nopperl/arxiv-image-text · Datasets at Hugging Face

dedup-isc-ft-v107-score float64 0.32 1	uid stringlengths 32 32	text stringlengths 0 32.5k	paper_id stringlengths 1 14	original_image_filename stringlengths 5 222
0.378794	d26d79b11ed3443ea6a39aaf74a6cfd0	The first two panels from the left illustrate the different formation paths described in section [sec:formation_channels]3.1. In the direct physical collision scenario, the dominant process by which the principal object increases in mass is through physical collisions, as is shown by the initial binary-binary interaction that leads to the merger of two stars. In contrast, in the middle, for the formation path labeled binary coalescence, the main increase in mass comes from binary coalescence.The right-most column illustrates the collision history of the most massive IMBH formed in our simulations.	2012.10497	tree.jpg
0.464724	679aec000b89491d89931845bdb479ab	Specific heat capacity of NdCoGe_3 for (a) magnetic field within the basal plane and (b) field applied along the c axis.	2012.10499	Cp1.jpg
0.543801	d7f70238ad7d4e4b8ab2a685ff2931d9	Zero-field data revealing magnetic phase transitions at T_N1 = 3.70 K and T_N2 = 3.50 K in NdCoGe_3 indicated by the vertical dashed lines. (a) Specific heat capacity and (b,c,d) the in-plane ac magnetic susceptibility measured in zero applied DC field. The ac data are shown as the (b) in-phase contribution χ' with (c) displaying the derivative of χ', and (d) contains the out-of-phase contribution χ”.	2012.10499	Figure1.jpg
0.497287	cd469f8ad55240808cba91c5442c6d1e	(a,c,e) Neutron diffraction intensity as a function of h, k or l at various temperatures and (b,d,f) the fitted position of h, k, and l as a function of temperature. The red lines in (a,c,e) are the corresponding Gaussian fit(s) and the fitted error bars in (b,d,f) are smaller than the size of the symbols. In (a), the small shoulders for smaller h are artifacts from the sample analyzer collimator.	2012.10499	Figure3.jpg
0.497347	8ae2fc5a5ea54ee08f6e69c1a84ad08c	Temperature-dependent magnetization in zero-field-cooled warming (red) and field-cooled cooling (blue) measurements with an applied field (a) H ⊥ c and (b) H\|\|c. The inset in (a) shows the magnetization to T = 0.4 K using the same units as the main panel.	2012.10499	Figure4.jpg
0.415548	3dd2f767d1db43b78e4e48ac8e416a97	Isothermal magnetization M(H) and derivative of isothermal magnetization dM/dH of NdCoGe_3 with an applied field H⊥c. (a) M(H) (left axis) and dM/dH (right axis) at T = 0.4 K (b) The in-plane ac magnetic susceptibility χ' as a function of applied DC field at T = 0.4 K (c) dM/dH from 1.8 K to 2.8 K with a temperature step of Δ T = 0.04 K. (d) dM/dH at T = 3 K.	2012.10499	Figure5.jpg
0.473994	288057c4e3104d7bb4181e037434fdfc	(a,b) Isothermal magnetization data (left axis) and derivative dM/dH (right axis) on increasing field (red line) and decreasing field (blue line) of NdCoGe_3 with an applied field H⊥c.	2012.10499	Figure6.jpg
0.421257	f2fadf026cc34d79aa26a9da0edf7240	Neutron powder diffraction pattern of NdCoGe_3 at T = 1.8 K < T_N2 . Inset: Data for a limited Q range revealing magnetic reflections at 1.8 K (bottom) through a comparison to the pattern obtained at 40 K (top) for which the Bragg peaks are entirely from the crystal lattice.	2012.10499	NeutronPowder.jpg
0.410236	387e066b43d04a809a4d89a19710903c	Magnetic phase diagrams of NdCoGe_3 as inferred from anomalies in the specific heat, derivatives of temperature-dependent and isothermal magnetization as indicated in the legends for (a) H⊥c and (b) H\|\|c. The black dotted line in (a) indicates the presence of a first order phase boundary based upon observations of thermal or field hysteresis.	2012.10499	Phase.jpg
0.445534	e726616ccac54a5a9af51274485f77cd	Neutron single crystal diffraction data for NdCoGe_3 from the WAND^2. A section of the reciprocal h0l plane between 00l and -10l with the magnetic reflections close to the -0.5 0 l position. The color scale is adjusted to show both the weak magnetic intensity and the stronger nuclear scattering. The rings originate from the Al-sample holder. Purple arrows indicate the propagation vector from the crystallographic -103 and 004 reflections.	2012.10499	Wand2_new.jpg
0.449911	13622d1a9f814824a85273c24bc5575f	Candidate magnetic structures resulting from (a) 1k⃗, (b) 2k⃗, (c) 3k⃗ and (d) 4k⃗ models. The vector moments (blue arrows) on Nd atoms (blue spheres) are shown for a segment of the magnetic unit cell; the nuclear (paramagnetic) unit cell is outlined in grey, the Co atoms are red and Ge atoms are grey. The various models describe the neutron powder diffraction data at 1.8 K equally well.	2012.10499	mag_str.jpg
0.413801	95c2b092a1e64911a5efcba08ff64bfa	Location of the FASERν detector and event topology. Top: The FASER experiment is placed about 500 m downstream of the ATLAS interaction point in the previously unused side tunnel TI12, which connects the SPS with the LHC tunnel. Center: The detector is centered around the beam collision axis where the neutrino flux is maximal. It consists of the FASERν emulsion neutrino detector, followed by a magnetized spectrometer and a calorimeter. Bottom: The emulsion detector consists of tungsten plates interleaved with nuclear emulsion films. Both interactions of neutrinos and neutral hadrons lead to the appearance of a neutral vertex at which several charged particles emerge. Different types of events can be distinguished based on the event topology, as explained in the text.	2012.10500	Figure.jpg
0.517272	776b9d59f2344eb09fbc35259b8c168f	Comparison of 3 observables for NH interactions generated using , , and and NC interactions using , and . Top: NH interactions at 60, 300. Middle: Different NHs (mesons and baryons) generated via at 300. Bottom: NC interactions at 100, 1 generated via , and . See text for more details.	2012.10500	appendix_fig.jpg
0.429211	c674bcc9d693488ebece1fa7a7245cc5	Left: Correlation matrix showing the linear relationship between observables presented in <ref>, the incoming particle energy (E_beam) and the event type (1 for NC neutrino interactions, 0 for neutral hadrons). Right: Permutation feature importance (the normalized mean score decrease for each of the observables) for the signal identification classifier (blue) and neutrino energy estimator (red) network. We use accuracy as the score metric for the classifier network and mean average percents error for the estimator. Scores decreases are normalized so that they sum to 1.	2012.10500	corr_perm_importance.jpg
0.385364	be11346bd4b049aab3847874a59d5d15	Left: Signal selection efficiencies as a function of beam energy. Each line indicates the fraction of events passing the following criteria sequentially: i) neutral vertex identification (blue) requiring ≥ 5 charged tracks, ii) lepton veto (green) requiring no electron candidate and no non-interacting charged track, and iii) signal identification (red) as performed by the NN classifier. The dashed black line shows the combined efficiency. Right: The energy spectra of particles interacting within the FASERν detector. We show the expected numbers of neutral hadron interactions (green) and NC neutrino interactions (red) with the FASERν detector during LHC Run 3 as dashed lines. The solid lines show the spectra for events passing the signal selection (including neutral vertex identification, lepton veto and signal identification). The uncertainty associated to the background generation is shown as a shaded band.	2012.10500	efficiency_intrct_spectrum.jpg
0.393763	063b20839c7140338a9c2fcb9587e72d	Left: Neutrino energy reconstruction for NC neutrino interaction events obtained by a neural network-based multivariate analysis using the observables defined in <ref>. Right: Relative RMS energy resolution using the neural network-based multivariate analysis (solid) and only the hadronic energy of the events (dashed).	2012.10500	energy2D_resolution.jpg
0.458506	7257e621190c4eecbeb1c634f4685840	Expected energy spectrum of neutral hadrons interacting within the FASERν detector during LHC Run 3 with 150 of luminosity.	2012.10500	hadspec_072720_neutral.jpg
0.495289	e46949d501bf4c59b2511b6af8e1d293	Normalized kinematic distributions for the observables defined in <ref>. The dashed lines show the distributions obtained with for the NC neutrino interaction signal at incoming neutrino energies of E_ν=100 (blue) and E_ν=1 (red). The solid green lines correspond to the distributions for the neutral hadron interactions simulated with for the expected energy spectrum presented in <ref>. The shaded region shows the range of predictions for the background distributions obtained from different generators: , and .	2012.10500	obs.jpg
0.442296	e36519aa97ee4818a90ee63d7d142c74	Left: Stacked histogram of events passing the event signal selection described in <ref> as function of the reconstructed energy for LHC Run 3 with 150 integrated luminosity. The red and green shaded regions show the NC neutrino interaction signal and the neutral hadron interaction background, respectively. The hatched region indicates the uncertainty arising from the simulation of neutral hadron interactions, corresponding to the range of predictions obtained by three different generators. Right: FASERν's estimated neutrino-tungsten NC cross section sensitivity. Existing constraints are shown in gray. The black dashed curve is the theoretical prediction for the DIS cross section, averaged over neutrinos and anti-neutrinos, per tungsten nucleus. The inner red error bars correspond to statistical uncertainties, the blue error bars additional take into account uncertainties associated with the simulation of the background, and the outer green error bars show the combined uncertainties with the neutrino production rate (which corresponds to the range of predictions obtained from different MC generators as obtained in Ref. <cit.>).	2012.10500	signal_spectrum_crosssection.jpg
0.483784	c2cc8de2259e45aca2831214faf5470c	Spectra of radiation from sidewalls of the accretion columns, calculated simultaneously with velocity and temperature profiles plotted in Figs <ref> and <ref>. The results correspond to different sets of the model parameters: `1' (solid line), `2' (dashed), `3' (dot-dashed), `4' (dot-double-dashed), `5' (double-dot-dashed), `6' (dotted).	2012.10501	figsp1.jpg
0.455476	676006c692dd46b295d0a3b315e4fade	(a) Electron data (in blue) from left to right: SUMER on-disk “correlation heights” in open funnels <cit.>, SUMER off-limb temperatures from (; dark blue) and (; light blue), UVCS <cit.>, PSP <cit.>, and Helios/Ulysses <cit.>. Proton data (in red) from left to right: UVCS <cit.>, PSP (dark red), Helios (light red), Ulysses (dark red), and Voyager (light red) high-speed wind data. Oxygen-ion data (in green) from left to right: SUMER off-limb kinetic temperatures <cit.>, UVCS <cit.>, and Wind <cit.>. (b) From left to right, O^+5 and proton anisotropy data from UVCS <cit.>, PSP data with v ≥ 450 km s^-1 <cit.>, Helios data with v ≥ 600 km s^-1 <cit.>, and Ulysses data with v ≥ 600 km s^-1 <cit.>. All black curves are solely to guide the eye.	2012.10509	cranmer_temps2020_fig1.jpg
