text
stringlengths
54
548k
label
stringclasses
4 values
id_
stringlengths
32
32
This shift of the modality from the structured (image/video) to unstructured (skeleton) data type provides benefits in terms of data compaction, computation, storage, scalability, and recognition accuracy. Furthermore, the skeleton-related attributes mimic actual physical traits in human body and can be utilized as a s...
i
8a436f303f71e59c34d4a780629b7521
Unimodular gravity (UG) can be defined by imposing {{formula:8a0d0f5d-9fe7-42b1-9502-a62039d8c496}} as a constraint, where {{formula:ebd10828-6443-4d06-bc71-e4d9b9918f0b}} is a nondynamic background volume element. One often chooses {{formula:6013a195-a154-4ad0-ad1b-5e986a46394b}} , hence the name unimodular gravity....
i
b516706a208061f8187713622e567c15
Baselines We first benchmark FeTA against vanilla transformer {{cite:a6f2a7e1d088d18fcaf2c6d57bb757eb88a545f3}}. To study impact of position encoding on FeTA, we compare against recent PE-induced vanilla transformer based approaches, namely Graph-BERT {{cite:e5e85405b16a2f9e52c11caa23f035c739e61e45}}, GT {{cite:2a33340...
r
7c02ba395bdb38ca6318264bcb4d3567
When attempting to derive the Hawking radiation, as far as our quantum improved black hole can be viewed as an effective classical geometry and geometrical optics for quantum fields can be applied as in the Hawking's original derivation {{cite:4514b8c1151477229a172e844ea0fe8c78fe20ab}}, one can naturally expect that th...
d
4f259cbfc557609c3e816f8d682d8b9b
Digital traces such as social media posts or interactions appear to be extremely powerful in gaining knowledge about our personal lives and predicting political beliefs {{cite:0eba81c6615d2ff202d97eea00ccf6f8e813c727}}. Automated analysis of electoral support on social media could effectively replace traditional survey...
i
058437b5c668c167c8388718f69f3ecf
For thorough evaluation, we report the results in terms of different attributes in Table REF . It can be seen that the best five performers achieve the best performance on various subsets. Specifically, improved from state-of-the-art L2S method {{cite:de817bab38898e6b5065dbc206d843ba564e4500}}, FPNCC (REF ) obtains the...
r
efaae8de62895f14179e2550269118b2
In this part we state some known results, which are important to prove Theorem REF and Corollary REF . As for the proofs for these known results, we would like to refer the interested readers to {{cite:c58760a2cf83650aae3f9fd22767a6586d9f8af4}} and {{cite:3993601a78740a917c3356f9a709a4d1513a4b75}}.
r
9e88845c0dbcbc724b9dbca7881487f2
return the combined coset {{formula:f3ae5605-233f-4276-b3af-9d5d3833b8ff}} That the procedure works correctly follows directly from lemma:isoaut-item2 of lemma:isoaut given the coset decomposition {{formula:fdab1baf-26ac-411a-ad74-5866a9edfc02}} . Line REF is justified by the fact that {{formula:22d8f36c-6de2-4977-...
m
6cbe29a2d0940dab3e88ce6501c6ae65
We can propose several avenues in continuing this research. In our work we calculated the relativistic effects in the context of perturbed {{formula:32829e71-ac5f-44bd-837c-7d0c0e149932}} CDM to obtain an estimate of this effect. Implicitly this assumes that fluctuating dark energy does not alter this effect. A complet...
d
0810dae6a875f5d3b1c824cd854b7e93
Split Mini-Imagenet: Mini-Imagenet {{cite:474972c882b4f162daefbd4a3a71b39726ae8942}} is a subset of ImageNet {{cite:d631f519a6545ceaa9ddcc0b737c046d2052395e}} with a total of 100 classes and 600 images per class. Each task consists of 10 disjoint subset of classes from these 100 classes. Similar to CIFAR variants, 2500...
r
13353cb3cbe36f04039e871a632d2f69
Theorems REF and REF are two new additions to a long line of results identifying tropical moduli spaces with non-Archimedean skeletons of the Berkovich analytic spaces associated to their algebraic counterparts. We refer the reader to {{cite:24a9c530037aee981b22718e8c8a16a27c431e2f}} for the case of Jacobians and to ...
