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
|
|---|---|---|---|---|
0.423739 | 130285259c62402c891670302540c7c8 | Charge drift as a function of the depth at which the charge was generated. The plot is for modules in L2 at the nominal CMS B-field of 3.8 T. The slope of the linear fit is interpreted as the tangent of the Lorentz angle. | 2012.14304 | hMean_LYR-2.jpg |
0.446931 | 89821aa0debc485e8fd57c797d05222a | Module temperatures, measured with the thermal mockup and averaged over all modules on a ladder, as a function of the module position along the cooling loop. The module position is defined by the ladder number, which is counted starting from top to bottom. The two plots show the temperature measurements for the loop entering the detector from the +z (left) and -z (right) side. The loops entering the detector from opposite sides (± z) differ in their geometry <cit.>. The direction of flow of the coolant is left to right for the +z side and right to left for the -z side. The measurements are performed for different mass flows and a nominal power load, expected for detector operation after having collected data corresponding to an integrated luminosity of 300 . The temperature measurements for ladder number 12 are affected by a bad thermal contact between the temperature sensor and the cooling loop. | 2012.14304 | mockuptemp.jpg |
0.494989 | 72dd43cf264c4d33a438db4f7b1bbb04 | Picture of a L2 module after assembly of the protection cap. The amber-colored protection cap covers the wire bonds of ROCs and TBMs. In the picture, the flat polyimide flex cable used for module testing is connected to the module. | 2012.14304 | module-with-cap.jpg |
0.422683 | 2ece83cd3deb4f1790594e8de1a4fa12 | Cross section of a pixel detector module for L2-4 cut along the short side of the module. | 2012.14304 | modulexsec.jpg |
0.413873 | eada2cf896b24cf9abbc50e7ff8cd7c2 | Overview of the readout and control system of the . The components in the control and readout chain that have been specifically designed for the are highlighted in yellow. All other electronics components on the service half-cylinders are identical to those used in the original pixel detector. | 2012.14304 | pixeldaq.jpg |
0.407643 | 8f3855f0e92747659b8245b1c4c61f02 | Measured and simulated efficiencies for (used in the original pixel detector) and (used in the in the outer layers and the detector) as a function of X-ray hit rates <cit.>. | 2012.14304 | psi46digeff.jpg |
0.455516 | c9788abb267042e88c3d7dcb5bde2107 | One half of the detector with the service half-cylinders connected. The detector system is placed on a rail system inside a 5 m long custom-built transport box. | 2012.14304 | psilab.jpg |
0.386782 | 6375149928d44167960c14da0fac7619 | Probability of occurrences of similarities among same (auto) and different identities (cross) | 2012.14305 | Histogram.jpg |
0.453661 | e2cb61e310c842948bb803cb9fcd0abe | Model performance at different threshold (adaptive) | 2012.14305 | adaptive-metrics-comparision.jpg |
0.519992 | bbfb00eef2434d458cb4044c52b7731c | Flow-diagram of online optimization based adaptive threshold search method | 2012.14305 | adaptive_threshold_flowdiagram.jpg |
0.442162 | b0b61b65524f4a9a91845708d7784066 | Auto similarities pairing | 2012.14305 | auto-similarity-diagram.jpg |
0.392028 | f5821fa72f22432b99cdbb4dff21b279 | Auto-similarity distribution among faces of same identities | 2012.14305 | auto-similarity_distribution.jpg |
0.443281 | e4b0f332d8ef42d69307d66a6817a08d | F1-score curve at fixed threshold @{0.3, 0.5, 0.7} and adaptive threshold | 2012.14305 | comparative-study-f1score.jpg |
0.43287 | 69880f61ef6e47d3aab856dd3317c199 | Precision curve at fixed threshold @{0.3, 0.5, 0.7} and adaptive threshold | 2012.14305 | comparative-study-precision.jpg |
0.495288 | ee5451ab72e5496cb5559197cf03094f | Recall curve at fixed threshold @{0.3, 0.5, 0.7} and adaptive threshold | 2012.14305 | comparative-study-recall.jpg |
0.388645 | 9537a39d1c7a441fbf7b950825055217 | Comparative study of fixed threshold @{0.3, 0.5, 0.7} with adaptive threshold | 2012.14305 | comparative-study-threshold-vs-f1score.jpg |
0.396316 | a1dac55ba69d4dd6b26c7b32bfded3d6 | Cross similarities pairing | 2012.14305 | cross-similarity-diagram.jpg |
0.415915 | 789918f5291c45cc82b488d39cff8e41 | Cross-similarity distribution among faces of different identities | 2012.14305 | cross-similarity_distribution.jpg |
0.3971 | ebee849f21dc4c958d37e8f1797bbbe0 | Estimated actual and Gaussian distribution functions for standardized auto and cross similarities functions | 2012.14305 | histogram-gaussian-curve.jpg |
0.444114 | e46dc759bf9340ebb23f39542d9dc16f | Flow-diagram of identity recognition/re-identification | 2012.14305 | model_inference_flowchart.jpg |