0.459751	0a57cfccc5fd4e298dcbd19763830963	Comparison of the HL (purple) and HM (dark blue) methods to the modified LCHO-CI method for evaluating J. The system is a double QD in Si/SiO_2 described by a quartic potential, where l_0= 6 nm. a) Probability densities of the ground single-electron orbital states for the HL/HM and LCHO-CI calculations at different dot separations. At small dot separations, the HL/HM states overestimate the separation of the two localized wave functions compared to the numerically calculated orbitals. As dot separation increases, the approximate HL/HM states converge to the numerical results. b) J versus the interdot separation 2d is plotted. Three different LCHO-CI calculations are done for N=2 (light blue), N=4 (green), and N=10 (yellow). For all LCHO-CI calculations, M=15^2.	2012.10512	F_HL_HM_comp.jpg
0.430391	49ce28a13a74468ba463c69ddd820f0e	Convergence of J with respect to the number of single-electron orbitals N when M=16^2. The device parameters are V_p1= V_p2= 0.15 V, V_t= 0.09455 V, D_x= D_y= 40 nm, D_t= 20 nm, and T= 15 nm. Inset: The convergence of J with respect to M when N=18 (star in the main figure).	2012.10512	F_Jconvergance.jpg
0.403861	f490e4da394046ef9c9b58455438b9b3	Using harmonic orbitals (HO) to approximate single-electron states. a) The operator Â transforms between the HO basis {\|ϕ_i⟩} and the approximate single-electron orbitals \|ξ_j'⟩. The first three single-electron orbitals for a quartic potential are shown at left, and the eight lowest energy HO states are shown at right. b) Convergence of the first twelve approximate single-electron energies ϵ_j' versus the number of HOs, M, used to compose the basis {\|ϕ_i^ω⟩}. Here, M = M_xM_y and M_x=M_y are the numbers of 1-dimensional HOs taken along the x and y axes, respectively, to construct the 2-dimensional HOs.	2012.10512	F_SEconvergance.jpg
0.532902	ddd86be3f3c245ae97598d4f37f3ef30	Relationship between bias voltage V_ bias and inter-dot detuning ϵ for different dot sizes. The calculated data points are indicated by circles and the solid line is a fit to the equation ϵ = α V_ bias where α is the lever arm. The region in the dashed black box is enlarged and shown in the inset to demonstrate the linearity of the fitted data.	2012.10512	F_bias_fit.jpg
0.492798	ae5b43f664c747ec88c74b501594b8ad	Schematic of a double QD device formed by plunger, tunnel, and screening gates. The geometric parameters are D_x= D_y= 40 nm, D_t= 20 nm, and T= 15 nm. Applied voltages in (a) and (b) are V_p1= V_p2= 0.150 V, and V_ tun= 0.09455 V. a) 3D device model, where the plunger, tunnel and screening gates are shaded differently for contrast, along with a semi-transparent SiO_2 layer. A 2D electrostatic potential obtained by the self-consistent 3D Poisson calculation is plotted beneath. b) The 2D potential overlaid with an outline of the gate structure, showing the heads of the plunger and tunnel gates as well as the screening gate. The gate lengths (extending beyond the screening gate) are denoted D_y, and the widths of the plunger and tunnel gates are D_x and D_t, respectively. The 2D potential is taken 1 nm below the Si/SiO_2 interface. c) Side profile of the gate structure taken along the dashed line in b). The SiO_2 layer thickness is labelled T. The QD (yellow ellipse) is formed underneath the head of the plunger gate.	2012.10512	F_deviceSchematic.jpg
0.453271	2977cd872fdc484f90d5fe0bb1efbb4d	Dependence of J on tunnel gate voltage as the dot size D=D_x=D_y is varied. a) J versus V_ tun, where the latter is varied from 0.03-0.150 V. The fixed device parameters are V_p=0.150 V, V_ bias=0 V, D_t= 20 nm, and T= 15 nm. The dashed black line indicates J= 1 μeV. b) Derivative of J with respect to V_ bias. For each device geometry, V_ tun is tuned so that J= 1 μeV at V_ bias=0. c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the lever arm between the applied bias voltage and the effective inter-dot detuning. The upper-right table provides a legend and includes the calculated charging energy U and lever arm for each geometry.	2012.10512	F_dot_size.jpg
0.431734	7b42e4e164de46e9b7f06de1334bd47d	Dependence of J on the plunger gate eccentricity, D_y/D_x. For D_y/D_x> 1.0, D_x= 40 nm, while for D_y/D_x< 1.0, D_y= 40 nm. a) J versus V_ tun, as the latter is varied from 0.03-0.150 V. The fixed device parameters are V_p=0.150 V, V_ bias=0 V, D_t= 20 nm, and T= 15 nm. b) Derivative of J with respect to V_ bias. Each device has V_ tun tuned so that J= 1 μeV at V_ bias=0, as indicated by the dashed black line in panel (a). c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the lever arm between the applied bias voltage and the effective inter-dot detuning. The upper-right table provides a legend and includes the charging energy and lever arm versus dot eccentricity.	2012.10512	F_ecc.jpg
0.45815	5ec99ebfd071406a86b4d5efcdcccca3	Dependence of the optimization function f_min(ω) on ω and the size of the harmonic orbital basis (M_xM_y=M). N'=6 for a quartic potential with parameters m^* = 0.191m_0, ħω_0= 0.375 meV, and d = 50 nm. Color indicates a different number of HO basis states ranging from M_x=M_y=1 (purple) to M_x=M_y=16 (yellow).	2012.10512	F_opt_omega.jpg
0.489292	29d5eb415a9e4689aed3f3ecc7810fd4	Dependence of J on varying oxide thickness T. a) J versus V_ tun, as the latter is varied from 0.0-0.100 V. The fixed device parameters are V_p=0.100 V, V_ bias=0 V, D_x= D_y= 40 nm, and D_t= 20 nm. b) Derivative of J with respect to V_ bias. Each device has V_ tun tuned so that J= 1 μeV at V_ bias=0, as indicated by the dashed black line in panel (a). c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the lever arm between the applied bias voltage and the effective inter-dot detuning. The upper-right table provides a legend and includes the charging energies and lever arms versus T.	2012.10512	F_oxide.jpg
0.446911	62b0d1dcc8d1495b9f8f4151f45f35ce	Dependence of J on a varying tunnel gate width D_t. a) J versus V_ tun, as the latter is varied from 0.03-0.150 V. The fixed device parameters are V_p=0.150 V, V_ bias=0 V, D_x= D_y= 40 nm, and T= 15 nm. The dashed black line indicates J= 1 μeV. b) Derivative of J with respect to V_ bias. For each device geometry, V_ tun is tuned so that J= 1 μeV at V_ bias=0. c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the gate lever arm. The upper-right table provides a legend and includes the charging energy U and lever arm α versus D_t.	2012.10512	F_tun_gate.jpg
0.464564	75f09d3902b8461088a2384535482996	1D line cuts of the 2D potentials that give J= 1 μeV at V_ bias= 0 V. Unless varied, the default device parameters are D_x=D_y= 40 nm, D_t= 20 nm, and T= 15 nm. For D_y/D_x>1.0, D_x= 40 nm, and for D_y/D_x<1.0, D_y= 40 nm. The corresponding gate voltages are given in Table <ref>. The 1D slices are taken along the x-axis and pass through the lowest potential minima in the 2D potentials.	2012.10512	F_zero_bias_pots.jpg
0.492839	1e35ce088fef4839a7efc94280d4c3e7	Consort diagram for the studies reviewed	2012.10517	DTIPRISMAv2.jpg
0.503996	b5c957da6ea64a0f845396a51f6efd9c	Our model predicts a 3D Gaussian for each keypoint parameterized by μ, Σ. To supervise this prediction using 2D annotations, we differentiably project the 3D distribution onto each of the C camera planes using para-perspective projection, yielding C Gaussians with parameters {(μ_c, Σ_c)}. We then maximize the likelihood of the annotated keypoints in each image under the projected Gaussians. move images beside above the cameras, this image needs to be much smaller (half its current height) update notation	2012.10518	2d_proj.jpg
0.525169	d75e4300c2834995b718e7115bb9bb13	Model architecture – We predict keypoints as probability distributions in 3D parameterized by a mean μ and covariance Σ representing the position and spatial uncertainty of a keypoint. As input, our model accepts multiple images of a subject taken from multiple (known) camera views. We feed these images through a 2D backbone using the architecture from <cit.> which predicts one heatmap per keypoint within each view. To predict μ, we triangulate the spatial mean of each heatmap using known camera parameters. To predict Σ, we aggregate the bottleneck features of the backbone network, and use a fully connected network to predict a decomposition of Σ that is guaranteed to be positive definite.	2012.10518	architecture.jpg
0.466384	b64d68fc1b1d4d0fa2c0c887b82f19d3	(a): Schematic of arrival times of excitation PL pulse and STED pulse, with delay τ_s between them. (b): Time-resolved ZPL-PL of defect-A with positive (red) and negative (black) τ_s. The STED pulse switches off the emission (c): Time resolved PL of defect-A at five different values of τ_s, showing rapid depletion at arrival time of STED pulse. (d): Plot of time-averaged ZPL-PL intensity with varying τ_s. Moving the STED pulse through the PL pulse results in significant quenching of the PL signal that recovers as τ_s is increased. (e): Sweep of STED ratio with varying time-averaged power of the STED laser.	2012.10520	Fig2.jpg
0.452006	9e5eb6a4ae614bb89682e3f10959f1e4	(a): Calculated optical phonon dispersion for bulk hBN with LO(E_1u) mode highlighted. Taken from Serrano et al. <cit.> (b): Comparison of PL(grey), PLE (blue) and STED (red) OPSB spectra of defect-A, which has a ZPL energy of 2.170 eV. Red (blue) dashed line shows position of the LO(E_1u, Γ) transition in STED (PLE).	2012.10520	Fig3.jpg
0.443387	3b16f414795c4bf9bf427e11c9c83eb7	(a)-(d): Comparison of PL, PLE and STED spectra from four similar defects with ZPL energies of (a) 2.166 eV (b) 2.175eV (c) 2.142 eV and (d) 2.171eV. Each defect shows shift in 200meV peak between absorption and emission. In (b) the peaks marked with ∗ are from another nearby defect or defects.	2012.10520	Fig4.jpg
0.420593	c8fdfd67d32c42d88b338e23980ecd06	Level schematic of three-level system with rate parameters labeled.	2012.10520	SI-fig1.jpg
0.46498	098561ba5b57426fa3a29063a3269a60	(a): Franck-Condon energy diagram comparing PL, PLE, and STED techniques. In PL, the non-resonant excitation (green) populates the excited state and emission from the ZPL (orange) and PSB (red) is collected. In PLE, the excitation probes the excited vibronic states and the ZPL is collected. In STED, the non-resonant excitation is again used, and the STED pulses (red) deplete the excited state through stimulated emission into the ground vibronic states, reducing the ZPL intensity. Grey dashed arrows show fast relaxation from higher vibronic states. (b): Diagram of the experimental set-up. (c): Representative PL (grey), PLE (blue) and STED (red) spectrum from defect-A with ZPL at ∼2.17eV, emission OPSB between 1.97 and 2.03eV and absorption OSPB between 2.32 and 2.38eV.	2012.10520	fig1.jpg
0.499467	7d75c3ab28a24c3ca80e2f6821110582	Example of S for 𝔽_2=⟨ a,b,a^-1,b^-1⟩	2012.10522	freegroup.jpg
0.438251	1364e6d7c42e44ccaa084c8372cd6ae4	Shift on 2^	2012.10522	shiftmap.jpg
0.455737	e6e25cce207a41c48f5ddbe1d48d7f10	▹_s^4(x) with examples of S_x and S_1 x circled	2012.10522	tiles.jpg
0.430229	adadb959b61247bba86bfa8dd6373dc9	Illustration of the proofs of (a) Observation <ref>; (b) Theorem <ref>.	2012.10525	Caterpillar-2.jpg
0.448253	cceecba803d14c2ba3e1821bcbf0d11d	A schematic illustration of a embedding of T on S.	2012.10525	nice.jpg
0.469806	45d7ddcd0b004dd9aaab0b582b1b9f88	An illustration of the tree T in the proof of Theorem <ref>.	2012.10525	np-fixed-tree.jpg
0.444408	4fd2fcc04412445fb42707740b414aec	(a) A partial embedding Γ_i+1 of Γ that is not , while Γ_i is. (b) A modification Γ' of Γ such that Γ'_i+1 is .	2012.10525	np-tree-final-v3.jpg