d
14fe1edd9df4a82bee0fa969f158b5a5
Learning representations by combining quality diversity (here, MAP-Elites) and generative models (here, a VAE) opens promising research avenues for domains in which optimizations of the same cost function are launched continuously. This is, for example, the case of Model Predictive Control {{cite:75d9a4051cbe7bfd896d65...
d
a94f7def08f3cd3b4524057448e2934d
We also add the the test–particle lines (TP), which are given by {{cite:e5bbbaa1b3a546acd12c1ae1b8c8be68c6b75220}} {{formula:7b3efc27-9aaa-4f52-9a14-0e0f181735b5}}
r
676e3ea343e24a82a62caf2732132619
With the emergence of deep learning, DML has gained significant popularity because it can seamlessly incorporate the strength of deep learning into metric learning {{cite:3c465b15848be2894b5c0c26a9aa0ce0bf8e4dc1}}. The DML automatically constructs task-specific distance metrics in a weakly supervised manner. This resul...
i
8ca163bd681c26256c3397158d614001
In this paper, we did not fully explore the effect of the threshold {{formula:a3abd31b-66df-4918-8dc2-589b089e4ab5}} , but rather chose to focus on methods which are less sensitive to this parameter. In our case, there were no significant differences between the results obtained with the two tested threshold, which is ...
d
2101bab5770e5d12ed6740c963f85c65
The spectrum of known heavy-quark exotic hadrons continues to expand frequently, with about 50 candidates observed to date. Almost all have a valence light-flavor content consisting of only {{formula:0a3f2a9f-fc34-4ad9-b3ab-61ed3460625b}} and {{formula:62a82c3a-e7f5-44b9-83db-7edf3d621d86}} quarks, but very recently ...
i
778173094cf01ea0fb0f4826ca0f8dba
Results on FER+. In both settings (with and without data augmentation), the local SVM approach yields better accuracy rates than the global SVM approach on FER+, but the differences are not as high as for FER 2013. Without data augmentation, our combination of deep and handcrafted features attains an accuracy of {{form...
r
a3c07d4ef85c8edabe48da5ac15e795b
As we wrote in the main paper that we also conducted experiments of many-shot and 20-shot, here we show the results in Table REF . From the results we can make the following observations: (i) The downstream transfer performance is approximately correlated to the in-domain self-supervised learning evaluation, i.e., the ...
r
cdc0d59addbffff0e1135e9fa1ea9102
where the angular quantities {{formula:f702d9de-95c2-42ee-b9b9-54a035f56aa7}} describe a direction of {{formula:6ada4c25-9ba2-45e7-83fe-492c03bf38b7}} in the {{formula:4dbdc2eb-313b-43b8-b9c2-4702fadf2c19}} rest frame with respect to some specified axes, {{formula:ce948a9a-189f-41bd-a02f-bdc42e01839d}} are the angu...
i
18f5366c07adab2a85c927f650571335
Furthermore, left-symmetric algebras are a kind of natural algebraic systems appearing in many fields in mathematics and mathematical physics. Perhaps this is one of the most attractive and interesting places. As it was pointed out in a paper of Chapoton and Livernet {{cite:eb9c722f033a20df601c9d9b5421a5032a00c0ee}}, t...
i
25c1f31ca8f4211c430265085e178e47
According to {{cite:943a756ac70cb4ecda746664710e23bbd688abb5}}, for instance, there is a good correlation between the radius of the H i and the stellar disc for spiral galaxies. Thus, this (local closed-box) model suggests indeed a correlation between the local metallicity at the effective radius and the H i mass frac...
r
ea863f5b397b783b79200a135c5f7ba8
Explainable Artificial Intelligence (XAI), especially Explainable Reinforcement Learning (XRL) {{cite:b7476be0c3d2e503f83e41214aa7f9770d078b4c}} is attracting more attention recently. How to interpret the action choices in reinforcement learning (RL) policies remains a critical challenge, especially as the gradually in...
i
911aea4b9f916be916b3644430f102c6
Regularization IOT consists of different ordered blocks in a weight tying method, which may looks like a parameter regularization to some extent. However, we show that IOT is more than regularization and can be complementary with other regularization methods. Setting (1): We first train a Transformer model on IWSLT14 D...