0.426041 | 640d240ce8d54f35a60ab1f5cdf7ffc9 | ROC curve at fixed threshold @{0.3, 0.5, 0.7} and adaptive threshold | 2012.14305 | roc_curve.jpg |
0.418112 | fbd6c09b2a644032a632a3534ba08e12 | The A-GL^0.15/(MV_1) plot for corpora segmented by different algorithms<cit.>. We use CKIP<cit.>, the best Chinese algorithm, as the standard in cases (a, b, c, d), while (e, f) employ the word bank<cit.> segmented by linguists in Academia Sinica. Initially we were alarmed by the violation of positive correlation in (e, f). But soon it dawned on us that the version of standard bank we used is older than the one by CKIP. This realization reveals that the A-GL^0.15/(MV_1) plot has the potential to function as an empirical index for the accuracy of algorithm. | 2012.14309 | Accuracy.jpg |
0.408944 | b971575cee024ef1938dadaf022d416f | The dotted/crossed (red/blue) curves in the log-log plots of (a, b) show FRD of word/syllable (character) in Chinese novels, Demi-Gods and Semi-Devils and Frog. Similarly, (c, d) are for word/syllable in English novels, Moby-Dick and The Hobbit. The fitting function is for the red points. | 2012.14309 | FRD.jpg |
0.493021 | 290e473fcc7249a09b38acb9737a619a | Left figures show the weighted average scaling relation of r_H≡ H_n+1/H_n, while right ones depict r_f with standard error, σ_H and σ_f. Panels (a, b) are for Moby-Dick, (c, d) for Frog, (e, f) for Angels-Demons, (g, h) for The scholar. Note that f_2/f_1 is excluded when calculating r_f. Detail definitions of r_f and SP, a measure of goodness of scaling, are given in METHODS. | 2012.14309 | Geo-seq.jpg |
0.429246 | 8fef9d9b280a429ba40c88faac3c654d | Left figures depict Col-rank plot, while right ones are for Link-rank. Panels (a, b) are for Harry Potter 3, (c, d) work by Lung Yingtai, (e, f) David-Copperfield, and (g, h) work by Yu Guangzhong. Dashed lines are fitting curves. | 2012.14309 | LC.jpg |
0.413468 | 995b782ca4634eb7a121a67c80253dbe | Panels (a, b) are the RRD plot for Frog and Moby-Dick. Their construction is demonstrated schematically in panels (c, d). | 2012.14309 | RRD.jpg |
0.446254 | da21e4ec716f4677b97fc512067086e8 | Dataframe of RRD, word and syllable in Moby-Dick, where SeqOrder is sequence order that follows Rule 1, and y_n is the rank of n-th syllable. | 2012.14309 | dataframe.jpg |
0.418841 | 81127d1e025d494fbf8c18756b7f0b9c | Yellow (grey) points in (a, b) sketch out the scaling lines and are selected by Step 1. (b) The fitting form of scaling lines follows f(x)=A× r_f^N(x) where N(x)=ax^-b. Step 3 chose f_2 as the best base, while r_f=0.870 that also appeared in Fig. <ref>(b) is obtained in Step 2. The fitting score is defined in Eq. (<ref>). | 2012.14309 | fitting.jpg |
0.548107 | 6485fc85e0aa486192b1f62462918377 | Flow chart of our model. When we refresh the words, each content will have probability P to be refreshed. | 2012.14309 | flow_chart.jpg |
0.381478 | c5a7d7b7218448adbba14f8df7adfa42 | The connection between three contents on the first syllgram and second syllgram. Content at the middle and right may have strong connection on the first syllgram, but have poor connection on the second one. Left and middle content may have strong connection on the second syllgram but poor connection on the first one. | 2012.14309 | logic_connect.jpg |
0.443091 | 1e22bca315da42e7b3d5b78bbf510e2c | Cases with SP<0.75 due to different reasons: (a, b) L is too small, (c, d, e) FRD is neither Zipf nor Zipf-like, (f) V_1 is too small. Their parameters can be found in Tab. <ref>. Note that the computer-generated paper in (e) is created by SCIgen<cit.> which has succeeded at deceiving the editors of many journals and conference organizers. But our RRD analysis can tell it from real paper. | 2012.14309 | messy.jpg |
0.398015 | 50999071eef9420097be3c1c92672bf9 | Schematic demonstration of segmenting Toccata and Fugue in D minor, BWV 565, written by J. S. Bach. With the aid of waveform, we can display music in 1-D data type and segment it into possible units that have scaling structure. To deal with such 1-D type data without precise definition of main-unit, we need to segment their main-unit before segment their sub-unit. We guess the red grids are main-unit while the green grids are sub-unit. This method provides us with a tool to analyze the statistical property of music. The pictures of violin, cello and double bass are accessed from wikipedia. The waveform is displayed by Sonic Visualiser. | 2012.14309 | music.jpg |
0.452148 | c95a34be714f4fcaa8edd1bffab5dca5 | The result of our model with N=10000, λ=0,495, P=0.0005 and z=1000. The probability of creating a 1-syllgram, 2-syllgram, 3-syllgram, 4-syllgram and 5-syllgram word are 0.15, 0.40, 0.25, 0.15 and 0.05 respectively. | 2012.14309 | sim_result.jpg |