0.529461	8b9edd93cd274e7e82fd9c9c58da870d	Actual basketball game trajectories. The trajectory data is from a Toronto Raptors NBA game in 2015, which contains offense (red), defense (blue), and ball (green) 2D overhead-view trajectories.	2012.10531	basketball_trajectory.jpg
0.476134	02d5ec800e70479b93c0a70e0e15c491	An overview of the (3D) ITN and STN architectures. Note, 3×3×3 conv. layers (red and green arrows) are followed by a ReLU activation, except for in the final layer of the STN. Outputs of the STN, v and T, are passed to the Transformation Computation Module in Fig. <ref>.	2012.10533	architectures_updated.jpg
0.389598	6c768e0fe1ab4b4bbed733cfe9a9f0fe	An axial slice through the initial (top row) and final (bottom row) atlas image (first column) and the 6 channels of the atlas labelmap produced when training Atlas-ISTN on multi-label CCTA data.	2012.10533	atlas_2d.jpg
0.449408	ce013004b84240ac91f9e1f88d576099	Scatter plots showing HD (left), ASD (middle) and DSC (right) results of Atlas-ISTN on the 1000 case test set, comparing the LVM label from the ITN (x-axis) versus refinement (y-axis). The green/red gradients indicate increase/decrease in performance with refinement. HD and ASD of the ITN predictions almost always improve with refinement. Degradation observed for some cases in terms of DSC is always small, whereas improvements in DSC can be significant.	2012.10533	ccta_scatter_plots_horizontal.jpg
0.421059	9e492f818800487fb606a7e4c6560c9d	An overview of the Atlas-ISTN framework in training (top) and test time refinement (bottom) illustrated with CCTA image data. During training, both ITN and STN weights are optimized by leveraging ground-truth labelmaps for each subject, while the atlas is updated at the end of each epoch. A symmetric loss is used to register the atlas to the subject and vice versa. At test time, instance-specific optimization (indicated by *) leverages the ITN prediction as a target labelmap to update the STN weights, providing a refined transformation from the atlas to a given subject. Spurious segmentations in the ITN prediction can be circumvented via registration of the atlas to the subject. References to related sections and figures are provided in grey text. Red boxes: atlas labelmap (deformed and undeformed). Blue boxes: input image or corresponding GT labelmap (deformed and undeformed). Green boxes: ITN logits (deformed and undeformed).	2012.10533	full_framework_updated.jpg
0.442538	2e6d0471a6e64f6caa7e16a779d4a0d9	Axial (a) and sagittal (b) planes showing examples where refinement corrects for holes (false negatives) in the LVM channel predicted by the ITN, for the Atlas-ISTN trained with an additional 2000 cases. Red contour: LVM label of deformed atlas labelmap after refinement. Yellow contour: ITN prediction of the LVM channel. Heatmap: ITN logits of LVM channel. In both (a) and (b), the ITN prediction contour consists of only a single connected component of the LVM.	2012.10533	lv_hole_segmentation.jpg
0.383227	377369b1433c4925b12f5113e64e89ed	Test cases in order of increasing difficulty (top to bottom), displaying an axial slice from a CCTA image. In columns 3, 5 and 7, manual and predicted contours of the LVM are shown in green and red, respectively, while white contours are used for other predicted structures for clarity of the LVM comparison. Column descriptions are provided above. White arrows highlight false positive and false negative LVM predictions from the ITN. Orange arrows highlight errors between the 1-pass deformed LVM and the manual LVM contours. Rows 1-2 contain examples for which ITN, 1-pass and refinement predictions all perform well, representative of a significant proportion of test cases. Rows 3-4 show cases where the ITN produces spurious segmentations, but refinement is able to circumvent this. In row 4 however, the 1-pass LVM contour is visibly offset, while the refinement produces a better fit. Rows 5-6 show challenging cases, where row 5 contains an example where the heart is particularly small in the field-of-view. On top of this, the ITN predicts a large spurious segmentation extending from the base of the LVM. The deformed contours after 1-pass do not fit the target structures accurately, but after refinement the LVM and other chambers are well positioned. Row 6 shows a case which required a significant rotation and translation of the atlas, so much so that in column 2 a different axial slice is shown to include the undeformed atlas. Additionally the LVM wall is particularly thin, where the ITN predicts a hole near the apex, and an over-segmention in the basal septal region. The significant transformation proves challenging for a single pass of the model, resulting in poor alignment of the atlas to the target SoI. The deformed contours after refinement however align well with the target SoI and bridges the ITN's predicted hole in the LVM (albeit not perfectly fitting to the manual LVM contour). Notice for cases which require larger global transformations (rows 3-6), the deformed grid after refinement contains a more noticeable affine transformation, and the non-rigid component appears to deform the grid less compared to the 1-pass.	2012.10533	seg_results_all.jpg
0.38002	de9994939a134386b35ce9dc93ed4234	Qualitative results for the brain structure segmentation for fives cases from the five different datasets (one case per row) comparing U-net, VML (i.e. a VoxelMorph-like STN-only baseline), initial atlas alignment `Id', and the three outputs ITN, `1-pass' and `Refine' of our Atlas-ISTN approach. The reference segmentation is displayed with green contours and the predicted segmentation boundaries with red contours. Note how the voxel-wise segmentation methods, U-net and ITN, tend to merge neighboring structures, while the VML baseline produces less accurate atlas alignment than `1-pass' or `Refine'. The test time refinement of the Atlas-ISTN seems to produce the visually best results, followed by the `1-pass' prediction which acts as an initialization for the refinement.	2012.10533	segmentation_brain_mri.jpg
0.457451	9a5c50eeb9874d199c03a4fb0628baf2	Axial (a) and sagittal (b) planes showing examples where refinement corrects for (a) holes (false negatives) and (b) spurious segmentations (false positives) in the LVM channel predicted by the ITN. Red contour: LVM label of deformed atlas labelmap after refinement. Yellow contour: ITN prediction of the LVM channel. Heatmap: ITN logits of LVM channel. In both (a) and (b), the ITN prediction contour consists of only a single connected component of the LVM.	2012.10533	spurious_and_hole_segmentation_landscape_2.jpg
0.369804	45bb61d8803d48de822a7a9eb3250772	The left-most column illustrates the base letter B (top), the initial atlas labelmap at the start of training defined in Eq. (<ref>) at t=0 (middle), and the recovered letter B atlas labelmap after training (bottom). The other images are examples from the training set (top row) and clean (middle row) and corrupt test set (bottom row) generated randomly from the base letter B.	2012.10533	synth2d-overview.jpg
0.379657	eb3edc2adffd4ef7be5e4d60fcc7f30f	Qualitative results for the 2D toy data with test data coming from the same distribution as the training data. Both the ITN and 1-pass Atlas-ISTN yield accurate segmentations. test time refinement with increasing regularization weight λ affects the smoothness of the final transformation.	2012.10533	synth2d-results-clean.jpg
0.407635	899618f61cf64fe7be42155b4cc30c69	Qualitative results for the 2D toy data with corrupted, out-of-distribution test data. The ITN yields many false positives and false negatives and is topologically implausible. The 1-pass Atlas-ISTN yields a reasonable atlas alignment despite the corrupted input data. test time refinement with increasing regularization weight λ can yield accurate and topologically plausible segmentations.	2012.10533	synth2d-results-noisy.jpg
0.507786	68116ed96b594e9599b2aa24c916924a	The Transformation Computation Module. The predicted SVF (v) is integrated via scaling and squaring, before linear upsampling. The resulting transformation (ϕ) is composed with the predicted affine parameters (T_i). Computation of the inverse and forward transformations are shown at the top and bottom, respectively. Compositions are defined in Eqs. (<ref>) and (<ref>).	2012.10533	transformation_computation_module.jpg
0.423184	9ee07ca589e441da8555172040083ad3	Intensity projections of images (top) and labelmaps (bottom) of a CCTA training case (left) and the constructed atlas (right). Structures depicted include the left ventricle myocardium (red), right ventricle blood pool (blue), right atrial blood pool (yellow) and left atrial blood pool (green).	2012.10533	volumes.jpg
0.507564	dbd90be8cd7d4f14b66b397151af2818	User_x and User_y is in the same GPS co-ordinate block. They connect with the same wireless access point which provides them C^τ_k_gps and K^τ_k_net. User_x generates a random number Rand_x^τ_k and User_y generates a random number Rand_y^τ_k. Both users concatenate the parameters sent by the server with the network address of the access point and pass them to a hash function to generate a hashed token. Both users upload their generated random numbers and hashed tokens to the server. The server stores the hashed tokens and the random values sent by the users in a database to later use them for contact detection.	2012.10534	Sys_design.jpg
0.444102	753ede67009541f79ce3e09142a520ce	Workflow of the system once an user reports to be positively diagnosed and chooses to share the hashed tokens and pseudo-random numbers generated during the contact detection period of the system.	2012.10534	positive_diagnosis.jpg
0.447425	b8242e463e8e4fbb92c83ec4ed1b432c	Evolution of mean density (upper plot), mass (middle plot), and mean temperature (lower plot) of gravitationally instable regions.	2012.10535	collapsed_regions.jpg
0.513371	70d9203490f247cd9bce5dc3ea777f7a	Numerically obtained evolution of a three-dimensional Gaussian temperature distribution being subject to diffusion only (black diamonds) compared with the analytical solution (red solid line) at different times. The blue dashed line shows the relative error between both solutions. The temperature profile is shown along the x-direction.	2012.10535	cond_evol_2.jpg
0.44717	c26e4431c2ef457ea849750898fe6029	Evolution of Hα intensity as measured for the simulated reference clouds M0 (black solid line) and L0 (red solid line). The dashed-lines and dotted lines show the respective intensities at distances of 1kpc and 10kpc, respectively, in order to compare to <cit.>.	2012.10535	halpha_intensity_t.jpg
0.44227	aa2be47d3ef44359ab0701debac4bd75	The same as Fig. <ref>, but for the low-mass model clouds. The arrow lengths are scaled to v_ ini=248 (model L1b_vel).	2012.10535	hvc_dens_plane_xy_3.jpg
0.389286	8e6868be19774a509c8e501a5138eb1e	Evolution of density in the massive model clouds. The plots show the section plane (x,y) through cloud centre. The upper most plot shows the initial stage. The evolution is shown in snapshots after 20Myr (left column) and after 50Myr (right column). Model M3_hom_tc_sg is shown at 55Myr, because the finger-like structures, which are due to , are substantially developed. Model M1_dec is shown at 100Myr when the head-tail structure is more prominently developed. The black solid lines enclose cloud material (gravitationally bound and physically connected) and the light blue solid lines border gravitationally instable regions. Black arrows illustrate the motion of gas. The arrow lengths are scaled to v_ ini=333 (model M1c_vel). The grey square in the upper left corner is spanned by 10× 10 cells of finest numerical resolution (Δ x=1.3pc).	2012.10535	hvc_dens_plane_xy_4.jpg