d
d17eb88be9f6ddb8b4d78f45864f2fb4
It is well-known that the Type-II weak topology on the probability space {{formula:d0cc670e-a380-4dcf-9399-b41e2f54d6ed}} can be induced by the Prohorov metric if the ambient space {{formula:049909d7-052f-4d70-a0e2-9febe8bec674}} is a separable and complete metric space, see {{cite:95941a31aaf878e1de561da7672fe2a6474...
d
0fa6ad386db2d9f79ad45667bebfe0af
In addition to the binary-based representation, which makes it difficult to capture the inter-category relationship, {{formula:d32b0fbd-f125-424d-acb4-60bc66436557}} -Net uses the Euclidean distance (see Table REF of Appendix ), which is prone to be less discriminant in high-dimensional data {{cite:e283a45532c2bd517bb...
m
756c8aaacbb098fcbe943197beb6f681
At 1 AU, the narrowband oblique whistlers are most often associated with SIRs, often filling the downstream region of increased solar wind speed, and often variable magnetic field and lower electron parallel beta. The waves are also seen within CMEs {{cite:b461949e55e2fdc0be4f5d591c747ac458526cad}}. As shown in Figure ...
d
715a94c5483a64e07e985d698bc0fa9f
(Part REF ) We show that 3 implies 1. Denote {{formula:9c671e68-6963-40e2-acc4-226695511e27}} . Let {{formula:ff6408da-d908-481b-a897-28d87ca6a3c5}} small enough that {{formula:ddc52cb1-4cce-414a-ad1e-74fa96a9eaac}} . As the pointwise limit of convex functions, {{formula:a6522e94-89e4-48e4-9883-41a524fd9f8b}} is conv...
r
9f85ed7123a16dbc126e85edfaf44cd2
Accreting black holes (BHs) are bright X-ray sources powered by the gravitational potential energy of accreting materials. The luminosities of Galactic BHs are strongly limited to the Eddington luminosity ({{formula:a6f4348d-ede6-4975-a090-8889cedb9d4d}} ), at which the radiation pressure balances the inward gravitatio...
i
681d7b6555169841fdf462c3b2430d8f
The recent new effects at the LHC have shown signals of di-bosons that have masses around 2 TeV {{cite:86a80056fc864d8b13b87a8c112695e554ebc663}}. It has raised hopes for the existence of new bosons in the Zee-babu model. An interesting effect in this model is the small mass generation of neutrino at the two-loop level...
d
b471a6a97b81b740dc012e5b4860b225
Code properties. Our findings suggest that while certain code characteristics can be extracted from pre-trained code transformers with linear classifiers, implying that they are firmly encoded in the hidden layers, others, such as cyclomatic complexity, cannot be extracted as effectively. We plan to investigate whether...
d
c60fbf3419d91563b68c734d470d0c55
A quasi-deformation of {{formula:5a32b7eb-9940-43cb-84b2-0ff1158db79d}} is a diagram {{formula:e2903128-ed6e-4124-9117-7534f478a4f8}} of local homomorphisms, in which {{formula:84f65bf2-b49a-4b55-a66a-3d59e811e66b}} is faithfully flat, and {{formula:6804421a-7046-41eb-a32a-baf24f8d7ced}} is surjective with kernel g...
r
cf544af4ceaa7b124757dd7a637864f7
In Section Methodology we introduce a novel secure inference protocol that aims to improve the time and communication cost while preserving the privacy of all parties. We consider that a server has a well-trained deep neural network and would like to expose it as a service while preserving the privacy of both the data ...
i
9d175e44343776d96e832024ecc63544
The anisotropic part of the Knight shift comes from the magnetic dipolar hyperfine interaction with {{formula:f386495e-1ad0-43bc-9767-2029ce32134b}} spins. The different anisotropy of the hyperfine coupling tensor suggests the site-dependent {{formula:ec393eba-c8df-4b70-84a9-8e49c3af2992}} orbital occupation ({{formu...
d
1e1be24f33b52299158608d0e143d6ca
To leverage inertial, energetic, and cost benefits of small-scale robots, critical future applications of autonomous technologies may depend on coordinating large numbers of agents with minimal onboard sensing and communication resources. However, a critical problem for autonomous multi-robot groups is that state-of-th...