0.468967 | e353b7d1ab224b9497dfed1bb144e57a | The topological characteristics of multilayer network in Harry Potter 1. Panels (a, b) show the degree distribution of (word, syllgram) where vertices with zero degree have been excluded because they present as a singularity. Note that (b) exhibits the feature of a scale-free network<cit.> P(k)∼ k^-γ. The distributions of local clustering coefficient of (word, syllgram) are plotted in panels (c, d). The high average in (c) exemplifies a small-world network<cit.>. As in (e, f), their average path length (L_w, L_s) = (3.04, 3.35). | 2012.14309 | topology.jpg |
0.590966 | cc8d3da0a85f4ce8941f40405dabef0b | Numerical results on the two-qubit isotropic states. The blue line represents the precisely calculated logarithmic negativity of isotropic states with parameter p∈[0,1]. The yellow markers depict the estimated logarithmic negativity by simulations of Algorithm <ref> on selected isotropic states. | 2012.14311 | 2qubit_iso_logneg.jpg |
0.448171 | e12f270aaac8420f8843e13cd6ab5724 | Numerical results on the four-qubit Breuer states defined in (<ref>). Each line depicts the smallest eigenvalue of every Breuer state with parameter p∈[0,1] under the corresponding map. This line of the smallest eigenvalues is a lower bound of the loss function L(). Each marker depicts the minimized loss value obtained by simulations (sim) of Algorithm <ref> on a chosen Breuer state, aligning with the theoretical line. | 2012.14311 | 4qubit_breuer.jpg |
0.473154 | cabddab3d17a4077b6b4e6ec9415c023 | Numerical results on the four-qubit isotropic states defined in (<ref>). Each line depicts the smallest eigenvalue of every isotropic state with parameter p∈[0,1] under the corresponding map. This line of the smallest eigenvalues is a lower bound of the loss function L(). Each marker depicts the minimized loss value obtained by simulations (sim) of Algorithm <ref> on a chosen isotropic state, aligning with the theoretical line. | 2012.14311 | 4qubit_iso.jpg |
0.447575 | 20245114e21a41418cd744f7f46723f2 | The simplified quantum circuit that estimates the overlap ⟨ψ|(ρ_AB) |ψ⟩ in (<ref>) for a given implementable operation , where |ψ⟩:=U_|0⟩^ 2n is the parameterized input state. | 2012.14311 | overlap-simple.jpg |
0.4007 | 00f9b27628cf4a05a958fc3df9de9c6b | Proportion of papers per year per topic. | 2012.14312 | SM_topics25_timeline.jpg |
0.425646 | 02ae202c2a144cf09e48d3afb3ec76fe | Panarchy is an heuristic of nested adaptive cycles that serves to represent a variety of systems and environmental problems. Adaptive cycles (A) at different scales of the hierarchy (B) can be connected through remember and revolt cross-scale interactions. | 2012.14312 | fig1_panarchy.jpg |
0.517733 | c1c78c5ef0304a97a542b9874f7dbbb7 | Algorithm and number of topic selection Gibbs sampling maximizes entropy and log-likelihood estimation, making it a suitable algorithm for our data (A). Increasing the number of topics (from 5-100) shows that α decreases, suggesting that despite the larger number of topics, a few of them suffice to describe most papers (B). Log-likelihood is maximized for 50 topics followed closely with 25. While 50 topics is marginally better, for visualization purposes we restrict our analysis to 25. Note that perplexity cannot be calculated for Gibbs sampling, hence the missing value in A and absence in B. | 2012.14312 | fig2_topicnumber.jpg |
0.427645 | 079917f3257a47b6a7e27df4d76f84fe | Panarchy topics over time The number of papers per year (A) with a maximum of 324 in 2019. The proportion of papers per year (B) and the relative proportion of topic content per year (C) do not show strong trends for the time window with most papers (gray area in A). Each topic is summarized in (D) by the top 10-words that best describe them according to the posterior probability of our model fit. See FigS1 for a panel of time series per topics | 2012.14312 | fig3_topics25.jpg |
0.443718 | 62df55185c544a59b43c37d9cf62096a | Qualitative results Document analysis was used to disentangle the different panarchy features addressed by a sub-sample of papers (N = 42). Over half of the papers are conceptual work, and most empirical papers fall into qualitative methods. | 2012.14312 | fig4_qual_results.jpg |