0.485961	4575ebd3873a48a5a91eae0777bc5f39	The same as Fig. <ref>, but for respective clouds without thermal conduction.	2012.10535	hvc_mass_hist_3.jpg
0.437948	01167c2e07d24c658a881586b24ec441	Evolution of mass (upper plot) for reference clouds. The corresponding mass-transportation rates (middle plot) show both loss (Ṁ<0) and gain (Ṁ>0) of mass with the average values (dashed lines) indicating a mass loss. The relative mass loss (lower plot) is higher in low-mass clouds.	2012.10535	hvc_mdot_hist_2.jpg
0.460958	08a5019127f943f6bb980fa9de831877	Distributions of metallicity in the stripped material of massive clouds (upper panel) and low-mass clouds (lower panel) after 20Myr (left column) and after 50Myr (right column).	2012.10535	hvc_metl_hist_mass.jpg
0.497577	ef4f543439a442b1afae28e4a4f66122	Metallicity profiles weighted with the square of particle density along the direction of motion for the massive clouds (upper plot) and low-mass clouds (lower plot) after 50Myr.	2012.10535	hvc_metl_profile.jpg
0.441484	9e02bcd7bcb2482bac7cffae71c58c07	Column densities of Hi for selected massive clouds after 50Myr (M1_dec after 100Myr). The black isodensity contours represent log(N_Hi/)=18,19,… Left column: Actual N_Hi for the clouds. Right column: Diluted N_Hi as being obtained at a distance of 100kpc by a beam of opening angle 9arcmin. The black circle shows the area covered by this beam.	2012.10535	hvc_ncol_plane_xy_HI_5_countours.jpg
0.4946	bd6f5c40198243319ac82ee5f38b6bf8	Evolution of maximum particle density (upper plot) and mass (lower plot) of homogeneous and isothermal clouds with and without self-gravity.	2012.10535	hvc_nmax.jpg
0.448687	b3d85be5de1a466293badb8b7a9c9d62	Distributions of temperature in the stripped material of massive clouds (upper panel) and low-mass clouds (lower panel) with thermal conduction (red dashed lines) and without thermal conduction (black solid lines) after 20Myr (left column) and after 50Myr (right column). The standard deviation in every bin is too low to be visible in the plots.	2012.10535	hvc_temp_hist_mass_3.jpg
0.524737	9ece5a76543347208002b894f64cfdf8	Upper panel: Mean density of model clouds M0 (black solid line) and M1_me (red dashed line). Lower panel: Distributions of temperature within the clouds (left column) and in the stripped material (right column) after 20Myr (upper row) and after 50Myr (lower row). The standard deviation in every bin is too low to be visible in the plots.	2012.10535	hvc_temp_hist_mass_5.jpg
0.450071	affb6f2ad74142628429b2366dc8430f	Free-fall time versus disruption time in fastest massive clouds.	2012.10535	hvc_times.jpg
0.520236	2e710fd7275248cc99d2b95aa15695b7	From top to bottom: Initial radial profiles of density, temperature, pressure, and mass for all model clouds.	2012.10535	initialvalues2.jpg
0.463352	01d957ad98c8422fa233599bac4fbca4	Ratio of saturated to classical mass-transportation rate <cit.>.	2012.10535	mdotratio.jpg
0.5049	bd2ca510f8e7406d83e054381cd365d6	Radial cloud-to-Bonnor-Ebert mass ratios for all model clouds. The x-axis is normalized to the respective radius of each cloud.	2012.10535	prof_mbe.jpg
0.454673	43fdb5b93aae4c5381bbce3674a59444	Both disruption time τ_ dis and distance D_ dis for clouds developing a gravitationally instable region. The light-green stars indicate the onsets of collapse.	2012.10535	time_disruption_2.jpg
0.541475	a1fe436c4f0a459794ebf6ec8fd429ce	Deceleration of model clouds that experience drag.	2012.10535	vdot.jpg
0.439199	c2e79457e57f4bd1b256e1ae1177998e	PA concept in urban and rural areas with the vehicles served by the terrestrial BS and satellite, respectively.	2012.10537	Figure1_0814.jpg
0.554139	a953e097b6ac43588e043ef8e510fde2	E2E throughput of PA system in urban areas with spatially-correlated Rayleigh fading conditions. We set SNR = 21 dB, codeword length = 10^4 channel use, the minimum processing delay of the BS = 5 ms, antenna separation = 1.5 times the wavelength, and carrier frequency = 2.68 GHz.	2012.10537	Figure2_0910.jpg
0.41963	feefb69673cb46fa8c7e688c270544c4	Effect of spatial mismatch on the MRT and DFT BF in an NLoS multi-path propagation environment.	2012.10537	Figure3_0814.jpg
0.443605	7b511e21e78a4c41b8b12a7ef7cdbbbb	Received power at the vehicle for various BF schemes, prediction schemes (ideal prediction, without prediction, i.e., with spatial mismatch), and antenna sizes N, in a spatially-correlated Rayleigh fading environment.	2012.10537	Figure4_1217.jpg
0.453729	745905a9087840fea3f4aa657a107118	E2E throughput of the last wagon in a 10-wagon train served by the satellite, with SNR = 26 dB, codeword length = 10^4 channel use, and the minimum processing delay = 10 ms, antenna separation between the first and the last antenna in the same wagon is 10 times the wavelength, and carrier frequency is 2.68 GHz. The distance for adjacent PA and RA in closest wagons is 0.2 m.	2012.10537	Figure5_0814.jpg
0.392001	a7165da2504945f28b0ef9978959c423	Open symbols: Compilation of decoupling (freeze-out) times, τ_f, plotted as a function of the cubic root of the charged particle density, (dN_ch/dη)^1/3, observed in various collision systems and for a wide range of beam energies <cit.>. Filled symbols: extrapolated values of τ_f corresponding to each of the centrality classes used in our computation of η/s based on Au – Au collisions at √(s_ NN) =0.2 TeV and Pb – Pb collisions at √(s_ NN) =2.76 TeV, measured by the STAR and ALICE collaborations, respectively. Red/blue solid lines represent systematic uncertainties of the polynomial fits to the data.	2012.10542	Fig1.jpg
0.42156	da294b4c785a4feb82c46cc7b87b0a4f	Longitudinal width, σ_Δη, of the G_2^CI correlator vs. the estimated number of participants measured in Au – Au collisions at √(s_ NN) = 200 GeV <cit.> and in Pb – Pb collisions at √(s_ NN)=2.76 TeV <cit.>, reported by the STAR and ALICE collaborations, respectively. Error bars and error boxes represent statistical and systematic uncertainties, respectively.	2012.10542	Fig2.jpg
0.416456	7caba646109344828177d64fb7898519	Values of the shear viscosity per unit of entropy density, η/s, computed in this work, as a function of the cubic root of the charged particle density, (dN_ ch/dη)^1/3, measured in Pb – Pb collisions at √(s_ NN)=2.76 TeV <cit.> and in Au – Au collisions at √(s_ NN) = 200 GeV <cit.>. Error bars and error boxes represent statistical and systematic uncertainties, respectively.	2012.10542	Fig3.jpg
0.450413	c1d25d706dde4f53b5c8f503bb6ba307	Comparison of η/s obtained in this work with a collection of values published since 2000 <cit.>, including both theoretical values and experimental values obtained from direct comparisons of theory to data. Some more details are given in the text.	2012.10542	Fig4.jpg
0.418976	0c2c853f36634d2dac40b655ebbcb209		2012.10544	backdoor-code.jpg
0.623664	ea0d1765c4544a8694900c5c349fd303		2012.10544	backdoor-square.jpg
0.426311	4feb316cc9ef4a268778ddd9e41f1f21	In Panel (a), the tunneling (phase) time through the structure air(LR)^7air, whose transmission coefficient is shown in the lower panel of Figure <ref>. The tunneling times of the wave packet (WP) peaks through the air(LR)^7air structure, obtained from τ=∂ t_aSa/∂ω, at the resonant frequencies, are ≅ -0.28fs. In b), a Gaussian wave packet with the centroid at the resonance with frequency ω≃ 5.32 × 10^15rad/s, and the envelope as shown here is prepared, at t=0, at a distance z_0=L=40l_c from the SL.	2012.10546	Fig10e.jpg
0.448874	54f70d68461d439cbb701231daa4794b	A Gaussian WP with the centroid at the resonance with frequency ω≃ 5.32 × 10^15rad/s is prepared, at t=0, at a distance z_0=L=40l_c from the SL (see the blue curve). The WP moves towards D, passing through the SL (LR)^6, and reaches the point D (see the red curves), as predicted, at t=τ_D=2z_0/v_g+τ≃ 54.02fs. The WP is partially transmitted and partially reflected. Not all frequencies are transmitted because the transmission coefficients at the center of the packet are larger than in the tails. Thus, a dip is formed in the reflected (red) WP. The distance between the peak of the (blue) WP and the dip in the reflected (red) WP depends on τ.	2012.10546	Fig11e.jpg
0.45091	0d020de56cc749a9be513fe7efa13b2c	Incoming, reflected, and transmitted fields at the interface dielectric-conductor. b) The incidence θ_i and the effective refraction angle ψ for a silver slab with the dielectric constant shown here.	2012.10546	Fig1e.jpg
0.438625	b0b13a8ae05e4f188a92b1af3d203165	Transmittance as a function of the frequency through a single slab of silver with thickness d_c=30nm (left-hand side graph) and d_c=80nm (right-hand side graph), for incidence angles θ_i=π/6, π/4, π/3, and slightly less than π/2. For the slab thicknesses considered here, the EM field is highly attenuated below the plasma frequency ω_p= 5.75 10^15rad/s and oscillating for ω > ω_p. Near the incidence angle of π/2, the narrow and isolated resonances correspond to localized surface plasmons. Notice that as the incidence angle grows, the resonances move towards the isolated surface plasmon resonances.	2012.10546	Fig2e.jpg
0.482448	ac8a2e3c1ce144b88e0b4c2123219684	Transmittance through a single slab as a function of the incidence angle θ_i, for three different EM frequencies. For the graphs in the left-hand side column, the silver slab thickness is d_c=80nm, and in the right-hand side column, d_c=1000nm. Notice that the number of oscillations at high frequencies is of the order of d_c/λ.	2012.10546	Fig3e.jpg
0.465238	cd3e3bcf80f845c4b57f22b01b3654df	Transmittance as a function of the frequency through the metallic superlattices (air/silver/air)^n. In a) and b), the thickness of the silver and air slabs is equal to d_a=100nm and d_c=10nm. In a), the transmittance is shown for n=8 and incidence angles θ_i=π/6, π/4, π/3, and slightly less than π/2. Below the plasma frequency ω_p= 5.75 10^15rad/s, the transmittance is an oscillating function of ω. Near the incidence angle of π/2, the transmittance is highly resonant. In b), we repeat the transmittances for θ_i=π/6, π/4, but now for n=16, and we plot also the Kramer condition (black curves) and the dispersion relation of Equation (<ref>) derived in the theory of finite periodic systems (TFPS) (red lines). It is clear that this recurrence relation predicts the bands and the frequencies of all the resonant plasmons.	2012.10546	Fig4e.jpg
0.455987	31aca98afd244ea8b85789d41fa792d7	Transmittance as a function of the frequency through the metallic superlattices (air/silver/air)^n where the silver and dielectric widths are a) small and equal and b) different with the silver slabs' width being larger. The transmittances are shown for different incidence angles θ_i indicated on the graphs. Below the plasma frequency ω_p= 5.75 10^15rad/s, for small widths, the transmittance is independent of the incidence angle, while for larger silver width, the metallic superlattice is almost completely transparent. For frequencies above ω_p, we have wider bands for small layers' widths and thinner bands for larger silver widths.	2012.10546	Fig5e.jpg

