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
0.387175	ce3ce71e4c3d447ab34628f346aa22cd	Properties derived from the RHESSI SXR source (red curve, top plot). The emission measure stayed surprisingly constant, although the temperature followed the expected behaviour. The last three panels, which depended on imaging, are shown when the flare fluxes clearly exceed the flux of the on-disk source. The compact thermal source almost doubled in size as it rose above the flaring region and the solar limb. The density, calculated using a filling factor of unity, follows the temperature profile. Long iron ionization times might explain why AIA did not observe the thermal source.	2012.10179	figure_time_ionization.jpg
0.458894	5cd1e854a6ff4390b5ced0d8ef589d34	RHESSI and HEND photon spectra. The RHESSI photon spectrum was fit with both a single thermal component plus a broken power-law.	2012.10179	hend_rhessi_paper.jpg
0.421765	e5956ce29e5b434f968f62294c8711f5	From top-left, clockwise: The flare peak observed directly by STEREO-B 195 Å. RHESSI imaging of the flare peak at 02:33UT showed an extended non-thermal source leading a compact thermal source. No corresponding sources were visible in AIA 193 Å. However, plasma observed at 17 GHz by NoRH resembled the cooler-temperature AIA 304 Å images. The view from STEREO-A was occulted by 3 less than AIA, so it saw slightly more of the flare. An animated version of this figure is available, following the flare evolution. From an hour before the flare erupts, the full-disk view offered by STEREO-B shows the active region changing. During and after the flare peak, short-lived X-ray emission and longer-lived radio emission is seen. The filament eruption is seen especially well from the STEREO-A viewpoint, with EUV plasmoids stretched along the filament axis seven minutes after the peak.	2012.10179	map_peak.jpg
0.536688	7aa99af1b95341eba3a5674f0cd0fbcb	Non-thermal electron density (top) and collisional loss timescales (bottom) calculated for bremsstrahlung emission at different energies.	2012.10179	new_figure.jpg
0.427163	023d56382c0b4941a850fb740668a294	Time evolution of the soft 4-8 keV X-ray source. Contour levels are 50, 70 and 90%, as is common with RHESSI imaging. SXR images, made every 45 seconds starting at 02:32UT, show that the compact source expanded as it moved upward above the flare region. The background images are from AIA 304 Å, corresponding to the timestamps.	2012.10179	sxr_evolution_only.jpg
0.471687	d1e92614f72045b1b1ef1e6e941a6750	Ca_3Ru_2O_7 RIXS spectrum, measured at and at , is compared with Ca_2RuO_4. The Ca_2RuO_4 data is scaled by a factor of 0.7. Modes associated with spin-wave, t_2g and e_g excitations as well as charge-transfer are present in both samples. The low-energy splitting of the excitations in Ca_2RuO_4 are associated with spin-orbit coupling and tetragonal crystal field terms.	2012.10180	fig-comp.jpg
0.507232	2be47eb593084135b7bd543567f1af87	a) Temperature dependence of the magnon at the magnetic zone center, corresponding to the spin wave gap. Above T_N the gap closes and strongly increases in intensity. At weak paramagnetic excitations remain. b) Fitted magnitude and intensity of the spin wave gap as a function of temperature. Results reveal that the gap closes at T_S, while the bulk of the increase in intensity happens above T_N. c) Momentum dependence of the RIXS spectra at collected by varying θ. Each spectrum was normalized to account for self-absorption and shifted vertically for clarity. Overlaid on each spectrum is the data collected at . d) Intensity of the t_2g features as a function of θ plotted in comparison to the spin-orbit excitation in Ca_2RuO_4. Only moderate increase in spectral weight is observed in Ca_3Ru_2O_7 at low θ.	2012.10180	fig-deps.jpg
0.435886	7118d01dfe6146e7bd9b8d1b78765f2b	a) The momentum dependence of the RIXS signal below at from zone center to zone boundary shows the dispersion of the magnon. b) The fitted magnon dispersion captured with RIXS corresponds well with inelastic neutron spectroscopy. Solid white lines plot the acoustic and optical modes of the bilayer spin wave model as discussed in the text, while the colormap represents the intensity calculated from the spin-spin correlation function that takes into account instrumental momentum and energy resolutions. The Ca_2RuO_4 spin wave position determined from Fig. <ref>, and model spin wave based on Ref. Jain17 (dashed white line) are shown for comparison.	2012.10180	fig-mag.jpg
0.409358	f5eb0675e190429086ad47defbde4a09	a) RIXS energy map taken at plots the incident-energy dependence of the energy loss spectra across the Ru L_3 edge. The quasielastic magnon signal and t_2g feature at ∼0.4 eV (horizontal dashed line) resonate around , while e_g excitations around 3 eV strongly resonate at 2841 eV. At higher energies (∼6 eV) charge transfer excitations are present. b) The sample was oriented such that the Ru-O-Ru bonds were aligned 45^∘ to the scattering plane. Outgoing photons k_f were detected at a fixed angle of 90^∘ with respect to the incoming beam k_i. The momentum-transfer was selected by varying the angle θ between k_i and the RuO_2 planes. c) Bulk magnetic response of Ca_3Ru_2O_7, with applied along the b-axis. d) RIXS energy map conducted above T_N (region indicated by a dashed rectangle in (a)) shows a dramatic enhancement of the quasielastic magnetic excitation and an apparent shift of the t_2g spectral weight to smaller energy loss. e-g) Comparison of individual RIXS spectra from (a) and (d). The quasielastic contribution has been removed. Dashed vertical line at 0.4 eV lies at the t_2g maximum at T=70 K. The arrows identify a second mode with an independent incidence-energy response.	2012.10180	fig-rixs.jpg
0.465708	dd45b94773924649b68f2735f3662871	A compact body (circle, left) is in a weak gravitational interaction with a distribution of external matter (ellipse, right). The body zone extends from the body out to r̅ = r̅_ max; gravity is strong deep in the body zone, and it becomes progressively weaker as the boundary is approached. The post-Newtonian zone extends inward to r̅ = r̅_ min; gravity is weak everywhere in this zone. The overlap zone (shaded box) corresponds to the interval r̅_ min < r̅ < r̅_ max.	2012.10184	fig1.jpg
0.438901	efbd3f35951d48a19e05b9cbeb9674b7	The four-month smoothed sunspot-area data are used to identify the Gnevyshev gaps and is shown here for the whole Sun (top), northern hemisphere (middle) and southern hemisphere (bottom) for Cycles 20 and 21.	2012.10190	fig_10.jpg
0.488846	6aabbd80ecaf4635b10d7e4b484e742d	The total area occupied by the sunspots in the northern (plus (+)) and southern (star (∗)) hemisphere over each cycle from 16 – 23. The value of the total sunspot-area is divided by 10 to show the other curves in the same plot. The diamond (♢) and triangle() symbols represent the maximum area occupied by the monthly averaged sunspot-area in each cycle in both the northern and southern hemispheres respectively.	2012.10190	fig_3.jpg
0.463692	d35ab1ba5dc14e4b88d1d726a8afb7da	Cumulative monthly counts of KOSO sunspot-areas in the northern and southern hemispheres for Cycles 16 – 23.	2012.10190	fig_5.jpg
0.42742	c1eda45b213740fc82de6341b788d044	The cumulative N–S asymmetry index is plotted for each Cycle, 16 – 23. The bottom right plot is the cumulative N–S asymmetry index for all the cycles.	2012.10190	fig_6.jpg