0.486868 | c4915a1609ab4bdc8e1d7468be4e8ca5 | Architecture of the sensor network and heteroscedastic observation noise model for planar pushing. We use 6-channel RGBXYZ images as input for computing the object position and contact related state components. The object orientation is estimated relative to the initial orientation by comparing the RGB glimpse centered on the current estimated object position to the initial one. White boxes represent tensors, green arrows and boxes indicate network layers, whereas black arrows represent dataflow without processing. For convolution (conv) and deconvolution (deconv) layers, the numbers in each tensor are the kernel size and number of output channels of the layer that produced it. For fully connected layers (fc), the number corresponds to the number of output channels. With the exception of the output layers, all convolution, deconvolution and fully connected layers are followed by ReLU non-linearities. The (de)convolution layers also use layer normalization. | 2012.14313 | sensor_network.jpg |
0.451976 | f7e2f69bd94744ec95e70687be007a7e | Tracking result with MOT17-DPM training data compared with baseline approaches. 1) Baseline: IOU-based; 2) IOU and motion based association; 3) Combine the spatial feature, deep features and IOU; 4) Explicit features: the explicit weights of each feature are learned from network; 5) Implicit features: the cost function is obtained from the end-to-end learning GRU network. | 2012.14314 | Experiment1.jpg |
0.403497 | 2ad59419f1fd461bbcfa83f735a9d3de | The proposed MOT utilizes motion model, appearance model, as well as detection bounding box information. | 2012.14314 | MOTGRU2.jpg |
0.51494 | 2f09fe877ee6476485f545cdf5a4240f | The flow chart of the proposed MOT framework. Our online MOT algorithm consists of three parts as follow: 1) Gaussian statistics (mean and covariance) are given by auto-tuning Kalman prediction; 2) data association is realized by using Hungarian algorithm; and 3) The similarity between detections and predictions is evaluated through the recurrent neural network with GRU, which is trained with the ground-truth of the MOT17 data <cit.>. | 2012.14314 | Visio-framework1.jpg |
0.43952 | 8bc2e39540b2400cab6a4a9b6ef5a7b6 | Examples of multi-object tracking results. The curved line behind each person is the trajectory. | 2012.14314 | figure.jpg |
0.484093 | 6b1cfad740274cc7a5d478d80fddd580 | Flow chart of the Cost Matrix calculation with GRU. 1) The feature vector comes from the motion model and feature extractor. 2) Feature vector of tracked object i in frame t and the feature of detection j in frame t+1 are used as input of the GRU model to calculate the association cost C_ij of the cost matrix. | 2012.14314 | framework_gru.jpg |
0.450726 | e900d37dd49c470e91648665530d67a6 | Normalized growth and migration tracks from the scenario where the giant planet starts forming at 5 au. The growth tracks of both mass and radius are normalized to the final mass (1 M_J) and radius (1.6 R_J) of the giant planet. The migration track is instead normalized to the initial semimajor axis (5 au) of the giant planet. | 2012.14315 | Figure_1.jpg |
0.412381 | d8195a60be28468b8c30278aa2a1849a | Mass growth and planetesimal accretion tracks, normalized to their final values, from the scenario where the giant planet starts forming at 19 au. | 2012.14315 | Figure_10.jpg |
0.458569 | efbc6fddf502424db4e3d88aeb6f8201 | Mass fraction of solid material as a function of the distance from the star in our template protoplanetary disk. The mass fraction is expressed with respect to the total mass of gas assuming a solar composition for the latter. The mass fraction of condensed material is always lower than the protostellar Z as some elements (Ne) and molecules (CO and N_2) never condense in our disk (see main text and Table <ref> for further details). | 2012.14315 | Figure_2.jpg |
0.422363 | e942c195c7c54b0aafef598ebf59a3f3 | Fractions of C, O, N and S in solid phase (solid lines) and gas phase (dot dashed lines) in the protoplanetary disk at varying distances from the host star. The values are normalized to the protostellar abundances reported in Tables <ref> and <ref>. The vertical dashed lines indicate the position of the snowlines of H_2O (2.5 au), refractory organic carbon (5 au), NH_3 (9.4 au), CO_2 (10.5 au) and CH_4 (105 au). | 2012.14315 | Figure_3.jpg |
0.437971 | ba22494c970e4f4f86e134de305f0187 | Top: Final C/O ratios of the gaseous envelope of the giant planet as a function of its initial orbital position for the cases of solid-enriched giant planets (purple line and squares) and gas-dominated giant planets (green line and circles). Bottom: C/O ratios of the gas (green dot-dashed line) and the solids (purple solid line) throughout the protoplanetary disk. The dashed lines shows the stellar C/O ratio, which is the same as the solar one. We refer the readers to Sect. <ref> for details. | 2012.14315 | Figure_4.jpg |