0.378794

d26d79b11ed3443ea6a39aaf74a6cfd0

The first two panels from the left illustrate the different formation paths described in section [sec:formation_channels]3.1. In the direct physical collision scenario, the dominant process by which the principal object increases in mass is through physical collisions, as is shown by the initial binary-binary interaction that leads to the merger of two stars. In contrast, in the middle, for the formation path labeled binary coalescence, the main increase in mass comes from binary coalescence.The right-most column illustrates the collision history of the most massive IMBH formed in our simulations.

2012.10497

tree.jpg

0.464724

679aec000b89491d89931845bdb479ab

Specific heat capacity of NdCoGe_3 for (a) magnetic field within the basal plane and (b) field applied along the c axis.

2012.10499

Cp1.jpg

0.543801

d7f70238ad7d4e4b8ab2a685ff2931d9

Zero-field data revealing magnetic phase transitions at T_N1 = 3.70 K and T_N2 = 3.50 K in NdCoGe_3 indicated by the vertical dashed lines. (a) Specific heat capacity and (b,c,d) the in-plane ac magnetic susceptibility measured in zero applied DC field. The ac data are shown as the (b) in-phase contribution χ' with (c) displaying the derivative of χ', and (d) contains the out-of-phase contribution χ”.

2012.10499

Figure1.jpg

0.497287

cd469f8ad55240808cba91c5442c6d1e

(a,c,e) Neutron diffraction intensity as a function of h, k or l at various temperatures and (b,d,f) the fitted position of h, k, and l as a function of temperature. The red lines in (a,c,e) are the corresponding Gaussian fit(s) and the fitted error bars in (b,d,f) are smaller than the size of the symbols. In (a), the small shoulders for smaller h are artifacts from the sample analyzer collimator.

2012.10499

Figure3.jpg

0.497347

8ae2fc5a5ea54ee08f6e69c1a84ad08c

Temperature-dependent magnetization in zero-field-cooled warming (red) and field-cooled cooling (blue) measurements with an applied field (a) H ⊥ c and (b) H||c. The inset in (a) shows the magnetization to T = 0.4 K using the same units as the main panel.

2012.10499

Figure4.jpg

0.415548

3dd2f767d1db43b78e4e48ac8e416a97

Isothermal magnetization M(H) and derivative of isothermal magnetization dM/dH of NdCoGe_3 with an applied field H⊥c. (a) M(H) (left axis) and dM/dH (right axis) at T = 0.4 K (b) The in-plane ac magnetic susceptibility χ' as a function of applied DC field at T = 0.4 K (c) dM/dH from 1.8 K to 2.8 K with a temperature step of Δ T = 0.04 K. (d) dM/dH at T = 3 K.

2012.10499

Figure5.jpg

0.473994

288057c4e3104d7bb4181e037434fdfc

(a,b) Isothermal magnetization data (left axis) and derivative dM/dH (right axis) on increasing field (red line) and decreasing field (blue line) of NdCoGe_3 with an applied field H⊥c.

2012.10499

Figure6.jpg

0.421257

f2fadf026cc34d79aa26a9da0edf7240

Neutron powder diffraction pattern of NdCoGe_3 at T = 1.8 K < T_N2 . Inset: Data for a limited Q range revealing magnetic reflections at 1.8 K (bottom) through a comparison to the pattern obtained at 40 K (top) for which the Bragg peaks are entirely from the crystal lattice.

2012.10499

NeutronPowder.jpg

0.410236

387e066b43d04a809a4d89a19710903c

Magnetic phase diagrams of NdCoGe_3 as inferred from anomalies in the specific heat, derivatives of temperature-dependent and isothermal magnetization as indicated in the legends for (a) H⊥c and (b) H||c. The black dotted line in (a) indicates the presence of a first order phase boundary based upon observations of thermal or field hysteresis.

2012.10499

Phase.jpg

0.445534

e726616ccac54a5a9af51274485f77cd

Neutron single crystal diffraction data for NdCoGe_3 from the WAND^2. A section of the reciprocal h0l plane between 00l and -10l with the magnetic reflections close to the -0.5 0 l position. The color scale is adjusted to show both the weak magnetic intensity and the stronger nuclear scattering. The rings originate from the Al-sample holder. Purple arrows indicate the propagation vector from the crystallographic -103 and 004 reflections.

2012.10499

Wand2_new.jpg

0.449911

13622d1a9f814824a85273c24bc5575f

Candidate magnetic structures resulting from (a) 1k⃗, (b) 2k⃗, (c) 3k⃗ and (d) 4k⃗ models. The vector moments (blue arrows) on Nd atoms (blue spheres) are shown for a segment of the magnetic unit cell; the nuclear (paramagnetic) unit cell is outlined in grey, the Co atoms are red and Ge atoms are grey. The various models describe the neutron powder diffraction data at 1.8 K equally well.

2012.10499

mag_str.jpg

0.413801

95c2b092a1e64911a5efcba08ff64bfa

Location of the FASERν detector and event topology. Top: The FASER experiment is placed about 500 m downstream of the ATLAS interaction point in the previously unused side tunnel TI12, which connects the SPS with the LHC tunnel. Center: The detector is centered around the beam collision axis where the neutrino flux is maximal. It consists of the FASERν emulsion neutrino detector, followed by a magnetized spectrometer and a calorimeter. Bottom: The emulsion detector consists of tungsten plates interleaved with nuclear emulsion films. Both interactions of neutrinos and neutral hadrons lead to the appearance of a neutral vertex at which several charged particles emerge. Different types of events can be distinguished based on the event topology, as explained in the text.

2012.10500

Figure.jpg

0.517272

776b9d59f2344eb09fbc35259b8c168f

Comparison of 3 observables for NH interactions generated using , , and and NC interactions using , and . Top: NH interactions at 60, 300. Middle: Different NHs (mesons and baryons) generated via at 300. Bottom: NC interactions at 100, 1 generated via , and . See text for more details.

2012.10500

appendix_fig.jpg

0.429211

c674bcc9d693488ebece1fa7a7245cc5

Left: Correlation matrix showing the linear relationship between observables presented in <ref>, the incoming particle energy (E_beam) and the event type (1 for NC neutrino interactions, 0 for neutral hadrons). Right: Permutation feature importance (the normalized mean score decrease for each of the observables) for the signal identification classifier (blue) and neutrino energy estimator (red) network. We use accuracy as the score metric for the classifier network and mean average percents error for the estimator. Scores decreases are normalized so that they sum to 1.

2012.10500

corr_perm_importance.jpg

0.385364

be11346bd4b049aab3847874a59d5d15

Left: Signal selection efficiencies as a function of beam energy. Each line indicates the fraction of events passing the following criteria sequentially: i) neutral vertex identification (blue) requiring ≥ 5 charged tracks, ii) lepton veto (green) requiring no electron candidate and no non-interacting charged track, and iii) signal identification (red) as performed by the NN classifier. The dashed black line shows the combined efficiency. Right: The energy spectra of particles interacting within the FASERν detector. We show the expected numbers of neutral hadron interactions (green) and NC neutrino interactions (red) with the FASERν detector during LHC Run 3 as dashed lines. The solid lines show the spectra for events passing the signal selection (including neutral vertex identification, lepton veto and signal identification). The uncertainty associated to the background generation is shown as a shaded band.

2012.10500

efficiency_intrct_spectrum.jpg

0.393763

063b20839c7140338a9c2fcb9587e72d

Left: Neutrino energy reconstruction for NC neutrino interaction events obtained by a neural network-based multivariate analysis using the observables defined in <ref>. Right: Relative RMS energy resolution using the neural network-based multivariate analysis (solid) and only the hadronic energy of the events (dashed).

2012.10500

energy2D_resolution.jpg

0.458506

7257e621190c4eecbeb1c634f4685840

Expected energy spectrum of neutral hadrons interacting within the FASERν detector during LHC Run 3 with 150 of luminosity.

2012.10500

hadspec_072720_neutral.jpg

0.495289

e46949d501bf4c59b2511b6af8e1d293

Normalized kinematic distributions for the observables defined in <ref>. The dashed lines show the distributions obtained with for the NC neutrino interaction signal at incoming neutrino energies of E_ν=100 (blue) and E_ν=1 (red). The solid green lines correspond to the distributions for the neutral hadron interactions simulated with for the expected energy spectrum presented in <ref>. The shaded region shows the range of predictions for the background distributions obtained from different generators: , and .

2012.10500

obs.jpg

0.442296

e36519aa97ee4818a90ee63d7d142c74

Left: Stacked histogram of events passing the event signal selection described in <ref> as function of the reconstructed energy for LHC Run 3 with 150 integrated luminosity. The red and green shaded regions show the NC neutrino interaction signal and the neutral hadron interaction background, respectively. The hatched region indicates the uncertainty arising from the simulation of neutral hadron interactions, corresponding to the range of predictions obtained by three different generators. Right: FASERν's estimated neutrino-tungsten NC cross section sensitivity. Existing constraints are shown in gray. The black dashed curve is the theoretical prediction for the DIS cross section, averaged over neutrinos and anti-neutrinos, per tungsten nucleus. The inner red error bars correspond to statistical uncertainties, the blue error bars additional take into account uncertainties associated with the simulation of the background, and the outer green error bars show the combined uncertainties with the neutrino production rate (which corresponds to the range of predictions obtained from different MC generators as obtained in Ref. <cit.>).

2012.10500

signal_spectrum_crosssection.jpg

0.483784

c2cc8de2259e45aca2831214faf5470c

Spectra of radiation from sidewalls of the accretion columns, calculated simultaneously with velocity and temperature profiles plotted in Figs <ref> and <ref>. The results correspond to different sets of the model parameters: `1' (solid line), `2' (dashed), `3' (dot-dashed), `4' (dot-double-dashed), `5' (double-dot-dashed), `6' (dotted).

2012.10501

figsp1.jpg

0.455476

676006c692dd46b295d0a3b315e4fade

(a) Electron data (in blue) from left to right: SUMER on-disk “correlation heights” in open funnels <cit.>, SUMER off-limb temperatures from (; dark blue) and (; light blue), UVCS <cit.>, PSP <cit.>, and Helios/Ulysses <cit.>. Proton data (in red) from left to right: UVCS <cit.>, PSP (dark red), Helios (light red), Ulysses (dark red), and Voyager (light red) high-speed wind data. Oxygen-ion data (in green) from left to right: SUMER off-limb kinetic temperatures <cit.>, UVCS <cit.>, and Wind <cit.>. (b) From left to right, O^+5 and proton anisotropy data from UVCS <cit.>, PSP data with v ≥ 450 km s^-1 <cit.>, Helios data with v ≥ 600 km s^-1 <cit.>, and Ulysses data with v ≥ 600 km s^-1 <cit.>. All black curves are solely to guide the eye.

2012.10509

cranmer_temps2020_fig1.jpg

0.459751

0a57cfccc5fd4e298dcbd19763830963

Comparison of the HL (purple) and HM (dark blue) methods to the modified LCHO-CI method for evaluating J. The system is a double QD in Si/SiO_2 described by a quartic potential, where l_0= 6 nm. a) Probability densities of the ground single-electron orbital states for the HL/HM and LCHO-CI calculations at different dot separations. At small dot separations, the HL/HM states overestimate the separation of the two localized wave functions compared to the numerically calculated orbitals. As dot separation increases, the approximate HL/HM states converge to the numerical results. b) J versus the interdot separation 2d is plotted. Three different LCHO-CI calculations are done for N=2 (light blue), N=4 (green), and N=10 (yellow). For all LCHO-CI calculations, M=15^2.