0.480416	203bb802d59d49a8aa54537ce0755036	The phase lag between the occurrence of peak between the northern and southern hemisphere in each cycle. The phase lag is shown in months. The solid vertical line represents the time of maximum correlation. The peaks at positive and negative values of lag indicate north and south lead respectively.	2012.10190	fig_8.jpg
0.444634	91acff4e116a42f483d4a17a9a25d901	C=10	2012.10194	big_bounds.jpg
0.442499	1ba0b2a211a546929f13e149f2342f27	Multi-outcome multi-stage design realisation with N=60, C=2.256490, Δ=0. Operating characteristics: α^=0.025, 1-β^=0.827.	2012.10194	contours_MO_2outs.jpg
0.441793	059d06003d3040cfa1cd20406ed81c35	Multi-outcome multi-stage design realisation with N=42, C=1.579395, Δ=0. Operating characteristics: α^=0.025, 1-β^=0.801.	2012.10194	contours_MO_3outs.jpg
0.409998	43302a4d10ae46bab840ab03e333f82b	Composite multi-stage design realisation with N=60, C=3.240066, Δ=0. Operating characteristics: α^=0.025, 1-β^=0.814.	2012.10194	contours_comp_2outs.jpg
0.380771	c4db7520938c4556b699e586d5eb9411	Composite multi-stage design realisation with N=42, C=4.389363, Δ=0. Operating characteristics: α^=0.025, 1-β^=0.836.	2012.10194	contours_comp_3outs.jpg
0.412813	f91988c682e5490caec677cd048de262	Change in ESS_MO/ESS_comp as correlation varies. α=0.025, 1-β=0.8, design parameters J=3, δ_0=0.2, δ_1=0.4. Simulations: 10^5.	2012.10194	cor_plot_2.jpg
0.504409	5dc282445bd746e0ac427a6339fed5fc	Changes in ESS_DtL/ESS_single and ENM_DtL/ENM_single for various designs as correlation ρ is varied. Note: for {K=2, m=1}, K-1=K/2.	2012.10194	dtl_cor.jpg
0.390611	f12e9b9c659f47f2844794d1d0e13a20	Changes in ESS_DtL/ESS_single and ENM_DtL/ENM_single for fixed design with {K=2, K_max=1, m=1} and design is powered for outcome effects δ_β = (0.4, 0.2).	2012.10194	dtl_true_grid2.jpg
0.372922	d737cad58a4c443aba2fe385200dab91	Changes in the probability of rejecting H_0 for the DtL and single-stage designs with {K=2, K_max=1, m=1} and design is powered for outcome effects δ_β=(0.4, 0.2).	2012.10194	dtl_true_grid_preject2.jpg
0.406283	00b7cd6130ac4870b59b1c5c5500c4e8	Change in ESS_MO/ESS_comp as true outcome effect sizes vary. Required type-I error-rate α=0.025, power 1-β=0.8, design parameters {K=2, m=1, J=3}, δ_β 1=0.4, δ_β 2=0.2, ρ=0.3. Simulations: 10^5.	2012.10194	mo_true_grid_2.jpg
0.383517	e487bea8055c4312bed57ebedf272eb9	R(μ=μ_1, μ_2) as true outcome effect sizes vary. Required type-I error-rate α=0.025, power 1-β=0.8, design parameters {K=2, m=1, J=3}, δ_β = (0.4, 0.2). Simulations: 10^5.	2012.10194	mo_true_preject_grid_2.jpg
0.423273	c0e3bea9fc4841819c0c3ff6f468ffd5	Comparison of final rejection regions for multi-outcome design (blue) and composite design (red), for {K=2, m=1, J=1}.	2012.10194	rejection_region_2.jpg
0.408816	66a2bb3a460d48469415b0a731506290	Example of separate stopping approach. Go decision at stage 3.	2012.10194	separate_stopping_go.jpg
0.403915	41955a0c08624e2dbcf6ab208207b5a9	Example of simultaneous stopping approach. Go decision at stage 3.	2012.10194	simultaneous_stopping_go.jpg
0.375231	01f96e251c134f1881961b09cbb01086	Example of simultaneous stopping approach. No go decision at stage 4.	2012.10194	simultaneous_stopping_nogo.jpg
0.432543	19be090aa6ea4a2387bd3bce9af88841	C=2	2012.10194	small_bounds.jpg
0.397879	f527b80f9f964fcda650a5ea3916f000	Host-subtracted light curve of : B (blue squares), g (green circles), U (black diamonds), UVM2 (magenta triangles), UVW1 (orange pentagons) and UVW2 (purple crosses).	2012.10195	19dsg_subLum.jpg
0.459196	c3c9a35e5fef477ea504f146d506ef56	X-ray luminosity (0.3–1 keV) versus the BB temperature. The dashed line represents L∝ kT^4 with arbitrary normalisation. The observed rate of cooling decreases over time.	2012.10195	19dsg_xrayLvsT.jpg
0.503575	c5e2dea90f0b4a339dbc45bb44c6d203	Evolution of the BB temperature from the fit to the NICER (0.3–1 keV) data with time.	2012.10195	19dsg_xrayT.jpg
0.500707	84748ca211f847e5b4d1f25e307b6626	The 0.3-10 keV X-ray luminosity evolution with time of . In blue (circles for detections), the Swift data, in black (squares for detections), the NICER data. The downward arrows represent 3 upper limits.	2012.10195	19dsg_xraylum.jpg
0.489297	9fb3682002534de988741d46276db04d	Position of the persistent narrow emission lines in the host galaxy spectrum of on a BPT diagram. The spectrum used was the late time TNG spectrum at 411 days, when no broad components to the Balmer lines could be detected. The lines separating the different regions come from <cit.>.	2012.10195	BPT_19dsg.jpg
0.430563	a9f89af3f9e1456abe2bf9b8a905c29d	Evolution of the EW of the emission lines. From top to bottom: the broad (red circles) and (blue squares), the (black squares) and the (green circles), and the (magenta circles) emission lines caused by the TDE. Bottom panel: the narrow (red circles) and (blue squares) emission lines originating in the host galaxy. Note that the redshift of caused the region to coincide in wavelength with the prominent telluric feature, rendering the derived EW of uncertain. The number of days passed since the discovery of the transient is given on the X-axis.	2012.10195	EW_19dsg.jpg
0.644297	b2db4b55eb844d779ee1e435a751752f	Example of the fit to the and [Sii] emission lines in the EFOSC2 spectrum taken on 2019 June 10 (62 days after discovery), after performing the host galaxy subtraction. The dashed lines represent different Gaussian components, while the solid line represents the total fitting function. In the bottom panel, the residuals of the fit.	2012.10195	Ha_20190610_epessto_azsub1_err.jpg
0.51483	db8bb74ef78a45ac87986b8e5504242d	Example of the fit to the emission lines in the region of the EFOSC2 spectrum taken on 2019 June 08 (60 days after discovery), after the subtraction of the stellar component process. The dashed lines represent different Gaussian components, while the solid line represents the total fitting function. In the bottom panel, the residuals of the fit.	2012.10195	Hb_20190608_epessto_azsub1_err.jpg
0.524676	88d57c7400fb48ee9fbcf1c9a3d69d69	Example of a fit to the emission lines in the region of the EFOSC2 spectrum taken on 2019 June 08 (60 days after discovery), before performing the host galaxy subtraction. With B. blend we indicate the blend of unresolved lines previously associated with Bowen fluorescence. In the bottom panel, the residuals of the fit are shown.	2012.10195	Hb_20190608_unsub.jpg
0.435195	74a0abc61ccc4e5f81f2e4a93fa04ca6	region of the UVES spectrum with two different fits: in red, the fit with a Gaussian curve, similar to the function used in the other spectra and in blue, the fit using a flat-top Gaussian. The two bottom panels are the residuals of the two fit, the top one for the flat-top Gaussian fit and the bottom one for the Gaussian fit.	2012.10195	Hb_UVES_flatgauss-gauss.jpg
0.452722	9d0939e5102f4669bead126d2a325e88	Fit to the emission lines in the region of the host galaxy subtracted ISIS spectrum obtained 135 days after discovery of . The dashed lines represent different Gaussian components, while the solid line represents the total fitting function. In the bottom panel, the residuals of the fit.	2012.10195	ISIS_Hb_full.jpg
0.494438	c0293b859fec4a5ead68e0ff23c5753f	Comparison between the spectra of (orange) at +245 days and AT2018fyk (black) at +85 days and ASASSN 15oi (red) at +330 days, the other two TDES that have shown Fe lines in their optical spectra.	2012.10195	at2019dsg_Fe2_comparison.jpg
