparrow(0)$ and $\mu^\downarrow(0)$ at the interface. The origin of the transient XMCD effect is discussed in the text. | 1503.07275_4.jpg |
arxivcap_2202.00203_14 | Title: Constraining low redshift [CII] Emission by Cross-Correlating FIRAS and BOSS Data | Caption: The redshift diagonal of the binned angular power spectrum, $C_b(z=z')$, for both the CMASS and LOWZ redshift regions. The three angular bins, centered at $\ell=24$, $\ell=33$, and $\ell=42$ are plotted in red, green, and blue, respectively. The best-fit models are plotted as solid lines, and the data are plotted as triangles or circles, with error bars computed from the best-fit model. The $\ell=24$ data are artificially shifted backward by half a redshift bin for visual clarity. Dashed lines show the thermal noise portion of the best-fit model only. From the dashed lines, the effect of the FIRAS beam and scan convolution can be seen, as the thermal noise increases from low to high $\ell$. Only the $\ell=24$ bin has a significant foreground contribution. | 2202.00203_14.jpg |
arxivcap_2212.13138_11 | Title: Large Language Models Encode Clinical Knowledge | Caption: Clinician evaluation of potential harm from answers (a) Extent of possible harm and (b) likelihood of harm. While 29.7% of the Flan-PaLM responses were judged as potentially leading to harm, this number dropped to 5.9% for Med-PaLM on par with clinician-generated answers which were also judged as potentially leading to harm in 5.7% of the cases. | 2212.13138_11.jpg |
arxivcap_1111.0261_8 | Title: Extended Lagrangian free energy molecular dynamics | Caption: (Color online) The error amplitudes as a function of time step for the total nuclear and potential energy, $E_K + U$, and the total free energy, $E_K +U +T_e S$, of a 8 atom periodic silicon cell | 1111.0261_8.jpg |
arxivcap_2206.11975_3 | Title: Background-penalty-free waveguide enhancement of CARS signal in air-filled anti-resonance hollow-core fiber | Caption: (a) The CARS signal of nitrogen at 2331 cm$^{−1}$ (Q-branch vibrational transition) in a 2.96 cm long air- filled AR-HCF compared to the CARS signal from the ambient air (no fiber present). Inset: enlarged plot in the lower CCD count region for better resolution. The spikes in the region between 2400 cm$^{-1}$ and 2800 cm$^{-1}$ only appear at very short acquisition times (here 0.05s). This is because of the way that the CCD spectrometer operates. (b) Pump power dependent CARS signal measurement. Since the pump beam is used as both pump and the probe beams, the quadratic dependence of the CARS signal is expected. | 2206.11975_3.jpg |
arxivcap_2005.06188_4 | Title: Star clusters near and far; tracing star formation across cosmic time | Caption: The Galactic young massive star cluster Westerlund 2 as it is seen by the integral field spectrograph MUSE. The image represents H$\alpha$, NII$\lambda5583$, and [OIII$\lambda5007$] in red, green, and blue, respectively. This image was similarily published in <cit.>. | 2005.06188_4.jpg |
arxivcap_1211.6964_1 | Title: On the Bauschinger effect in supercoooled melts under shear: results from mode coupling theory and molecular dynamics simulations | Caption: Linear plot of the stress-strain relation $\sigma(\gamma)$ at fixed density and shear rate for various flow histories: startup flow with constant shear rate (red), flow reversal in the steady state (s, purple), from the elastic regime (el, blue), and from the point of the stress overshoot (max, green). The linear increase with the quiescent elastic modulus $G_\infty$ in startup, and with $G_\text{eff}$ after shear-reversal are indicated by dotted lines. Characteristic stress-strain points, whose transient modulus is discussed in the text, are marked by symbols: The red square gives the stationary stress. It corresponds to an integration along the line from A to B in Fig. <ref>. Triangles mark stresses after flow reversal in the linear regime and correspond to an integration along the line C' to E'. Discs mark stresses after flow reversal in the overshoot regime and are obtained via integrations along the line C to E in Fig. <ref>. | 1211.6964_1.jpg |
arxivcap_1902.04067_10 | Title: Multi-tier Caching Analysis in CDN-based Over-the-top Video Streaming Systems | Caption: A flowchart illustrates the online updates for an edge-cache when a file $i$ is requested at time $t_i$. Here, $t_{f,l_t}$ represents the time of the last request of file $i$, and $\mathcal{H}$ is the index set of all video files in the edge-cache. | 1902.04067_10.jpg |
arxivcap_2202.13752_2 | Title: Discrete unified gas kinetic scheme for the conservative Allen-Cahn equation | Caption: The phase-field contour ($\phi=0$) of diagonal translation of a circular interface obtained by (a) DUGKS-AC, (b) DUGKS-I and (c) DUGKS-II with different reconstruction schemes at $\text{Pe}=60$. The reconstruction schemes are 2CDI, 4CDI, WENO-Z3 and WENO-Z5 from left to right. Black solid line denotes the initial profile and red dashed line represents the reconstructed interface at 10T. | 2202.13752_2.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.