2012.10512

F_HL_HM_comp.jpg

0.430391

49ce28a13a74468ba463c69ddd820f0e

Convergence of J with respect to the number of single-electron orbitals N when M=16^2. The device parameters are V_p1= V_p2= 0.15 V, V_t= 0.09455 V, D_x= D_y= 40 nm, D_t= 20 nm, and T= 15 nm. Inset: The convergence of J with respect to M when N=18 (star in the main figure).

2012.10512

F_Jconvergance.jpg

0.403861

f490e4da394046ef9c9b58455438b9b3

Using harmonic orbitals (HO) to approximate single-electron states. a) The operator Â transforms between the HO basis {|ϕ_i⟩} and the approximate single-electron orbitals |ξ_j'⟩. The first three single-electron orbitals for a quartic potential are shown at left, and the eight lowest energy HO states are shown at right. b) Convergence of the first twelve approximate single-electron energies ϵ_j' versus the number of HOs, M, used to compose the basis {|ϕ_i^ω⟩}. Here, M = M_xM_y and M_x=M_y are the numbers of 1-dimensional HOs taken along the x and y axes, respectively, to construct the 2-dimensional HOs.

2012.10512

F_SEconvergance.jpg

0.532902

ddd86be3f3c245ae97598d4f37f3ef30

Relationship between bias voltage V_ bias and inter-dot detuning ϵ for different dot sizes. The calculated data points are indicated by circles and the solid line is a fit to the equation ϵ = α V_ bias where α is the lever arm. The region in the dashed black box is enlarged and shown in the inset to demonstrate the linearity of the fitted data.

2012.10512

F_bias_fit.jpg

0.492798

ae5b43f664c747ec88c74b501594b8ad

Schematic of a double QD device formed by plunger, tunnel, and screening gates. The geometric parameters are D_x= D_y= 40 nm, D_t= 20 nm, and T= 15 nm. Applied voltages in (a) and (b) are V_p1= V_p2= 0.150 V, and V_ tun= 0.09455 V. a) 3D device model, where the plunger, tunnel and screening gates are shaded differently for contrast, along with a semi-transparent SiO_2 layer. A 2D electrostatic potential obtained by the self-consistent 3D Poisson calculation is plotted beneath. b) The 2D potential overlaid with an outline of the gate structure, showing the heads of the plunger and tunnel gates as well as the screening gate. The gate lengths (extending beyond the screening gate) are denoted D_y, and the widths of the plunger and tunnel gates are D_x and D_t, respectively. The 2D potential is taken 1 nm below the Si/SiO_2 interface. c) Side profile of the gate structure taken along the dashed line in b). The SiO_2 layer thickness is labelled T. The QD (yellow ellipse) is formed underneath the head of the plunger gate.

2012.10512

F_deviceSchematic.jpg

0.453271

2977cd872fdc484f90d5fe0bb1efbb4d

Dependence of J on tunnel gate voltage as the dot size D=D_x=D_y is varied. a) J versus V_ tun, where the latter is varied from 0.03-0.150 V. The fixed device parameters are V_p=0.150 V, V_ bias=0 V, D_t= 20 nm, and T= 15 nm. The dashed black line indicates J= 1 μeV. b) Derivative of J with respect to V_ bias. For each device geometry, V_ tun is tuned so that J= 1 μeV at V_ bias=0. c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the lever arm between the applied bias voltage and the effective inter-dot detuning. The upper-right table provides a legend and includes the calculated charging energy U and lever arm for each geometry.

2012.10512

F_dot_size.jpg

0.431734

7b42e4e164de46e9b7f06de1334bd47d

Dependence of J on the plunger gate eccentricity, D_y/D_x. For D_y/D_x> 1.0, D_x= 40 nm, while for D_y/D_x< 1.0, D_y= 40 nm. a) J versus V_ tun, as the latter is varied from 0.03-0.150 V. The fixed device parameters are V_p=0.150 V, V_ bias=0 V, D_t= 20 nm, and T= 15 nm. b) Derivative of J with respect to V_ bias. Each device has V_ tun tuned so that J= 1 μeV at V_ bias=0, as indicated by the dashed black line in panel (a). c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the lever arm between the applied bias voltage and the effective inter-dot detuning. The upper-right table provides a legend and includes the charging energy and lever arm versus dot eccentricity.

2012.10512

F_ecc.jpg

0.45815

5ec99ebfd071406a86b4d5efcdcccca3

Dependence of the optimization function f_min(ω) on ω and the size of the harmonic orbital basis (M_xM_y=M). N'=6 for a quartic potential with parameters m^* = 0.191m_0, ħω_0= 0.375 meV, and d = 50 nm. Color indicates a different number of HO basis states ranging from M_x=M_y=1 (purple) to M_x=M_y=16 (yellow).

2012.10512

F_opt_omega.jpg

0.489292

29d5eb415a9e4689aed3f3ecc7810fd4

Dependence of J on varying oxide thickness T. a) J versus V_ tun, as the latter is varied from 0.0-0.100 V. The fixed device parameters are V_p=0.100 V, V_ bias=0 V, D_x= D_y= 40 nm, and D_t= 20 nm. b) Derivative of J with respect to V_ bias. Each device has V_ tun tuned so that J= 1 μeV at V_ bias=0, as indicated by the dashed black line in panel (a). c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the lever arm between the applied bias voltage and the effective inter-dot detuning. The upper-right table provides a legend and includes the charging energies and lever arms versus T.

2012.10512

F_oxide.jpg

0.446911

62b0d1dcc8d1495b9f8f4151f45f35ce

Dependence of J on a varying tunnel gate width D_t. a) J versus V_ tun, as the latter is varied from 0.03-0.150 V. The fixed device parameters are V_p=0.150 V, V_ bias=0 V, D_x= D_y= 40 nm, and T= 15 nm. The dashed black line indicates J= 1 μeV. b) Derivative of J with respect to V_ bias. For each device geometry, V_ tun is tuned so that J= 1 μeV at V_ bias=0. c) Derivative of J with respect to ϵ, where ϵ = α V_ bias and α is the gate lever arm. The upper-right table provides a legend and includes the charging energy U and lever arm α versus D_t.

2012.10512

F_tun_gate.jpg

0.464564

75f09d3902b8461088a2384535482996

1D line cuts of the 2D potentials that give J= 1 μeV at V_ bias= 0 V. Unless varied, the default device parameters are D_x=D_y= 40 nm, D_t= 20 nm, and T= 15 nm. For D_y/D_x>1.0, D_x= 40 nm, and for D_y/D_x<1.0, D_y= 40 nm. The corresponding gate voltages are given in Table <ref>. The 1D slices are taken along the x-axis and pass through the lowest potential minima in the 2D potentials.

2012.10512

F_zero_bias_pots.jpg

0.492839

1e35ce088fef4839a7efc94280d4c3e7

Consort diagram for the studies reviewed

2012.10517

DTIPRISMAv2.jpg

0.503996

b5c957da6ea64a0f845396a51f6efd9c

Our model predicts a 3D Gaussian for each keypoint parameterized by μ, Σ. To supervise this prediction using 2D annotations, we differentiably project the 3D distribution onto each of the C camera planes using para-perspective projection, yielding C Gaussians with parameters {(μ_c, Σ_c)}. We then maximize the likelihood of the annotated keypoints in each image under the projected Gaussians. move images beside above the cameras, this image needs to be much smaller (half its current height) update notation

2012.10518

2d_proj.jpg

0.525169

d75e4300c2834995b718e7115bb9bb13

Model architecture – We predict keypoints as probability distributions in 3D parameterized by a mean μ and covariance Σ representing the position and spatial uncertainty of a keypoint. As input, our model accepts multiple images of a subject taken from multiple (known) camera views. We feed these images through a 2D backbone using the architecture from <cit.> which predicts one heatmap per keypoint within each view. To predict μ, we triangulate the spatial mean of each heatmap using known camera parameters. To predict Σ, we aggregate the bottleneck features of the backbone network, and use a fully connected network to predict a decomposition of Σ that is guaranteed to be positive definite.

2012.10518

architecture.jpg

0.466384

b64d68fc1b1d4d0fa2c0c887b82f19d3

(a): Schematic of arrival times of excitation PL pulse and STED pulse, with delay τ_s between them. (b): Time-resolved ZPL-PL of defect-A with positive (red) and negative (black) τ_s. The STED pulse switches off the emission (c): Time resolved PL of defect-A at five different values of τ_s, showing rapid depletion at arrival time of STED pulse. (d): Plot of time-averaged ZPL-PL intensity with varying τ_s. Moving the STED pulse through the PL pulse results in significant quenching of the PL signal that recovers as τ_s is increased. (e): Sweep of STED ratio with varying time-averaged power of the STED laser.

2012.10520

Fig2.jpg

0.452006

9e5eb6a4ae614bb89682e3f10959f1e4

(a): Calculated optical phonon dispersion for bulk hBN with LO(E_1u) mode highlighted. Taken from Serrano et al. <cit.> (b): Comparison of PL(grey), PLE (blue) and STED (red) OPSB spectra of defect-A, which has a ZPL energy of 2.170 eV. Red (blue) dashed line shows position of the LO(E_1u, Γ) transition in STED (PLE).

2012.10520

Fig3.jpg

0.443387

3b16f414795c4bf9bf427e11c9c83eb7

(a)-(d): Comparison of PL, PLE and STED spectra from four similar defects with ZPL energies of (a) 2.166 eV (b) 2.175eV (c) 2.142 eV and (d) 2.171eV. Each defect shows shift in 200meV peak between absorption and emission. In (b) the peaks marked with ∗ are from another nearby defect or defects.

2012.10520

Fig4.jpg

0.420593

c8fdfd67d32c42d88b338e23980ecd06

Level schematic of three-level system with rate parameters labeled.

2012.10520

SI-fig1.jpg

0.46498

098561ba5b57426fa3a29063a3269a60

(a): Franck-Condon energy diagram comparing PL, PLE, and STED techniques. In PL, the non-resonant excitation (green) populates the excited state and emission from the ZPL (orange) and PSB (red) is collected. In PLE, the excitation probes the excited vibronic states and the ZPL is collected. In STED, the non-resonant excitation is again used, and the STED pulses (red) deplete the excited state through stimulated emission into the ground vibronic states, reducing the ZPL intensity. Grey dashed arrows show fast relaxation from higher vibronic states. (b): Diagram of the experimental set-up. (c): Representative PL (grey), PLE (blue) and STED (red) spectrum from defect-A with ZPL at ∼2.17eV, emission OPSB between 1.97 and 2.03eV and absorption OSPB between 2.32 and 2.38eV.

2012.10520

fig1.jpg

0.499467

7d75c3ab28a24c3ca80e2f6821110582

Example of S for 𝔽_2=⟨ a,b,a^-1,b^-1⟩

2012.10522

freegroup.jpg

0.438251

1364e6d7c42e44ccaa084c8372cd6ae4

Shift on 2^

2012.10522

shiftmap.jpg

0.455737

e6e25cce207a41c48f5ddbe1d48d7f10

▹_s^4(x) with examples of S_x and S_1 x circled

2012.10522

tiles.jpg

0.430229

adadb959b61247bba86bfa8dd6373dc9

Illustration of the proofs of (a) Observation <ref>; (b) Theorem <ref>.

2012.10525

Caterpillar-2.jpg

0.448253

cceecba803d14c2ba3e1821bcbf0d11d

A schematic illustration of a embedding of T on S.