0.442065	9e53f92dbb7241c8b6e93be74760ab40	Evolution of the FWHM of the emission lines. In the first panel, the broad base of (red circles) and (blue squares); in the second panel, (green circles) and (black squares); third panel, (magenta circles). These three lines are due to the TDE. In the bottom panel, the narrow peaks of (red circles) and (blue squares) are shown, these lines are due to host galaxy activity. On the x-axis, the number of days passed since the discovery of the transient.	2012.10195	fwhm_19dsg.jpg
0.413785	d78dadafd9b148378cba5c7bac3bcaf9	Evolution of the shift with respect to the laboratory central wavelength of the emission lines. From top to bottom: the broad (red circles), the (magenta circles), the (black squares), and (green circles) emission lines caused by the TDE. The central wavelength of the broad Hβ component was fixed during the fitting process and is not shown (see text). On the horizontal axis, the number of days passed since the discovery of the transient. Negative values indicate a blue-shift with respect to the rest frame wavelength.	2012.10195	offset_19dsg.jpg
0.395329	7de30daab4334cba8f7df57654b3f454	Ratio of the EW of Balmer lines. Top panel: narrow / in blue squares and broad / in black squares, both fitted with a straight line. The changes in the ratio between the narrow peaks are due to changes in the telluric absorption. Bottom panel: the ratio of the EW of the and the narrow emission line. On the X-axis, the number of days since the discovery of the transient.	2012.10195	ratio_19dsg.jpg
0.43791	69d092d05d194818a973362b3a358b97	Sequence of spectra taken with EFOSC2 (red), DOLORES (blue), Du Pont (cyan), UVES (green), X-shooter (magenta), ISIS (orange) and with ACAM (black). For each spectrum the phase with respect to the transient discovery date (2019 Apr 09, MJD 58582) is reported on the right. The dotted lines indicate the main emission lines and the grey band represents the area affected by telluric absorption. The line is affected by telluric absorption, but this is not shown in the plot for clarity. The UVES spectrum has been smoothed for clarity. All spectra have been corrected for foreground extinction. For plotting purposes, all spectra have been divided by their median value. The X-shooter, ISIS and UVES spectra are not flux calibrated.	2012.10195	spectra3_19dsg.jpg
0.41398	af9ed15f6b1347ec9d52d33b5a0f001e	Sequence of spectra taken with the FLOYDS spectrograph. For each spectrum the phase with respect to the transient discovery date (2019 Apr 09, MJD 58582) is reported on the right. For plotting purposes, all spectra have been divided by their median value.	2012.10195	spectra_19dsg_floyds.jpg
0.39403	defcb43876e641ac80d3638de0f7eccc	Sequence of host-subtracted spectra taken with EFOSC2 (red), DOLORES (blue), Du Pont (cyan), X-shooter (magenta), ISIS (orange) and with ACAM (black). For each spectrum we annotate to the right the number of days passed since the transient discovery (2019 Apr 09, MJD 58 582). The dotted lines indicate the main emission lines and the grey band represents the area affected by telluric absorption. The line is affected by telluric absorption, but this is not shown in the plot for clarity. All spectra have been corrected for foreground extinction. For plotting purposes, all spectra have been normalised. The X-shooter and ISIS spectra are not flux calibrated.	2012.10195	spectra_19dsg_sub.jpg
0.428177	c5eeecfd9bd840a3bdb78c538a80383e	Zoom in of the - region on the late time TNG spectrum (+411 days). In black, the source spectrum; in red, the convolved stellar library (the best-fit host galaxy synthetic spectrum); in green and blue, the subtracted spectrum. The grey vertical bands indicate the regions of the spectrum excluded from the fit. The - region is almost completely masked during the fit, while the rest of the spectrum, except for the line, is considered for the template fitting.	2012.10195	zoom_ppxf_TNG.jpg
0.411756	9fc7e03d95bf41ba8e10f97593d25735	(a) Identification of the area with Vickers indents (b) EBSD map (c,d) Correction of geometrical artefacts due to the SEM configuration used for EBSD measurements (e) 3D reconstructed microstructure: each color corresponds to a crystallographic orientation (f) Cross-sections of the 3D microstructure (see Appendix A for cross-sections without corrections of the geometrical distortions)	2012.10198	EBSD3D.jpg
0.413606	ec87f10461ef4f299be1b4c24d0c8165	Visualization of the GB (in gray) detected by the software on the first EBSD map, and projection of the intergranular cracks (in red)	2012.10198	GB.jpg
0.451739	181e41acd16f4bf7ae61224562d84b51	Stacking of successive 2D EBSD maps to reconstruct the 3D microstructure, (a) without and (b) with the use of Eq. <ref>	2012.10198	appendix.jpg
0.414705	b0ce0bd615204a0aa7cf97f24dc0c316	(a) SEM image of the area identified by the indents in the EBSD configuration (70^∘ tilt angle) (b) Correction of the SEM image	2012.10198	ebsd_correction.jpg
0.459903	695d95c8165c47939c075394288bfaee	(a, b) Detection of GB on the free-surface plane. Black lines denote calculated in-plane GB slopes, red points the identified triple points and white points a 3× 3 neighborhood of each triple point where GB detection is avoided (b) Enlarged section of the top plane sketching the calculation of the in-plane GB slope (c,d) GB on top and bottom planes: red lines denote vertical planes to detect the out-of-plane GB slope k_1 (e).	2012.10198	gbsamir.jpg
0.39193	79961b1695f1408c9894306a91474efb	The ATE1 bias, bucket availability correlation, and bucket sampling correlation plotted as a function of δ, for settings 3-6. The parameters for each setting are given in Table <ref>.	2012.10202	plots.jpg
0.435586	ccb70aa31a4c47c2bf21ebc92b53e4b0	Box plot of classifier accuracy with 4 features: under strategic reporting	2012.10203	0601_4box_pcc_bal_fixed.jpg
0.438979	f74b102b811d4aa282a77292051852c5	Box plot of classifier accuracy with all features: under strategic reporting	2012.10203	0925_4box_pcc_bal_NFS.jpg
0.449238	f815b4f43ec7413d82320bb0f262fe7e	Selected classifier accuracy w/ strategic behavior, balanced datasets, each classifier selects 4 best features	2012.10203	box_FS_BAL_ACC.jpg
0.422939	aa698e32058d496a8990a6484f69662e	Selected classifier AUC w/ strategic behavior, balanced datasets, each classifier selects 4 best features	2012.10203	box_FS_BAL_AUC.jpg
0.349904	07394dadc5e04c53b8c5963ab0b5983a	Selected classifier accuracy w/ strategic behavior, unbalanced datasets, each classifier selects 4 best features	2012.10203	box_FS_UNBAL_ACC.jpg
0.444041	83f70596a64a4e9aae15467bf22b03d8	Selected classifier AUC w/ strategic behavior, unbalanced datasets, each classifier selects 4 best features	2012.10203	box_FS_UNBAL_AUC.jpg
0.451692	1b6d8ea775154ca0bf5ffa64ab1d13c7	Selected classifier accuracy w/ strategic behavior, balanced datasets, all features	2012.10203	box_NFS_BAL_ACC.jpg
0.460784	32ba70686d034d13a0885ec1b180b348	Selected classifier AUC w/ strategic behavior, balanced datasets, all features	2012.10203	box_NFS_BAL_AUC.jpg
0.498131	6a6cb1f01be54b1cb398d0ff1142d3f5	Selected classifier accuracy w/ strategic behavior, unbalanced datasets, all features	2012.10203	box_NFS_UNBAL_ACC.jpg
0.437193	09a071be953447679264e44139a749ee	Selected classifier AUC w/ strategic behavior, unbalanced datasets, all features	2012.10203	box_NFS_UNBAL_AUC.jpg