0.46214 | 78cf3c64b8c040f38cebf5ed298f5a78 | C/N (left) and N/O (right) ratios of the gaseous envelope of the giant planet as a function of its initial orbital position. In the figure we show the cases of solid-enriched giant planets (purple line and squares) and gas-dominated giant planets (green line and circles). The dashed lines show the stellar C/N and N/O ratios, which are the same as the solar ones. We refer the readers to Sect. <ref> for details. | 2012.14315 | Figure_5.jpg |
0.485238 | c8e50b67f9d546d8b26cbdb45c396af2 | S/N ratio (left), metallicity and mass of accreted planetesimals (right) of solid-enriched giant planets as a function of their initial orbital positions. The dashed lines show the stellar S/N ratio and the mass of high-Z elements equivalent to the stellar metallicity, which are the same as the solar ones. We refer the readers to Sect. <ref> for details. | 2012.14315 | Figure_6.jpg |
0.517137 | 36da35bd93ee470d93400b6cca938d94 | Left: normalized elemental ratios of the gaseous envelope when the metallicity of the giant planet is dominated by the accretion of planetesimals. Right: normalized elemental ratios in the gaseous envelope when the metallicity of the giant planet is dominated instead by the accretion of gas. Each elemental ratio is normalized to the relevant stellar elemental ratio. | 2012.14315 | Figure_7.jpg |
0.518208 | 56f44a1717034bada38b6ceb333cc799 | Same as Fig. <ref> but assuming a 9:2 N_2:NH_3 ratio instead of the 1:1 ratio from <cit.> adopted in our standard compositional model. The contrast among the different elemental ratios in the solid-enriched metallicity case is enhanced by a factor of two, while the effects of this alternative partitioning are more limited in the gas-dominated metallicity case. | 2012.14315 | Figure_8.jpg |
0.486614 | 20efd730e8c8422e88a3ddf50aca938e | Elemental ratios of the gas (dot-dashed lines) and the solids (solid lines) in the protoplanetary disk as a function of the radial distance from the star. Each elemental ratio is normalized to the relevant stellar elemental ratio. | 2012.14315 | Figure_9.jpg |
0.469695 | 54ab9e3e8a16417d9bff15714f3d328e | As Fig. <ref> for the mass balance of water in the framework of our 5-layer model. Compared to Fig. <ref>, the main differences arise from the fact that water precipitates out of layers 2 and 1 so that its abundance in these layers is, on average, extremely small. Also, the evaporation of mushballs removes a fraction a_ w of the water which is incorporated into layer 4, with the remaining 1-a_ w being incorporated into layer 5. | 2012.14316 | fig2_columns_h2o.jpg |
0.441466 | 9b4564749a9c4c57bb43d1e2d72f4d41 | Mass balance of ammonia in the framework of our 5-layer model. We consider that three main processes transport material between layers: In yellow, small-scale convection is modeled as an updraft and its reciprocal downdraft between adjacent layers. We consider that it is characterized by an upward mass flux ṁ_ deep between layers 5 and 4 and layers 4 and 3, and by an upward mass flux ṁ_ conv between layers 3 and 2 and layers 2 and 1. In blue, strong storms due to water condensation lead to a transport of material directly from layer 3 to layer 1 and to a compensating subsidence mass flux from layer 1 to layer 2 and to layer 3. These storms also lead to the formation of mushballs and evaporative downdrafts which deliver ammonia and water directly to layers 4 and 5. The terms in each layer correspond to the mass balance of ammonia described by Eq. (<ref>). | 2012.14316 | fig2_columns_nh3.jpg |
0.430429 | 0d0bc8c00f194d418f068a73c4154942 | As Fig. <ref> for the balance of static energy in the framework of our 5-layer model. In addition to the terms due to a transport of static energy, terms resulting from the condensation or evaporation of water are highlighted. A flux F_ tot arising from internal heat is added to layer 5 and removed from layer 1. | 2012.14316 | fig2_columns_s.jpg |
0.491304 | 8f7236d66a6b4be48d84ef6be3fc6878 | Average map of ammonia abundance in Jupiter retrieved by the Juno MWR during PJ1 to PJ9 as a function of latitude and pressure. Overlaid are indications of altitude and temperature as well as the layers and mechanisms (small-scale convection and/or storms in the water condensation region, dry convection deeper) considered in this work (see text). Water vapor condenses to ice particles at ∼ 5 bar level (0^∘C), ∼ 50 km below the 1 bar level. | 2012.14316 | fig2_model5layers.jpg |