2012.10525

nice.jpg

0.469806

45d7ddcd0b004dd9aaab0b582b1b9f88

An illustration of the tree T in the proof of Theorem <ref>.

2012.10525

np-fixed-tree.jpg

0.444408

4fd2fcc04412445fb42707740b414aec

(a) A partial embedding Γ_i+1 of Γ that is not , while Γ_i is. (b) A modification Γ' of Γ such that Γ'_i+1 is .

2012.10525

np-tree-final-v3.jpg

0.529461

8b9edd93cd274e7e82fd9c9c58da870d

Actual basketball game trajectories. The trajectory data is from a Toronto Raptors NBA game in 2015, which contains offense (red), defense (blue), and ball (green) 2D overhead-view trajectories.

2012.10531

basketball_trajectory.jpg

0.476134

02d5ec800e70479b93c0a70e0e15c491

An overview of the (3D) ITN and STN architectures. Note, 3×3×3 conv. layers (red and green arrows) are followed by a ReLU activation, except for in the final layer of the STN. Outputs of the STN, v and T, are passed to the Transformation Computation Module in Fig. <ref>.

2012.10533

architectures_updated.jpg

0.389598

6c768e0fe1ab4b4bbed733cfe9a9f0fe

An axial slice through the initial (top row) and final (bottom row) atlas image (first column) and the 6 channels of the atlas labelmap produced when training Atlas-ISTN on multi-label CCTA data.

2012.10533

atlas_2d.jpg

0.449408

ce013004b84240ac91f9e1f88d576099

Scatter plots showing HD (left), ASD (middle) and DSC (right) results of Atlas-ISTN on the 1000 case test set, comparing the LVM label from the ITN (x-axis) versus refinement (y-axis). The green/red gradients indicate increase/decrease in performance with refinement. HD and ASD of the ITN predictions almost always improve with refinement. Degradation observed for some cases in terms of DSC is always small, whereas improvements in DSC can be significant.

2012.10533

ccta_scatter_plots_horizontal.jpg

0.421059

9e492f818800487fb606a7e4c6560c9d

An overview of the Atlas-ISTN framework in training (top) and test time refinement (bottom) illustrated with CCTA image data. During training, both ITN and STN weights are optimized by leveraging ground-truth labelmaps for each subject, while the atlas is updated at the end of each epoch. A symmetric loss is used to register the atlas to the subject and vice versa. At test time, instance-specific optimization (indicated by *) leverages the ITN prediction as a target labelmap to update the STN weights, providing a refined transformation from the atlas to a given subject. Spurious segmentations in the ITN prediction can be circumvented via registration of the atlas to the subject. References to related sections and figures are provided in grey text. Red boxes: atlas labelmap (deformed and undeformed). Blue boxes: input image or corresponding GT labelmap (deformed and undeformed). Green boxes: ITN logits (deformed and undeformed).

2012.10533

full_framework_updated.jpg

0.442538

2e6d0471a6e64f6caa7e16a779d4a0d9

Axial (a) and sagittal (b) planes showing examples where refinement corrects for holes (false negatives) in the LVM channel predicted by the ITN, for the Atlas-ISTN trained with an additional 2000 cases. Red contour: LVM label of deformed atlas labelmap after refinement. Yellow contour: ITN prediction of the LVM channel. Heatmap: ITN logits of LVM channel. In both (a) and (b), the ITN prediction contour consists of only a single connected component of the LVM.

2012.10533

lv_hole_segmentation.jpg

0.383227

377369b1433c4925b12f5113e64e89ed

Test cases in order of increasing difficulty (top to bottom), displaying an axial slice from a CCTA image. In columns 3, 5 and 7, manual and predicted contours of the LVM are shown in green and red, respectively, while white contours are used for other predicted structures for clarity of the LVM comparison. Column descriptions are provided above. White arrows highlight false positive and false negative LVM predictions from the ITN. Orange arrows highlight errors between the 1-pass deformed LVM and the manual LVM contours. Rows 1-2 contain examples for which ITN, 1-pass and refinement predictions all perform well, representative of a significant proportion of test cases. Rows 3-4 show cases where the ITN produces spurious segmentations, but refinement is able to circumvent this. In row 4 however, the 1-pass LVM contour is visibly offset, while the refinement produces a better fit. Rows 5-6 show challenging cases, where row 5 contains an example where the heart is particularly small in the field-of-view. On top of this, the ITN predicts a large spurious segmentation extending from the base of the LVM. The deformed contours after 1-pass do not fit the target structures accurately, but after refinement the LVM and other chambers are well positioned. Row 6 shows a case which required a significant rotation and translation of the atlas, so much so that in column 2 a different axial slice is shown to include the undeformed atlas. Additionally the LVM wall is particularly thin, where the ITN predicts a hole near the apex, and an over-segmention in the basal septal region. The significant transformation proves challenging for a single pass of the model, resulting in poor alignment of the atlas to the target SoI. The deformed contours after refinement however align well with the target SoI and bridges the ITN's predicted hole in the LVM (albeit not perfectly fitting to the manual LVM contour). Notice for cases which require larger global transformations (rows 3-6), the deformed grid after refinement contains a more noticeable affine transformation, and the non-rigid component appears to deform the grid less compared to the 1-pass.

2012.10533

seg_results_all.jpg

0.38002

de9994939a134386b35ce9dc93ed4234

Qualitative results for the brain structure segmentation for fives cases from the five different datasets (one case per row) comparing U-net, VML (i.e. a VoxelMorph-like STN-only baseline), initial atlas alignment `Id', and the three outputs ITN, `1-pass' and `Refine' of our Atlas-ISTN approach. The reference segmentation is displayed with green contours and the predicted segmentation boundaries with red contours. Note how the voxel-wise segmentation methods, U-net and ITN, tend to merge neighboring structures, while the VML baseline produces less accurate atlas alignment than `1-pass' or `Refine'. The test time refinement of the Atlas-ISTN seems to produce the visually best results, followed by the `1-pass' prediction which acts as an initialization for the refinement.

2012.10533

segmentation_brain_mri.jpg

0.457451

9a5c50eeb9874d199c03a4fb0628baf2

Axial (a) and sagittal (b) planes showing examples where refinement corrects for (a) holes (false negatives) and (b) spurious segmentations (false positives) in the LVM channel predicted by the ITN. Red contour: LVM label of deformed atlas labelmap after refinement. Yellow contour: ITN prediction of the LVM channel. Heatmap: ITN logits of LVM channel. In both (a) and (b), the ITN prediction contour consists of only a single connected component of the LVM.

2012.10533

spurious_and_hole_segmentation_landscape_2.jpg

0.369804

45bb61d8803d48de822a7a9eb3250772

The left-most column illustrates the base letter B (top), the initial atlas labelmap at the start of training defined in Eq. (<ref>) at t=0 (middle), and the recovered letter B atlas labelmap after training (bottom). The other images are examples from the training set (top row) and clean (middle row) and corrupt test set (bottom row) generated randomly from the base letter B.

2012.10533

synth2d-overview.jpg

0.379657

eb3edc2adffd4ef7be5e4d60fcc7f30f

Qualitative results for the 2D toy data with test data coming from the same distribution as the training data. Both the ITN and 1-pass Atlas-ISTN yield accurate segmentations. test time refinement with increasing regularization weight λ affects the smoothness of the final transformation.

2012.10533

synth2d-results-clean.jpg

0.407635

899618f61cf64fe7be42155b4cc30c69

Qualitative results for the 2D toy data with corrupted, out-of-distribution test data. The ITN yields many false positives and false negatives and is topologically implausible. The 1-pass Atlas-ISTN yields a reasonable atlas alignment despite the corrupted input data. test time refinement with increasing regularization weight λ can yield accurate and topologically plausible segmentations.

2012.10533

synth2d-results-noisy.jpg

0.507786

68116ed96b594e9599b2aa24c916924a

The Transformation Computation Module. The predicted SVF (v) is integrated via scaling and squaring, before linear upsampling. The resulting transformation (ϕ) is composed with the predicted affine parameters (T_i). Computation of the inverse and forward transformations are shown at the top and bottom, respectively. Compositions are defined in Eqs. (<ref>) and (<ref>).

2012.10533

transformation_computation_module.jpg

0.423184

9ee07ca589e441da8555172040083ad3

Intensity projections of images (top) and labelmaps (bottom) of a CCTA training case (left) and the constructed atlas (right). Structures depicted include the left ventricle myocardium (red), right ventricle blood pool (blue), right atrial blood pool (yellow) and left atrial blood pool (green).

2012.10533

volumes.jpg

0.507564

dbd90be8cd7d4f14b66b397151af2818

User_x and User_y is in the same GPS co-ordinate block. They connect with the same wireless access point which provides them C^τ_k_gps and K^τ_k_net. User_x generates a random number Rand_x^τ_k and User_y generates a random number Rand_y^τ_k. Both users concatenate the parameters sent by the server with the network address of the access point and pass them to a hash function to generate a hashed token. Both users upload their generated random numbers and hashed tokens to the server. The server stores the hashed tokens and the random values sent by the users in a database to later use them for contact detection.

2012.10534

Sys_design.jpg

0.444102

753ede67009541f79ce3e09142a520ce

Workflow of the system once an user reports to be positively diagnosed and chooses to share the hashed tokens and pseudo-random numbers generated during the contact detection period of the system.

2012.10534

positive_diagnosis.jpg

0.447425

b8242e463e8e4fbb92c83ec4ed1b432c

Evolution of mean density (upper plot), mass (middle plot), and mean temperature (lower plot) of gravitationally instable regions.

2012.10535

collapsed_regions.jpg

0.513371

70d9203490f247cd9bce5dc3ea777f7a

Numerically obtained evolution of a three-dimensional Gaussian temperature distribution being subject to diffusion only (black diamonds) compared with the analytical solution (red solid line) at different times. The blue dashed line shows the relative error between both solutions. The temperature profile is shown along the x-direction.

2012.10535

cond_evol_2.jpg

0.44717

c26e4431c2ef457ea849750898fe6029

Evolution of Hα intensity as measured for the simulated reference clouds M0 (black solid line) and L0 (red solid line). The dashed-lines and dotted lines show the respective intensities at distances of 1kpc and 10kpc, respectively, in order to compare to <cit.>.

2012.10535

halpha_intensity_t.jpg

0.44227

aa2be47d3ef44359ab0701debac4bd75

The same as Fig. <ref>, but for the low-mass model clouds. The arrow lengths are scaled to v_ ini=248 (model L1b_vel).

2012.10535

hvc_dens_plane_xy_3.jpg

0.389286

8e6868be19774a509c8e501a5138eb1e

Evolution of density in the massive model clouds. The plots show the section plane (x,y) through cloud centre. The upper most plot shows the initial stage. The evolution is shown in snapshots after 20Myr (left column) and after 50Myr (right column). Model M3_hom_tc_sg is shown at 55Myr, because the finger-like structures, which are due to , are substantially developed. Model M1_dec is shown at 100Myr when the head-tail structure is more prominently developed. The black solid lines enclose cloud material (gravitationally bound and physically connected) and the light blue solid lines border gravitationally instable regions. Black arrows illustrate the motion of gas. The arrow lengths are scaled to v_ ini=333 (model M1c_vel). The grey square in the upper left corner is spanned by 10× 10 cells of finest numerical resolution (Δ x=1.3pc).

2012.10535

hvc_dens_plane_xy_4.jpg

0.485961

4575ebd3873a48a5a91eae0777bc5f39

The same as Fig. <ref>, but for respective clouds without thermal conduction.