0.450077	6a11456a1bec450887c69403222e226a	Selected classifier accuracy w/ strategic behavior, balanced datasets, pre-selected 4 features	2012.10203	box_preFS_BAL_ACC.jpg
0.419558	43fbde08c36a445ca2707a410416ba27	Selected classifier AUC w/ strategic behavior, balanced datasets, pre-selected 4 features	2012.10203	box_preFS_BAL_AUC.jpg
0.383666	60c819c6389e46bf8e62b9dfc1f1dc7a	Selected classifier accuracy w/ strategic behavior, unbalanced datasets, pre-selected 4 features	2012.10203	box_preFS_UNBAL_ACC.jpg
0.412085	1f59b1b1291e4c39b2c6792ceaaa805a	Selected classifier AUC w/ strategic behavior, unbalanced datasets, pre-selected 4 features	2012.10203	box_preFS_UNBAL_AUC.jpg
0.39767	2522cf53b06d419d9e75b229e23f1f22	Box plot of classifier accuracy with 4 features: under strategic reporting	2012.10203	main_FS_vert.jpg
0.38999	a92e2379b98b40bba8bb828f317a73d3	Box plot of classifier accuracy with all features: under strategic reporting	2012.10203	main_NFS_vert.jpg
0.49387	cb446a8f0c424218b8037154dcec98b7	These are possible plots of the quadratic function h(w)(Eq. (<ref>)) with u<1(left) and u>1(right). The red-color region represents for the interval of w giving h(w)<0. Alternatively speaking, the Heaviside function does not vanish in these w intervals marked by red. In the case of u<1, it is possible that h(w) has no roots, which means h(w) is always positive. However, this case is not what we are interested in. The other case of u>1 always has real roots with opposite signs.	2012.10204	analysis_of_delta_function.jpg
0.443767	4ab6546f1c6143f79a6d55b2f6e605cd	(color online): Schematic representation of the system under study. An atom at a distance z over a metallic plate moves at constant speed v parallel to the surface. The quantum frictional force opposes the motion of the atom. Inset: the internal state of the atom is a two-level system for which the excited state is threefold degenerate, in analogy with the 1s and 2p levels of the hydrogen atom.	2012.10204	figure1.jpg
0.495548	505f5aa9177344af9cf85e858c4d8d48	(Color online): Frictional force as a function of the relative velocity between the atom and the plate. The force is normalized by the static Casimir-Polder force between a perfectly conducting plate and an atom, and the velocity is normalized as u = v / β, where β is the sound speed in the material. We have considered parameters such that ω = ω_p / ω_b = 1, 5 and z̃ = z ω_p / β = 10, 100. We have taken the sound speed as β = 10^6 m/s.	2012.10204	plot.jpg
0.51978	5f72a39ea6954040852b3438478a487e	Influence of the learned decision boundary.	2012.10209	Aba_2.jpg
0.419799	3ad7e0a8a21f426fb147040645963791	The boundary learning process.	2012.10209	curve.jpg
0.456331	80df1a74d2ca454d859229f1a37c72ab	An example of open intent classification. Book flight and Restaurant reservation are two known intents. We should identify them correctly while detecting the sentences with the open intent.	2012.10209	example.jpg
0.42673	b87fad3ff7374681ba431c54af96f830	The model architecture of our approach. Firstly, we use BERT to extract intent features and pre-train the model with labeled samples. Then, we initialize the centroids {c_i}_i=1^K and the radius of decision boundaries {Δ_i}_i=1^K for each known class. Next, we propose the boundary loss to learn tight decision boundaries adaptive to the known intent feature space. Finally, we perform open classification with the learned decision boundaries to both identify known classes and detect the open class.	2012.10209	model.jpg
0.400759	49a2ab313ab6459c9476f2c59e7bd1b7	Error probability for (a) internet_source and (b) dangerous.	2012.10211	error_prob.jpg
0.453349	cfbf84bd71d64bac897ceb23790a3647	Histogram of Bernoulli misclassification test statistic for (a) internet_sourced and (b) dangerous. The horizontal line marks a value of λ=1.	2012.10211	ll_bern.jpg
0.431621	7a3b859fe69e46d483cb5cec4eb8a473	The relation matrix for (a) internet_sourced, (b) dangerous. Rows correspond to the error messages listed in Table <ref>. Columns correspond to files. In each matrix entry, white indicates no error, black indicates error.	2012.10211	matrix.jpg
0.501633	a0862a10337544e0a8e9380895014435	Correlation matrix between parser behaviors. Rows and columns are sorted according to median message correlation.	2012.10211	parser_correlation.jpg
0.461851	ca2264d0749f44928873ccd112f251e5	Parsers placed in three dimensions according to principal components analysis of the union of internet_sourced and dangerous.	2012.10211	parsers_pca.jpg
0.434125	d29dc51bf42e48faad2e422d995310a8	Principal components plots for (a) internet_sourced and (b) dangerous. Black indicate “valid”, and gray indicates “reject”. The axes correspond to the three principal vectors, and so are not plotted on the same scale.	2012.10211	pca.jpg
0.396956	3bd14b55e5234d05b1ad16172e5f69df	Receiver operating curves for Bernoulli likelihood ratio misclassification statistic for (a) internet_sourced and (b) dangerous.	2012.10211	roc_bern.jpg
0.48948	0f41f8404cad4563ab184ddd55bbfdb3	Scree plots for (a) internet_sourced and (b) dangerous.	2012.10211	scree.jpg
0.469706	64fbcb69f1df41a096af43efc83742dd	Illustration of our method. (1) Traders predict the return of assets using simple formulas. Companies manage Traders and combine Traders' predictions into one value. The Company algorithm consists of four functions: (2) prediction by aggregation, (3) education of bad Traders, (4) dismissal of bad Traders, and (5) recruitment of new Traders.	2012.10215	fig_paper_idea.jpg
0.388369	b28a5bd4bf954955aa9b9316450ef652	Cumulative returns on UK market	2012.10215	fig_tc_only.jpg
0.377582	8e9218f0f107417d9ad00632a3f796a8	Cumulative returns achieved by individual Traders extracted from Companies. “TC” is the overall performance of our method. “TC # term = M” stands for a single best performing Trader extracted from a Company with fixed parameter M.	2012.10215	fig_trader.jpg
0.409988	00adf096b86e41b8a63b122816dffcf6	Cumulative returns on US market in the online prediction setting	2012.10215	fig_us_daily.jpg
0.451308	ff1e324ad17c477086dbdfb7df9ad156	Cumulative returns on US market	2012.10215	fig_us_offline.jpg
0.482226	f2498f80c7ba48968288c99223b7447e	Error on subsets of varying sizes (): x axis denotes points in ascending order of scores. y axis denotes error accrued on the subset containing points up to x.	2012.10216	final_pf_vs_fp_edited.jpg
0.502069	8898b18c2c9a4d62ad0f5fbe2a10f8a3	Given two groups Blue and Yellow (with true labels as shown), we have to choose just between the two classifiers x=0 and y=0. The Blue group is inherently harder to classify. Equalized odds makes us choose the classifier x=0, thereby hurting the Yellow group. We could choose y=0 with no aggregate effect on Blue, doing much better on Yellow.	2012.10216	harm_example.jpg
0.449404	c21be706cace45109f7b420e590d42db	Illustration of the real-time video streaming process in a cell with n users.	2012.10219	System_model_final_update_2.jpg
0.456595	405441623ead466d86960d1a386a638f	The impact of number of users on playout rate.	2012.10219	bar_playout_2.jpg
0.48451	0276b941e9f84944af1570522c5745ee	Drop rate as a function of buffer size.	2012.10219	buffer_seconds.jpg
0.510137	40cde20c8eec4779813fca3bee9ca7ca	QoE for our approach and ABR.	2012.10219	qoe_1.jpg
0.47262	366150e1048944e6a3ddb452cad3b5e9	Number of users with maximum playout rate.	2012.10219	tech_report.jpg
0.466404	81dd2df0127545d28602807bcfdcd6b4	The playout rate for two classes of users.	2012.10219	two_classes_delta_2.jpg