0.403996 | d3fab8f47f6942fcbaa44693f9e88d18 | Abundances of water (top row), ammonia (middle row) and potential-temperature anomalies (bottom row) obtained with our model, as a function of f_ storm, a parameter assessing the mass flux in large water storms relative to that of dry convection below the water cloud base. The left column corresponds to a situation in which no small-scale convection is present in the water condensation region (f_ conv=0) and pertains to mid-latitude regions of Jupiter. The right column assumes that both small-scale convection and storms occur, so that f_ storm+f_ conv=1. The curves show the different layers considered in Fig. <ref>: 1) upper layer (purple); 2) mushball-seed layer (blue, dashed); 3) water cloud layer (light blue); 4) downdraft layer (orange, dashed); 5) deep (red). The potential-temperature anomalies are calculated assuming that intrinsic heat flux transport occurs with negligible superadiabaticity (see text). | 2012.14316 | fig2_mush_model.jpg |
0.472436 | f21c8cfcf30e425bab8e06ea99a25b56 | Distribution of ammonia concentration obtained with our conceptualmass-balance model. The top panel shows the NH_3 mixing ratio as a function of latitude and pressure in the 5 layers of our model. The bottom panel indicates the value of f_ storm (black line) obtained to reproduce the 1-3 bar MWR ammonia mixing ratio compared to the number of flashes per second detected by the MWR instrument between PJ1 and PJ16 <cit.>. The large and uniform ammonia concentration in the Equatorial Zone is well reproduced by assuming a scarcity of storms (f_ storm∼ 0), in line with the absence of lightning there. At mid-latitudes, frequent storms and subsequent mushball formation lead to a depletion of ammonia. | 2012.14316 | fig2_reproduce_MWR_map.jpg |
0.456624 | f98d8490766e4e29986bf28c2d165bac | t-SNE projection of patterns. | 2012.14320 | figure6.jpg |
0.514385 | f65d2e6eec9a46e5ac34c6fbfdf415cc | An illustration of n-gram pre-training. | 2012.14320 | model.jpg |
0.452745 | 98f2d0cdee534ddaa0b5c08f62412bae | An example from the Chinese Wikipedia corpus. n-grams of different lengths are marked with dashed boxes in different colors in the upper part of the figure. During training, we randomly mask n-grams and only the longest n-gram is masked if there are multiple matches, as shown in the lower part of the figure. | 2012.14320 | tmp.jpg |
0.447086 | fbc3cc3387784dddabe0f9fdf7eb4717 | Examples of Question-Answer pairs from our insurance FAQ dataset. The correct match to the query is highlighted. | 2012.14320 | tmp2.jpg |
0.52684 | 8cfc9c849fc44222871448b599d765da | Active instabilities in two-dimensional layered states. (a) Instability arising from the chiral active force (ζ_c, here positive). The layers are indicated by dark red lines, while the linear flow field at instability overlaid (blue arrows). The thick black lines indicate confining walls, which create a finite threshold for the instability. (b) Instability arising from the achiral active force (ζ) for comparison; same stylings. | 2012.14321 | 2d_instability_flows.jpg |
0.440985 | da975c7ca8494045af13396f584bd574 | The flow field due to the chiral, active force for a distortion of a two-dimensional layered state purely along the normal direction. | 2012.14321 | 2dflows.jpg |
0.468998 | 50ffd774bbab46ccaaed05775a2c3277 | The active chiral flow field due to one dimensional undulations in three dimensional layered states | 2012.14321 | Flows2.jpg |
0.473708 | 6e4cf24f5b8049f190764510176f5fdf | The spontaneous vortex-lattice state: the chiral active force density (governed by z_c) generates counterrotating circulatory flows, with vorticity Ω_z obeying η∇^2Ω_z=-z_c∇^2∇_⊥^2 u, around the undulations u of the active Helfrich-Hurault instability. | 2012.14321 | activeHH3d.jpg |
0.410861 | 9911952750054704beedec5fdf20f502 | Mean photon number at 1^st and 20^th waveguide with respect to time when two photons are injected to the 1^st waveguide. | 2012.14326 | 2d2ph.jpg |
0.378742 | 10477443608149e7b0157c89a4e1e393 | Mean photon number variation for coherent state when input is given at first waveguide. |α|^2 is set to be 2. | 2012.14326 | coherent.jpg |
0.420059 | da55b550679f43309e8051c32c3911d7 | Mean photon number plot for two photon N00N state for p=1 and q=2. The plots for N_2 and N_19 are similar as that of N_1 and N_20. | 2012.14326 | n2d.jpg |
0.421953 | 5b86c37e218d4762adc32d0d0440b2b8 | Schematic view of J_x photonic lattice with parabolic coupling distribution. | 2012.14326 | waveguidep.jpg |
0.414682 | 64865cb5e74441d1a88d3a11c86ced02 | Graph shows the average number of targets placed at the retrieval points over iterations, for rescuer strategy 1 and scenario 3. x scale limited to 10000 iterations. | 2012.14329 | AvgTimeSeries1.jpg |
0.481282 | fc75ea60f35f47d78222569710bac787 | Efficiency estimate of scenario 3 | 2012.14329 | CostGraph_Scenario_3_Mix.jpg |
0.451505 | e6281609966f437ea8cdf410a32cfb1f | Heterogeneity in Scenario 2 vs. {R:S} ratio. | 2012.14329 | HeterogenityMeasureScenario2.jpg |
0.473516 | 42aac2be60c947c7b95ff448dcbd3f08 | Heterogeneity in Scenario 3 vs. {R:S} ratio. | 2012.14329 | HeterogenityMeasureScenario3.jpg |