2012.10535

hvc_mass_hist_3.jpg

0.437948

01167c2e07d24c658a881586b24ec441

Evolution of mass (upper plot) for reference clouds. The corresponding mass-transportation rates (middle plot) show both loss (Ṁ<0) and gain (Ṁ>0) of mass with the average values (dashed lines) indicating a mass loss. The relative mass loss (lower plot) is higher in low-mass clouds.

2012.10535

hvc_mdot_hist_2.jpg

0.460958

08a5019127f943f6bb980fa9de831877

Distributions of metallicity in the stripped material of massive clouds (upper panel) and low-mass clouds (lower panel) after 20Myr (left column) and after 50Myr (right column).

2012.10535

hvc_metl_hist_mass.jpg

0.497577

ef4f543439a442b1afae28e4a4f66122

Metallicity profiles weighted with the square of particle density along the direction of motion for the massive clouds (upper plot) and low-mass clouds (lower plot) after 50Myr.

2012.10535

hvc_metl_profile.jpg

0.441484

9e02bcd7bcb2482bac7cffae71c58c07

Column densities of Hi for selected massive clouds after 50Myr (M1_dec after 100Myr). The black isodensity contours represent log(N_Hi/)=18,19,… Left column: Actual N_Hi for the clouds. Right column: Diluted N_Hi as being obtained at a distance of 100kpc by a beam of opening angle 9arcmin. The black circle shows the area covered by this beam.

2012.10535

hvc_ncol_plane_xy_HI_5_countours.jpg

0.4946

bd6f5c40198243319ac82ee5f38b6bf8

Evolution of maximum particle density (upper plot) and mass (lower plot) of homogeneous and isothermal clouds with and without self-gravity.

2012.10535

hvc_nmax.jpg

0.448687

b3d85be5de1a466293badb8b7a9c9d62

Distributions of temperature in the stripped material of massive clouds (upper panel) and low-mass clouds (lower panel) with thermal conduction (red dashed lines) and without thermal conduction (black solid lines) after 20Myr (left column) and after 50Myr (right column). The standard deviation in every bin is too low to be visible in the plots.

2012.10535

hvc_temp_hist_mass_3.jpg

0.524737

9ece5a76543347208002b894f64cfdf8

Upper panel: Mean density of model clouds M0 (black solid line) and M1_me (red dashed line). Lower panel: Distributions of temperature within the clouds (left column) and in the stripped material (right column) after 20Myr (upper row) and after 50Myr (lower row). The standard deviation in every bin is too low to be visible in the plots.

2012.10535

hvc_temp_hist_mass_5.jpg

0.450071

affb6f2ad74142628429b2366dc8430f

Free-fall time versus disruption time in fastest massive clouds.

2012.10535

hvc_times.jpg

0.520236

2e710fd7275248cc99d2b95aa15695b7

From top to bottom: Initial radial profiles of density, temperature, pressure, and mass for all model clouds.

2012.10535

initialvalues2.jpg

0.463352

01d957ad98c8422fa233599bac4fbca4

Ratio of saturated to classical mass-transportation rate <cit.>.

2012.10535

mdotratio.jpg

0.5049

bd2ca510f8e7406d83e054381cd365d6

Radial cloud-to-Bonnor-Ebert mass ratios for all model clouds. The x-axis is normalized to the respective radius of each cloud.

2012.10535

prof_mbe.jpg

0.454673

43fdb5b93aae4c5381bbce3674a59444

Both disruption time τ_ dis and distance D_ dis for clouds developing a gravitationally instable region. The light-green stars indicate the onsets of collapse.

2012.10535

time_disruption_2.jpg

0.541475

a1fe436c4f0a459794ebf6ec8fd429ce

Deceleration of model clouds that experience drag.

2012.10535

vdot.jpg

0.439199

c2e79457e57f4bd1b256e1ae1177998e

PA concept in urban and rural areas with the vehicles served by the terrestrial BS and satellite, respectively.

2012.10537

Figure1_0814.jpg

0.554139

a953e097b6ac43588e043ef8e510fde2

E2E throughput of PA system in urban areas with spatially-correlated Rayleigh fading conditions. We set SNR = 21 dB, codeword length = 10^4 channel use, the minimum processing delay of the BS = 5 ms, antenna separation = 1.5 times the wavelength, and carrier frequency = 2.68 GHz.

2012.10537

Figure2_0910.jpg

0.41963

feefb69673cb46fa8c7e688c270544c4

Effect of spatial mismatch on the MRT and DFT BF in an NLoS multi-path propagation environment.

2012.10537

Figure3_0814.jpg

0.443605

7b511e21e78a4c41b8b12a7ef7cdbbbb

Received power at the vehicle for various BF schemes, prediction schemes (ideal prediction, without prediction, i.e., with spatial mismatch), and antenna sizes N, in a spatially-correlated Rayleigh fading environment.

2012.10537

Figure4_1217.jpg

0.453729

745905a9087840fea3f4aa657a107118

E2E throughput of the last wagon in a 10-wagon train served by the satellite, with SNR = 26 dB, codeword length = 10^4 channel use, and the minimum processing delay = 10 ms, antenna separation between the first and the last antenna in the same wagon is 10 times the wavelength, and carrier frequency is 2.68 GHz. The distance for adjacent PA and RA in closest wagons is 0.2 m.

2012.10537

Figure5_0814.jpg

0.392001

a7165da2504945f28b0ef9978959c423

Open symbols: Compilation of decoupling (freeze-out) times, τ_f, plotted as a function of the cubic root of the charged particle density, (dN_ch/dη)^1/3, observed in various collision systems and for a wide range of beam energies <cit.>. Filled symbols: extrapolated values of τ_f corresponding to each of the centrality classes used in our computation of η/s based on Au – Au collisions at √(s_ NN) =0.2 TeV and Pb – Pb collisions at √(s_ NN) =2.76 TeV, measured by the STAR and ALICE collaborations, respectively. Red/blue solid lines represent systematic uncertainties of the polynomial fits to the data.

2012.10542

Fig1.jpg

0.42156

da294b4c785a4feb82c46cc7b87b0a4f

Longitudinal width, σ_Δη, of the G_2^CI correlator vs. the estimated number of participants measured in Au – Au collisions at √(s_ NN) = 200 GeV <cit.> and in Pb – Pb collisions at √(s_ NN)=2.76 TeV <cit.>, reported by the STAR and ALICE collaborations, respectively. Error bars and error boxes represent statistical and systematic uncertainties, respectively.

2012.10542

Fig2.jpg

0.416456

7caba646109344828177d64fb7898519

Values of the shear viscosity per unit of entropy density, η/s, computed in this work, as a function of the cubic root of the charged particle density, (dN_ ch/dη)^1/3, measured in Pb – Pb collisions at √(s_ NN)=2.76 TeV <cit.> and in Au – Au collisions at √(s_ NN) = 200 GeV <cit.>. Error bars and error boxes represent statistical and systematic uncertainties, respectively.

2012.10542

Fig3.jpg

0.450413

c1d25d706dde4f53b5c8f503bb6ba307

Comparison of η/s obtained in this work with a collection of values published since 2000 <cit.>, including both theoretical values and experimental values obtained from direct comparisons of theory to data. Some more details are given in the text.

2012.10542

Fig4.jpg

0.418976

0c2c853f36634d2dac40b655ebbcb209

2012.10544

backdoor-code.jpg

0.623664

ea0d1765c4544a8694900c5c349fd303

2012.10544

backdoor-square.jpg

0.426311

4feb316cc9ef4a268778ddd9e41f1f21

In Panel (a), the tunneling (phase) time through the structure air(LR)^7air, whose transmission coefficient is shown in the lower panel of Figure <ref>. The tunneling times of the wave packet (WP) peaks through the air(LR)^7air structure, obtained from τ=∂ t_aSa/∂ω, at the resonant frequencies, are ≅ -0.28fs. In b), a Gaussian wave packet with the centroid at the resonance with frequency ω≃ 5.32 × 10^15rad/s, and the envelope as shown here is prepared, at t=0, at a distance z_0=L=40l_c from the SL.

2012.10546

Fig10e.jpg

0.448874

54f70d68461d439cbb701231daa4794b

A Gaussian WP with the centroid at the resonance with frequency ω≃ 5.32 × 10^15rad/s is prepared, at t=0, at a distance z_0=L=40l_c from the SL (see the blue curve). The WP moves towards D, passing through the SL (LR)^6, and reaches the point D (see the red curves), as predicted, at t=τ_D=2z_0/v_g+τ≃ 54.02fs. The WP is partially transmitted and partially reflected. Not all frequencies are transmitted because the transmission coefficients at the center of the packet are larger than in the tails. Thus, a dip is formed in the reflected (red) WP. The distance between the peak of the (blue) WP and the dip in the reflected (red) WP depends on τ.

2012.10546

Fig11e.jpg

0.45091

0d020de56cc749a9be513fe7efa13b2c

Incoming, reflected, and transmitted fields at the interface dielectric-conductor. b) The incidence θ_i and the effective refraction angle ψ for a silver slab with the dielectric constant shown here.

2012.10546

Fig1e.jpg

0.438625

b0b13a8ae05e4f188a92b1af3d203165

Transmittance as a function of the frequency through a single slab of silver with thickness d_c=30nm (left-hand side graph) and d_c=80nm (right-hand side graph), for incidence angles θ_i=π/6, π/4, π/3, and slightly less than π/2. For the slab thicknesses considered here, the EM field is highly attenuated below the plasma frequency ω_p= 5.75 10^15rad/s and oscillating for ω > ω_p. Near the incidence angle of π/2, the narrow and isolated resonances correspond to localized surface plasmons. Notice that as the incidence angle grows, the resonances move towards the isolated surface plasmon resonances.

2012.10546

Fig2e.jpg

0.482448

ac8a2e3c1ce144b88e0b4c2123219684

Transmittance through a single slab as a function of the incidence angle θ_i, for three different EM frequencies. For the graphs in the left-hand side column, the silver slab thickness is d_c=80nm, and in the right-hand side column, d_c=1000nm. Notice that the number of oscillations at high frequencies is of the order of d_c/λ.

2012.10546

Fig3e.jpg

0.465238

cd3e3bcf80f845c4b57f22b01b3654df

Transmittance as a function of the frequency through the metallic superlattices (air/silver/air)^n. In a) and b), the thickness of the silver and air slabs is equal to d_a=100nm and d_c=10nm. In a), the transmittance is shown for n=8 and incidence angles θ_i=π/6, π/4, π/3, and slightly less than π/2. Below the plasma frequency ω_p= 5.75 10^15rad/s, the transmittance is an oscillating function of ω. Near the incidence angle of π/2, the transmittance is highly resonant. In b), we repeat the transmittances for θ_i=π/6, π/4, but now for n=16, and we plot also the Kramer condition (black curves) and the dispersion relation of Equation (<ref>) derived in the theory of finite periodic systems (TFPS) (red lines). It is clear that this recurrence relation predicts the bands and the frequencies of all the resonant plasmons.

2012.10546

Fig4e.jpg

0.455987

31aca98afd244ea8b85789d41fa792d7

Transmittance as a function of the frequency through the metallic superlattices (air/silver/air)^n where the silver and dielectric widths are a) small and equal and b) different with the silver slabs' width being larger. The transmittances are shown for different incidence angles θ_i indicated on the graphs. Below the plasma frequency ω_p= 5.75 10^15rad/s, for small widths, the transmittance is independent of the incidence angle, while for larger silver width, the metallic superlattice is almost completely transparent. For frequencies above ω_p, we have wider bands for small layers' widths and thinner bands for larger silver widths.

2012.10546

Fig5e.jpg