0.387175

ce3ce71e4c3d447ab34628f346aa22cd

Properties derived from the RHESSI SXR source (red curve, top plot). The emission measure stayed surprisingly constant, although the temperature followed the expected behaviour. The last three panels, which depended on imaging, are shown when the flare fluxes clearly exceed the flux of the on-disk source. The compact thermal source almost doubled in size as it rose above the flaring region and the solar limb. The density, calculated using a filling factor of unity, follows the temperature profile. Long iron ionization times might explain why AIA did not observe the thermal source.

2012.10179

figure_time_ionization.jpg

0.458894

5cd1e854a6ff4390b5ced0d8ef589d34

RHESSI and HEND photon spectra. The RHESSI photon spectrum was fit with both a single thermal component plus a broken power-law.

2012.10179

hend_rhessi_paper.jpg

0.421765

e5956ce29e5b434f968f62294c8711f5

From top-left, clockwise: The flare peak observed directly by STEREO-B 195 Å. RHESSI imaging of the flare peak at 02:33UT showed an extended non-thermal source leading a compact thermal source. No corresponding sources were visible in AIA 193 Å. However, plasma observed at 17 GHz by NoRH resembled the cooler-temperature AIA 304 Å images. The view from STEREO-A was occulted by 3 less than AIA, so it saw slightly more of the flare. An animated version of this figure is available, following the flare evolution. From an hour before the flare erupts, the full-disk view offered by STEREO-B shows the active region changing. During and after the flare peak, short-lived X-ray emission and longer-lived radio emission is seen. The filament eruption is seen especially well from the STEREO-A viewpoint, with EUV plasmoids stretched along the filament axis seven minutes after the peak.

2012.10179

map_peak.jpg

0.536688

7aa99af1b95341eba3a5674f0cd0fbcb

Non-thermal electron density (top) and collisional loss timescales (bottom) calculated for bremsstrahlung emission at different energies.

2012.10179

new_figure.jpg

0.427163

023d56382c0b4941a850fb740668a294

Time evolution of the soft 4-8 keV X-ray source. Contour levels are 50, 70 and 90%, as is common with RHESSI imaging. SXR images, made every 45 seconds starting at 02:32UT, show that the compact source expanded as it moved upward above the flare region. The background images are from AIA 304 Å, corresponding to the timestamps.

2012.10179

sxr_evolution_only.jpg

0.471687

d1e92614f72045b1b1ef1e6e941a6750

Ca_3Ru_2O_7 RIXS spectrum, measured at and at , is compared with Ca_2RuO_4. The Ca_2RuO_4 data is scaled by a factor of 0.7. Modes associated with spin-wave, t_2g and e_g excitations as well as charge-transfer are present in both samples. The low-energy splitting of the excitations in Ca_2RuO_4 are associated with spin-orbit coupling and tetragonal crystal field terms.

2012.10180

fig-comp.jpg

0.507232

2be47eb593084135b7bd543567f1af87

a) Temperature dependence of the magnon at the magnetic zone center, corresponding to the spin wave gap. Above T_N the gap closes and strongly increases in intensity. At weak paramagnetic excitations remain. b) Fitted magnitude and intensity of the spin wave gap as a function of temperature. Results reveal that the gap closes at T_S, while the bulk of the increase in intensity happens above T_N. c) Momentum dependence of the RIXS spectra at collected by varying θ. Each spectrum was normalized to account for self-absorption and shifted vertically for clarity. Overlaid on each spectrum is the data collected at . d) Intensity of the t_2g features as a function of θ plotted in comparison to the spin-orbit excitation in Ca_2RuO_4. Only moderate increase in spectral weight is observed in Ca_3Ru_2O_7 at low θ.

2012.10180

fig-deps.jpg

0.435886

7118d01dfe6146e7bd9b8d1b78765f2b

a) The momentum dependence of the RIXS signal below at from zone center to zone boundary shows the dispersion of the magnon. b) The fitted magnon dispersion captured with RIXS corresponds well with inelastic neutron spectroscopy. Solid white lines plot the acoustic and optical modes of the bilayer spin wave model as discussed in the text, while the colormap represents the intensity calculated from the spin-spin correlation function that takes into account instrumental momentum and energy resolutions. The Ca_2RuO_4 spin wave position determined from Fig. <ref>, and model spin wave based on Ref. Jain17 (dashed white line) are shown for comparison.

2012.10180

fig-mag.jpg

0.409358

f5eb0675e190429086ad47defbde4a09

a) RIXS energy map taken at plots the incident-energy dependence of the energy loss spectra across the Ru L_3 edge. The quasielastic magnon signal and t_2g feature at ∼0.4 eV (horizontal dashed line) resonate around , while e_g excitations around 3 eV strongly resonate at 2841 eV. At higher energies (∼6 eV) charge transfer excitations are present. b) The sample was oriented such that the Ru-O-Ru bonds were aligned 45^∘ to the scattering plane. Outgoing photons k_f were detected at a fixed angle of 90^∘ with respect to the incoming beam k_i. The momentum-transfer was selected by varying the angle θ between k_i and the RuO_2 planes. c) Bulk magnetic response of Ca_3Ru_2O_7, with applied along the b-axis. d) RIXS energy map conducted above T_N (region indicated by a dashed rectangle in (a)) shows a dramatic enhancement of the quasielastic magnetic excitation and an apparent shift of the t_2g spectral weight to smaller energy loss. e-g) Comparison of individual RIXS spectra from (a) and (d). The quasielastic contribution has been removed. Dashed vertical line at 0.4 eV lies at the t_2g maximum at T=70 K. The arrows identify a second mode with an independent incidence-energy response.

2012.10180

fig-rixs.jpg

0.465708

dd45b94773924649b68f2735f3662871

A compact body (circle, left) is in a weak gravitational interaction with a distribution of external matter (ellipse, right). The body zone extends from the body out to r̅ = r̅_ max; gravity is strong deep in the body zone, and it becomes progressively weaker as the boundary is approached. The post-Newtonian zone extends inward to r̅ = r̅_ min; gravity is weak everywhere in this zone. The overlap zone (shaded box) corresponds to the interval r̅_ min < r̅ < r̅_ max.

2012.10184

fig1.jpg

0.438901

efbd3f35951d48a19e05b9cbeb9674b7

The four-month smoothed sunspot-area data are used to identify the Gnevyshev gaps and is shown here for the whole Sun (top), northern hemisphere (middle) and southern hemisphere (bottom) for Cycles 20 and 21.

2012.10190

fig_10.jpg

0.488846

6aabbd80ecaf4635b10d7e4b484e742d

The total area occupied by the sunspots in the northern (plus (+)) and southern (star (∗)) hemisphere over each cycle from 16 – 23. The value of the total sunspot-area is divided by 10 to show the other curves in the same plot. The diamond (♢) and triangle() symbols represent the maximum area occupied by the monthly averaged sunspot-area in each cycle in both the northern and southern hemispheres respectively.

2012.10190

fig_3.jpg

0.463692

d35ab1ba5dc14e4b88d1d726a8afb7da

Cumulative monthly counts of KOSO sunspot-areas in the northern and southern hemispheres for Cycles 16 – 23.

2012.10190

fig_5.jpg

0.42742

c1eda45b213740fc82de6341b788d044

The cumulative N–S asymmetry index is plotted for each Cycle, 16 – 23. The bottom right plot is the cumulative N–S asymmetry index for all the cycles.

2012.10190

fig_6.jpg

0.480416

203bb802d59d49a8aa54537ce0755036

The phase lag between the occurrence of peak between the northern and southern hemisphere in each cycle. The phase lag is shown in months. The solid vertical line represents the time of maximum correlation. The peaks at positive and negative values of lag indicate north and south lead respectively.

2012.10190

fig_8.jpg

0.444634

91acff4e116a42f483d4a17a9a25d901

C=10

2012.10194

big_bounds.jpg

0.442499

1ba0b2a211a546929f13e149f2342f27

Multi-outcome multi-stage design realisation with N=60, C=2.256490, Δ=0. Operating characteristics: α^*=0.025, 1-β^*=0.827.