0.462198 | 4d918202de5a44d091540e5c0617cbfc | Iteration 4000 | 2012.14329 | SnapShot3.jpg |
0.505438 | db4a5ce154b1438299c97902557edff2 | Iteration 6000 | 2012.14329 | SnapShot4.jpg |
0.422804 | c65ad1ccfdec4fca928b7920e5362e0b | Average time constants for different strategies in scenario 1, for (rescuer, Searcher) combinations. | 2012.14329 | TimeConst_Scenario_1.jpg |
0.418613 | 26253ec59abe4e8aac78bc26a05cf27a | Average time constants for different strategies in scenario 2, for (rescuer, Searcher) combinations. | 2012.14329 | TimeConst_Scenario_2.jpg |
0.431871 | 5af8228457e1477894ee606b4edd3b89 | Average time constants for different strategies in scenario 3, for (Rescuer, Searcher) combinations. | 2012.14329 | TimeConst_Scenario_3.jpg |
0.456067 | 96149ff70b814cd599962de956b3365b | Behavior tree (BT) summarising searcher and generic rescuer models. The distance values for every BT node (d_0,...d_n) are used to calculate the inter-species distance in Eq. <ref> and Heterogeneity in the HMRS. | 2012.14329 | newBT.jpg |
0.422523 | b9260af3d4d04a6f8655817372aeecea | The designed configuration space has four collection points (green), with targets (yellow) randomly placed. Searchers and rescuers are colored magenta and red respectively. A ray projecting from the agents indicates the direction of heading and their color represents the agent's state (Red - Long range target, Green - Communication, White - Collection or Random Walk. | 2012.14329 | newConfig.jpg |
0.454536 | 7ebaad1b1ed840a18d4e55499be4d94d | The ranges of communication,collision detector, short range target detector, long range target detector are labeled as D_C,D_CL,D_TS,D_TL, respectively. | 2012.14329 | sensorDiagram.jpg |
0.412052 | dd8a136214c64301a68a6a574e272413 | Iteration 1 | 2012.14329 | snapShot1.jpg |
0.440051 | ebdff843a9e74971a706f1fc9f2bc178 | Iteration 2000 | 2012.14329 | snapShot2.jpg |
0.394931 | 972b63471803451a89ed3e1e9ed6e865 | ROC curves. a. Yes/no ROC curves for a single shifting criterion (black), vs. a shifting likelihood-ratio (green), between the two 1d normals of fig. <ref>a (adapted from Wickens <cit.>, fig. 9.3), and a shifting likelihood-ratio between a normal and a t distribution in 4d (purple). The optimal two-interval accuracies of the 1d normals (fig. <ref>c) and the 4d distributions (fig. <ref>b) are 0.74 and 0.97, equal to the areas under their likelihood-ratio curves here. The points marked on these curves are the farthest from the diagonal, and correspond to the Bayes discriminability. b. Distributions of the log likelihood ratio of the 4d t vs normal distribution. Sweeping the criterion corresponds to moving along the purple ROC curve of a. | 2012.14331 | ROC.jpg |
0.435196 | fc6d9c1380f54bf6b1993d763b6d0f39 | Binary yes/no and two-interval classification tasks. a. Optimal yes/no decision between two unequal-variance 1d normal distributions. b. The same task transformed to the log likelihood ratio axis (log vertical axis for clarity). c. Optimal two-interval discrimination between the same 1d normal distributions a and b is actually a discrimination between 2d normals ab and ba. d. The task transformed to the log likelihood ratio axis. | 2012.14331 | binary_class.jpg |
0.461298 | 16be97e008c84d5cab33303952478676 | Comparing discriminability indices. a. Plots of existing indices d'_a and d'_e as fractions of the Bayes index d'_b, with increasing separation between two 1d and two 2d normals, for different ratios s of their sd's. b. Left : two normals with 1 sd error ellipses corresponding to their sd matrices 𝐒_a and 𝐒_b, and their average and rms sd matrices. Right: the space has been linearly transformed, so that a is now standard normal, and b is aligned with the coordinate axes. c. Discriminating two highly-separated 1d normals. | 2012.14331 | discriminability.jpg |
0.405418 | 1a331006139248d49b73acd43ec514b6 | Toolbox outputs for some integration and classification problems. a. Top: the probability of a 3d normal (blue shows 1 sd error ellipsoid) in an implicit toroidal domain f_t(x)>0. Black dots are boundary points within 3 sd traced by the ray method, across Matlab's adaptive integration grid over angles. Inset: pdf of f_t(x) and its integrated part (blue overlay). Bottom: integrating a 2d normal (blue error ellipse) in a domain built by the union of two circles. b. Estimates of the 4d standard normal probability in the 4d polyhedral domain f_p(x)= ∑_i=1^4 | x_i |<1 using the ray-trace method with Monte Carlo ray-sampling, across 5 runs, converging with growing sample size of rays. Inset: pdf of f_p(x) and its integrated part. c. Left: heat map of joint pdf of two functions of a 2d normal, to be