2012.10194

contours_MO_2outs.jpg

0.441793

059d06003d3040cfa1cd20406ed81c35

Multi-outcome multi-stage design realisation with N=42, C=1.579395, Δ=0. Operating characteristics: α^*=0.025, 1-β^*=0.801.

2012.10194

contours_MO_3outs.jpg

0.409998

43302a4d10ae46bab840ab03e333f82b

Composite multi-stage design realisation with N=60, C=3.240066, Δ=0. Operating characteristics: α^*=0.025, 1-β^*=0.814.

2012.10194

contours_comp_2outs.jpg

0.380771

c4db7520938c4556b699e586d5eb9411

Composite multi-stage design realisation with N=42, C=4.389363, Δ=0. Operating characteristics: α^*=0.025, 1-β^*=0.836.

2012.10194

contours_comp_3outs.jpg

0.412813

f91988c682e5490caec677cd048de262

Change in ESS_MO/ESS_comp as correlation varies. α=0.025, 1-β=0.8, design parameters J=3, δ_0=0.2, δ_1=0.4. Simulations: 10^5.

2012.10194

cor_plot_2.jpg

0.504409

5dc282445bd746e0ac427a6339fed5fc

Changes in ESS_DtL/ESS_single and ENM_DtL/ENM_single for various designs as correlation ρ is varied. Note: for {K=2, m=1}, K-1=K/2.

2012.10194

dtl_cor.jpg

0.390611

f12e9b9c659f47f2844794d1d0e13a20

Changes in ESS_DtL/ESS_single and ENM_DtL/ENM_single for fixed design with {K=2, K_max=1, m=1} and design is powered for outcome effects δ_β = (0.4, 0.2).

2012.10194

dtl_true_grid2.jpg

0.372922

d737cad58a4c443aba2fe385200dab91

Changes in the probability of rejecting H_0 for the DtL and single-stage designs with {K=2, K_max=1, m=1} and design is powered for outcome effects δ_β=(0.4, 0.2).

2012.10194

dtl_true_grid_preject2.jpg

0.406283

00b7cd6130ac4870b59b1c5c5500c4e8

Change in ESS_MO/ESS_comp as true outcome effect sizes vary. Required type-I error-rate α=0.025, power 1-β=0.8, design parameters {K=2, m=1, J=3}, δ_β 1=0.4, δ_β 2=0.2, ρ=0.3. Simulations: 10^5.

2012.10194

mo_true_grid_2.jpg

0.383517

e487bea8055c4312bed57ebedf272eb9

R(μ=μ_1, μ_2) as true outcome effect sizes vary. Required type-I error-rate α=0.025, power 1-β=0.8, design parameters {K=2, m=1, J=3}, δ_β = (0.4, 0.2). Simulations: 10^5.

2012.10194

mo_true_preject_grid_2.jpg

0.423273

c0e3bea9fc4841819c0c3ff6f468ffd5

Comparison of final rejection regions for multi-outcome design (blue) and composite design (red), for {K=2, m=1, J=1}.

2012.10194

rejection_region_2.jpg

0.408816

66a2bb3a460d48469415b0a731506290

Example of separate stopping approach. Go decision at stage 3.

2012.10194

separate_stopping_go.jpg

0.403915

41955a0c08624e2dbcf6ab208207b5a9

Example of simultaneous stopping approach. Go decision at stage 3.

2012.10194

simultaneous_stopping_go.jpg

0.375231

01f96e251c134f1881961b09cbb01086

Example of simultaneous stopping approach. No go decision at stage 4.

2012.10194

simultaneous_stopping_nogo.jpg

0.432543

19be090aa6ea4a2387bd3bce9af88841

C=2

2012.10194

small_bounds.jpg

0.397879

f527b80f9f964fcda650a5ea3916f000

Host-subtracted light curve of : B (blue squares), g (green circles), U (black diamonds), UVM2 (magenta triangles), UVW1 (orange pentagons) and UVW2 (purple crosses).

2012.10195

19dsg_subLum.jpg

0.459196

c3c9a35e5fef477ea504f146d506ef56

X-ray luminosity (0.3–1 keV) versus the BB temperature. The dashed line represents L∝ kT^4 with arbitrary normalisation. The observed rate of cooling decreases over time.

2012.10195

19dsg_xrayLvsT.jpg

0.503575

c5e2dea90f0b4a339dbc45bb44c6d203

Evolution of the BB temperature from the fit to the NICER (0.3–1 keV) data with time.

2012.10195

19dsg_xrayT.jpg

0.500707

84748ca211f847e5b4d1f25e307b6626

The 0.3-10 keV X-ray luminosity evolution with time of . In blue (circles for detections), the Swift data, in black (squares for detections), the NICER data. The downward arrows represent 3 upper limits.

2012.10195

19dsg_xraylum.jpg

0.489297

9fb3682002534de988741d46276db04d

Position of the persistent narrow emission lines in the host galaxy spectrum of on a BPT diagram. The spectrum used was the late time TNG spectrum at 411 days, when no broad components to the Balmer lines could be detected. The lines separating the different regions come from <cit.>.

2012.10195

BPT_19dsg.jpg

0.430563

a9f89af3f9e1456abe2bf9b8a905c29d

Evolution of the EW of the emission lines. From top to bottom: the broad (red circles) and (blue squares), the (black squares) and the (green circles), and the (magenta circles) emission lines caused by the TDE. Bottom panel: the narrow (red circles) and (blue squares) emission lines originating in the host galaxy. Note that the redshift of caused the region to coincide in wavelength with the prominent telluric feature, rendering the derived EW of uncertain. The number of days passed since the discovery of the transient is given on the X-axis.

2012.10195

EW_19dsg.jpg

0.644297

b2db4b55eb844d779ee1e435a751752f

Example of the fit to the and [Sii] emission lines in the EFOSC2 spectrum taken on 2019 June 10 (62 days after discovery), after performing the host galaxy subtraction. The dashed lines represent different Gaussian components, while the solid line represents the total fitting function. In the bottom panel, the residuals of the fit.

2012.10195

Ha_20190610_epessto_azsub1_err.jpg

0.51483

db8bb74ef78a45ac87986b8e5504242d

Example of the fit to the emission lines in the region of the EFOSC2 spectrum taken on 2019 June 08 (60 days after discovery), after the subtraction of the stellar component process. The dashed lines represent different Gaussian components, while the solid line represents the total fitting function. In the bottom panel, the residuals of the fit.

2012.10195

Hb_20190608_epessto_azsub1_err.jpg

0.524676

88d57c7400fb48ee9fbcf1c9a3d69d69

Example of a fit to the emission lines in the region of the EFOSC2 spectrum taken on 2019 June 08 (60 days after discovery), before performing the host galaxy subtraction. With B. blend we indicate the blend of unresolved lines previously associated with Bowen fluorescence. In the bottom panel, the residuals of the fit are shown.

2012.10195

Hb_20190608_unsub.jpg

0.435195

74a0abc61ccc4e5f81f2e4a93fa04ca6

region of the UVES spectrum with two different fits: in red, the fit with a Gaussian curve, similar to the function used in the other spectra and in blue, the fit using a flat-top Gaussian. The two bottom panels are the residuals of the two fit, the top one for the flat-top Gaussian fit and the bottom one for the Gaussian fit.

2012.10195

Hb_UVES_flatgauss-gauss.jpg

0.452722

9d0939e5102f4669bead126d2a325e88

Fit to the emission lines in the region of the host galaxy subtracted ISIS spectrum obtained 135 days after discovery of . The dashed lines represent different Gaussian components, while the solid line represents the total fitting function. In the bottom panel, the residuals of the fit.

2012.10195

ISIS_Hb_full.jpg

0.494438

c0293b859fec4a5ead68e0ff23c5753f

Comparison between the spectra of (orange) at +245 days and AT2018fyk (black) at +85 days and ASASSN 15oi (red) at +330 days, the other two TDES that have shown Fe lines in their optical spectra.

2012.10195

at2019dsg_Fe2_comparison.jpg

0.442065

9e53f92dbb7241c8b6e93be74760ab40

Evolution of the FWHM of the emission lines. In the first panel, the broad base of (red circles) and (blue squares); in the second panel, (green circles) and (black squares); third panel, (magenta circles). These three lines are due to the TDE. In the bottom panel, the narrow peaks of (red circles) and (blue squares) are shown, these lines are due to host galaxy activity. On the x-axis, the number of days passed since the discovery of the transient.

2012.10195

fwhm_19dsg.jpg

0.413785

d78dadafd9b148378cba5c7bac3bcaf9

Evolution of the shift with respect to the laboratory central wavelength of the emission lines. From top to bottom: the broad (red circles), the (magenta circles), the (black squares), and (green circles) emission lines caused by the TDE. The central wavelength of the broad Hβ component was fixed during the fitting process and is not shown (see text). On the horizontal axis, the number of days passed since the discovery of the transient. Negative values indicate a blue-shift with respect to the rest frame wavelength.

2012.10195

offset_19dsg.jpg

0.395329

7de30daab4334cba8f7df57654b3f454

Ratio of the EW of Balmer lines. Top panel: narrow / in blue squares and broad / in black squares, both fitted with a straight line. The changes in the ratio between the narrow peaks are due to changes in the telluric absorption. Bottom panel: the ratio of the EW of the and the narrow emission line. On the X-axis, the number of days since the discovery of the transient.

2012.10195

ratio_19dsg.jpg

0.43791

69d092d05d194818a973362b3a358b97

Sequence of spectra taken with EFOSC2 (red), DOLORES (blue), Du Pont (cyan), UVES (green), X-shooter (magenta), ISIS (orange) and with ACAM (black). For each spectrum the phase with respect to the transient discovery date (2019 Apr 09, MJD 58582) is reported on the right. The dotted lines indicate the main emission lines and the grey band represents the area affected by telluric absorption. The line is affected by telluric absorption, but this is not shown in the plot for clarity. The UVES spectrum has been smoothed for clarity. All spectra have been corrected for foreground extinction. For plotting purposes, all spectra have been divided by their median value. The X-shooter, ISIS and UVES spectra are not flux calibrated.

2012.10195

spectra3_19dsg.jpg

0.41398

af9ed15f6b1347ec9d52d33b5a0f001e

Sequence of spectra taken with the FLOYDS spectrograph. For each spectrum the phase with respect to the transient discovery date (2019 Apr 09, MJD 58582) is reported on the right. For plotting purposes, all spectra have been divided by their median value.

2012.10195

spectra_19dsg_floyds.jpg

0.39403

defcb43876e641ac80d3638de0f7eccc

Sequence of host-subtracted spectra taken with EFOSC2 (red), DOLORES (blue), Du Pont (cyan), X-shooter (magenta), ISIS (orange) and with ACAM (black). For each spectrum we annotate to the right the number of days passed since the transient discovery (2019 Apr 09, MJD 58 582). The dotted lines indicate the main emission lines and the grey band represents the area affected by telluric absorption. The line is affected by telluric absorption, but this is not shown in the plot for clarity. All spectra have been corrected for foreground extinction. For plotting purposes, all spectra have been normalised. The X-shooter and ISIS spectra are not flux calibrated.

2012.10195

spectra_19dsg_sub.jpg

0.428177

c5eeecfd9bd840a3bdb78c538a80383e

Zoom in of the - region on the late time TNG spectrum (+411 days). In black, the source spectrum; in red, the convolved stellar library (the best-fit host galaxy synthetic spectrum); in green and blue, the subtracted spectrum. The grey vertical bands indicate the regions of the spectrum excluded from the fit. The - region is almost completely masked during the fit, while the rest of the spectrum, except for the line, is considered for the template fitting.

2012.10195

zoom_ppxf_TNG.jpg

0.411756

9fc7e03d95bf41ba8e10f97593d25735

(a) Identification of the area with Vickers indents (b) EBSD map (c,d) Correction of geometrical artefacts due to the SEM configuration used for EBSD measurements (e) 3D reconstructed microstructure: each color corresponds to a crystallographic orientation (f) Cross-sections of the 3D microstructure (see Appendix A for cross-sections without corrections of the geometrical distortions)

2012.10198

EBSD3D.jpg

0.413606

ec87f10461ef4f299be1b4c24d0c8165

Visualization of the GB (in gray) detected by the software on the first EBSD map, and projection of the intergranular cracks (in red)

2012.10198

GB.jpg

0.451739

181e41acd16f4bf7ae61224562d84b51

Stacking of successive 2D EBSD maps to reconstruct the 3D microstructure, (a) without and (b) with the use of Eq. <ref>

2012.10198

appendix.jpg

0.414705

b0ce0bd615204a0aa7cf97f24dc0c316

(a) SEM image of the area identified by the indents in the EBSD configuration (70^∘ tilt angle) (b) Correction of the SEM image

2012.10198

ebsd_correction.jpg

0.459903

695d95c8165c47939c075394288bfaee

(a, b) Detection of GB on the free-surface plane. Black lines denote calculated in-plane GB slopes, red points the identified triple points and white points a 3× 3 neighborhood of each triple point where GB detection is avoided (b) Enlarged section of the top plane sketching the calculation of the in-plane GB slope (c,d) GB on top and bottom planes: red lines denote vertical planes to detect the out-of-plane GB slope k_1 (e).

2012.10198

gbsamir.jpg

0.39193

79961b1695f1408c9894306a91474efb

The ATE1 bias, bucket availability correlation, and bucket sampling correlation plotted as a function of δ, for settings 3-6. The parameters for each setting are given in Table <ref>.

2012.10202

plots.jpg

0.435586

ccb70aa31a4c47c2bf21ebc92b53e4b0

Box plot of classifier accuracy with 4 features: under strategic reporting

2012.10203

0601_4box_pcc_bal_fixed.jpg

0.438979

f74b102b811d4aa282a77292051852c5

Box plot of classifier accuracy with all features: under strategic reporting

2012.10203

0925_4box_pcc_bal_NFS.jpg

0.449238

f815b4f43ec7413d82320bb0f262fe7e

Selected classifier accuracy w/ strategic behavior, balanced datasets, each classifier selects 4 best features

2012.10203

box_FS_BAL_ACC.jpg

0.422939

aa698e32058d496a8990a6484f69662e

Selected classifier AUC w/ strategic behavior, balanced datasets, each classifier selects 4 best features

2012.10203

box_FS_BAL_AUC.jpg

0.349904

07394dadc5e04c53b8c5963ab0b5983a

Selected classifier accuracy w/ strategic behavior, unbalanced datasets, each classifier selects 4 best features

2012.10203

box_FS_UNBAL_ACC.jpg

0.444041

83f70596a64a4e9aae15467bf22b03d8

Selected classifier AUC w/ strategic behavior, unbalanced datasets, each classifier selects 4 best features

2012.10203

box_FS_UNBAL_AUC.jpg

0.451692

1b6d8ea775154ca0bf5ffa64ab1d13c7

Selected classifier accuracy w/ strategic behavior, balanced datasets, all features

2012.10203

box_NFS_BAL_ACC.jpg

0.460784

32ba70686d034d13a0885ec1b180b348

Selected classifier AUC w/ strategic behavior, balanced datasets, all features

2012.10203

box_NFS_BAL_AUC.jpg

0.498131

6a6cb1f01be54b1cb398d0ff1142d3f5

Selected classifier accuracy w/ strategic behavior, unbalanced datasets, all features

2012.10203

box_NFS_UNBAL_ACC.jpg

0.437193

09a071be953447679264e44139a749ee

Selected classifier AUC w/ strategic behavior, unbalanced datasets, all features

2012.10203

box_NFS_UNBAL_AUC.jpg

0.450077

6a11456a1bec450887c69403222e226a

Selected classifier accuracy w/ strategic behavior, balanced datasets, pre-selected 4 features