integrated over the implicit domain f_1-f_2>1 (overlay). Right: corresponding integral of the normal over the domain h(x)=x_1 sin x_2 - x_2 cos x_1 >1 (blue regions), `traced' up to 3 sd (black dots). Inset: pdf of h(x) and its integrated part. d. Classifying two 2d normals using the optimal boundary l (which yields the Bayes-optimal discriminability d'_b), and a custom linear boundary. d'_e and d'_a are approximate discriminability indices. e. Classification based on samples (dots) from non-normal distributions. Filled ellipses are error ellipses of fitted normals. γ is an optimized boundary between the samples. The three error rates are: of the normals with l, of the samples with l, and of the samples with γ. f. Classifying several 2d normals with arbitrary means and covariances. g. Top: 1d projection of a 4d normal integral over a quadratic domain q(x)>0. Bottom: Projection of the classification of two 4d normals based on samples, with unequal priors, and unequal outcome values (correctly classifying the blue class is valued 4 times the red, hence the optimal criterion is shifted), onto the axis of the Bayes decision variable β. Histograms and smooth curves are the projections of the samples and the fitted normals. The sample-optimized boundary γ=0 cannot be uniquely projected to this β axis. h. Classification based on four 4d non-normal samples, with different priors and outcome values, projected on the axis along (1,1,1,1). The boundaries cannot be projected to this axis. | 2012.14331 | examples.jpg |
0.413997 | f3f7ebe3065c41a7bbf87286767e1d1f | Applying the method and toolbox to visual target detection studies. a. Example image of a target on a natural background. b. Classification of images with the target present or absent, in the space of three cues. p(e) and d'_b correspond to the classification error of the sample points using γ. c. Example image for camouflage detection. d. Classifying these using 6 cues, viewed in terms of the log likelihood ratio l. e. Bootstrap mean ± sd of hit and false alarm fractions from applying a family of boundaries (corresponding to varying the criterion likelihood ratio) on 100 samples of the true 6d cue distributions (color fills), vs. their normal approximations (outlines). f. Classifying with only two cues computed at 2px. Grey curves are contours of the log likelihood ratio l. g. Combining the two cues of plot f into one using l (i.e. the space of plot f has been projected along the gray contours). h. Classifying with such combined cues at 3 scales. | 2012.14331 | examples_vision.jpg |
0.564224 | e79df1a0fa3d4cc395b9198894e19cc9 | Method schematic. a. Standard normal error ellipse is blue. Arrow indicates a ray from it at angle n in an angular slice dn, crossing the gray integration domain f̃(z)>0 at z_1 and z_2. b. 1d slice of this picture along the ray. The standard normal density along a ray is blue. f̃_n(z) is the slice of the domain function f̃(z) along the ray, crossing 0 at z_1 and z_2. | 2012.14331 | method_schematic.jpg |
0.483028 | b0df5a9d1a7544e5b0e554007ec7e49a | Testing normal approximations for classification. a. Classifying two empirical distributions (a is not normal). Gray curves are contours of l, i.e. the family of boundaries corresponding to varying likelihood ratios of the two fitted normals. b. Mean ± sd of hit and false alarm fractions observed (color fills) vs. predicted by the normal model (outlines), along this family of boundaries. Vertical line is the optimal boundary. c. Similar bands for class b hits and overall error, for the 4d 4-class problem of fig. <ref>h, across boundaries assuming different priors p_b, and d. across boundaries assuming different covariance scales (d's). | 2012.14331 | normality_check.jpg |
0.456726 | 1e72d496ac894f1082a4da73b0bf4f2e | Performance benchmarks of the generalized chi-square (denoted by χ̃^2) and ray-trace methods, against a standard Monte-Carlo method. a. Relative inaccuracies in d' estimates by the Monte-Carlo method (across multiple runs), and our two methods, as the true d' increases. Monte-Carlo estimates that take similar time rapidly become erroneous, failing beyond d'≈ 10. Our methods stay extremely accurate (around machine epsilon ϵ) up to d' ≈ 75, which corresponds to the smallest error rate representable in double precision (`realmin'). b. For several problems of fig. <ref>, relative differences in the outputs of the two methods from the Monte Carlo estimate, and computation times for 1% precision. | 2012.14331 | performance.jpg |
0.477286 | 2521014a0e1849028e9887def60665f8 | Illustrative deformation results of the simulation model at a time step. | 2012.14334 | DRM3D_motion3D_structure.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.