2012.10203

box_preFS_BAL_ACC.jpg

0.419558

43fbde08c36a445ca2707a410416ba27

Selected classifier AUC w/ strategic behavior, balanced datasets, pre-selected 4 features

2012.10203

box_preFS_BAL_AUC.jpg

0.383666

60c819c6389e46bf8e62b9dfc1f1dc7a

Selected classifier accuracy w/ strategic behavior, unbalanced datasets, pre-selected 4 features

2012.10203

box_preFS_UNBAL_ACC.jpg

0.412085

1f59b1b1291e4c39b2c6792ceaaa805a

Selected classifier AUC w/ strategic behavior, unbalanced datasets, pre-selected 4 features

2012.10203

box_preFS_UNBAL_AUC.jpg

0.39767

2522cf53b06d419d9e75b229e23f1f22

Box plot of classifier accuracy with 4 features: under strategic reporting

2012.10203

main_FS_vert.jpg

0.38999

a92e2379b98b40bba8bb828f317a73d3

Box plot of classifier accuracy with all features: under strategic reporting

2012.10203

main_NFS_vert.jpg

0.49387

cb446a8f0c424218b8037154dcec98b7

These are possible plots of the quadratic function h(w)(Eq. (<ref>)) with u<1(left) and u>1(right). The red-color region represents for the interval of w giving h(w)<0. Alternatively speaking, the Heaviside function does not vanish in these w intervals marked by red. In the case of u<1, it is possible that h(w) has no roots, which means h(w) is always positive. However, this case is not what we are interested in. The other case of u>1 always has real roots with opposite signs.

2012.10204

analysis_of_delta_function.jpg

0.443767

4ab6546f1c6143f79a6d55b2f6e605cd

(color online): Schematic representation of the system under study. An atom at a distance z over a metallic plate moves at constant speed v parallel to the surface. The quantum frictional force opposes the motion of the atom. Inset: the internal state of the atom is a two-level system for which the excited state is threefold degenerate, in analogy with the 1s and 2p levels of the hydrogen atom.

2012.10204

figure1.jpg

0.495548

505f5aa9177344af9cf85e858c4d8d48

(Color online): Frictional force as a function of the relative velocity between the atom and the plate. The force is normalized by the static Casimir-Polder force between a perfectly conducting plate and an atom, and the velocity is normalized as u = v / β, where β is the sound speed in the material. We have considered parameters such that ω = ω_p / ω_b = 1, 5 and z̃ = z ω_p / β = 10, 100. We have taken the sound speed as β = 10^6 m/s.

2012.10204

plot.jpg

0.51978

5f72a39ea6954040852b3438478a487e

Influence of the learned decision boundary.

2012.10209

Aba_2.jpg

0.419799

3ad7e0a8a21f426fb147040645963791

The boundary learning process.

2012.10209

curve.jpg

0.456331

80df1a74d2ca454d859229f1a37c72ab

An example of open intent classification. Book flight and Restaurant reservation are two known intents. We should identify them correctly while detecting the sentences with the open intent.

2012.10209

example.jpg

0.42673

b87fad3ff7374681ba431c54af96f830

The model architecture of our approach. Firstly, we use BERT to extract intent features and pre-train the model with labeled samples. Then, we initialize the centroids {c_i}_i=1^K and the radius of decision boundaries {Δ_i}_i=1^K for each known class. Next, we propose the boundary loss to learn tight decision boundaries adaptive to the known intent feature space. Finally, we perform open classification with the learned decision boundaries to both identify known classes and detect the open class.

2012.10209

model.jpg

0.400759

49a2ab313ab6459c9476f2c59e7bd1b7

Error probability for (a) internet_source and (b) dangerous.

2012.10211

error_prob.jpg

0.453349

cfbf84bd71d64bac897ceb23790a3647

Histogram of Bernoulli misclassification test statistic for (a) internet_sourced and (b) dangerous. The horizontal line marks a value of λ=1.

2012.10211

ll_bern.jpg

0.431621

7a3b859fe69e46d483cb5cec4eb8a473

The relation matrix for (a) internet_sourced, (b) dangerous. Rows correspond to the error messages listed in Table <ref>. Columns correspond to files. In each matrix entry, white indicates no error, black indicates error.

2012.10211

matrix.jpg

0.501633

a0862a10337544e0a8e9380895014435

Correlation matrix between parser behaviors. Rows and columns are sorted according to median message correlation.

2012.10211

parser_correlation.jpg

0.461851

ca2264d0749f44928873ccd112f251e5

Parsers placed in three dimensions according to principal components analysis of the union of internet_sourced and dangerous.

2012.10211

parsers_pca.jpg

0.434125

d29dc51bf42e48faad2e422d995310a8

Principal components plots for (a) internet_sourced and (b) dangerous. Black indicate “valid”, and gray indicates “reject”. The axes correspond to the three principal vectors, and so are not plotted on the same scale.

2012.10211

pca.jpg

0.396956

3bd14b55e5234d05b1ad16172e5f69df

Receiver operating curves for Bernoulli likelihood ratio misclassification statistic for (a) internet_sourced and (b) dangerous.

2012.10211

roc_bern.jpg

0.48948

0f41f8404cad4563ab184ddd55bbfdb3

Scree plots for (a) internet_sourced and (b) dangerous.

2012.10211

scree.jpg

0.469706

64fbcb69f1df41a096af43efc83742dd

Illustration of our method. (1) Traders predict the return of assets using simple formulas. Companies manage Traders and combine Traders' predictions into one value. The Company algorithm consists of four functions: (2) prediction by aggregation, (3) education of bad Traders, (4) dismissal of bad Traders, and (5) recruitment of new Traders.

2012.10215

fig_paper_idea.jpg

0.388369

b28a5bd4bf954955aa9b9316450ef652

Cumulative returns on UK market

2012.10215

fig_tc_only.jpg

0.377582

8e9218f0f107417d9ad00632a3f796a8

Cumulative returns achieved by individual Traders extracted from Companies. “TC” is the overall performance of our method. “TC # term = M” stands for a single best performing Trader extracted from a Company with fixed parameter M.

2012.10215

fig_trader.jpg

0.409988

00adf096b86e41b8a63b122816dffcf6

Cumulative returns on US market in the online prediction setting

2012.10215

fig_us_daily.jpg

0.451308

ff1e324ad17c477086dbdfb7df9ad156

Cumulative returns on US market

2012.10215

fig_us_offline.jpg

0.482226

f2498f80c7ba48968288c99223b7447e

Error on subsets of varying sizes (): x axis denotes points in ascending order of scores. y axis denotes error accrued on the subset containing points up to x.

2012.10216

final_pf_vs_fp_edited.jpg

0.502069

8898b18c2c9a4d62ad0f5fbe2a10f8a3

Given two groups Blue and Yellow (with true labels as shown), we have to choose just between the two classifiers x=0 and y=0. The Blue group is inherently harder to classify. Equalized odds makes us choose the classifier x=0, thereby hurting the Yellow group. We could choose y=0 with no aggregate effect on Blue, doing much better on Yellow.

2012.10216

harm_example.jpg

0.449404

c21be706cace45109f7b420e590d42db

Illustration of the real-time video streaming process in a cell with n users.

2012.10219

System_model_final_update_2.jpg

0.456595

405441623ead466d86960d1a386a638f

The impact of number of users on playout rate.

2012.10219

bar_playout_2.jpg

0.48451

0276b941e9f84944af1570522c5745ee

Drop rate as a function of buffer size.

2012.10219

buffer_seconds.jpg

0.510137

40cde20c8eec4779813fca3bee9ca7ca

QoE for our approach and ABR.

2012.10219

qoe_1.jpg

0.47262

366150e1048944e6a3ddb452cad3b5e9

Number of users with maximum playout rate.

2012.10219

tech_report.jpg

0.466404

81dd2df0127545d28602807bcfdcd6b4

The playout rate for two classes of users.

2012.10219

two_classes_delta_2.jpg