d
17ac4dfed79b90f48b506588d91852c7
The low error rate of SC processors makes them an ideal platform for quantum error correction (QEC) {{cite:1410f099dc8a46b66fcf72671fafcde374d13f9c}}, {{cite:f1063e3f9d04878f86b14b40deaa1085f023b538}}, {{cite:e1aa10a2ba607b0fcef5290e874ed784151fdd91}}, {{cite:94196dc8485dfb6cffa59ff42d17ab18886473ea}}, {{cite:3ed6ad776...
i
98876af23603f287d2bfc5ec15025061
The implementation details of our model are illustrated in Table REF . In the proposed model, the CNN sub-network consists a convolutional layer, a max-pooling layer and a flatten layer. The RNN sub-structure includes a hidden layer with 200 neurons. The MDN sub-model is composed of a hidden layer and an output layer. ...
r
1980bb150dffc230386072f4d3b0b618
In this paper, since the video source for our benchmark is Kinetics {{cite:e65b10cc3c485520a93f08cb2775fce6c39a9a1b}}, our target boundaries are event boundaries. This is because each video corresponds to one single, dominant event at the whole video level. Its videos usually are of 10s long, which is longer than the e...
d
f53dd21cc05837b2d0e14410c2c7e5c4
Lemma 1 (Lemma 48 in {{cite:b6fe36bc8d87156e54ca524c3c33cf7e85b804de}}) Let {{formula:394993f2-8f98-4732-87fd-da567403d432}} be a matrix that is {{formula:30cd8788-9215-44f3-8502-351b24c8e8b3}} -row-sparse and {{formula:ab84b822-e4fd-4119-a12d-69ebe52154ec}} -column-sparse, and each element of {{formula:cbf35044-482...
r
554089a9b386ab9aaed924d1dedbeb5f
For forward Euler type discretization, Assumption REF can be relaxed to {{formula:c75e0fa8-514e-4c77-afd8-61add1e32407}} only. The backword Euler type discretization needs {{formula:34546e39-3014-45b5-b9ba-8797675fc623}} , and other discretization methods need different {{formula:6c58a9a9-852b-411c-9183-1cf377ccdeed}...
m
8e4ce53290423ba4f7660af5c3697679
The investigation of the no-scalar hair theorem and interplay between gravity-scalar systems are not just of theoretical concern, and there are many reasons to examine them. Scalar fields play a central role in cosmology and particle physics {{cite:d65ff374cc65b4e786d27ae0e6086789709cfcf4}}. Scalar fields appear natura...
i
1df5048a8d9a7f3b0ac8ae2ba5f1adac
We would like to thank Mona Dentler, Colin Hill, Evan McDonough, and Mikhail Ivanov for useful comments and suggestions. RH is a CIFAR Azrieli Global Scholar, Gravity & the Extreme Universe Program, 2019, and a 2020 Alfred. P. Sloan Research Fellow. RH is supported by Natural Sciences and Engineering Research Council o...
d
6818197cfd7feabbf0c86b41a04b6282
Augmentations.   Using adversarial loss with very limited data might cause mode collapse. Thus, we use augmentations in the form of random affine transformations during training. However, when preserving the background is necessary as in Fig. REF (Bottom), these augmentations might leak to the generated frames resulti...
m
dee6ef22ad37839a7096ca7b864d65de
Lite-HRNet is the state-of-the-art in terms of complexity and accuracy trade-off on COCO and MPII human pose estimation and easily generalized to semantic segmentation task. Related Work Efficient blocks for classification. Separable convolutions and group convolutions have been increasingly popular in lightweight netw...
i
8540673ab829b02e195d207c6f872f0f
In design of our method, we are partly inspired by Monte Carlo tree search {{cite:595d173646bf1709e466c9b9b1f4a14f349e92db}} (MCTS), often used for exploring large trees, mainly for applications in combinatorial games. To the best of our knowledge, most research, where a tree search method is combined with ML, focuses ...
i
56d47b038925a935ff64f7e443e52361
In the end one needs to comment on possible importance of the strong electronic correlations in Na{{formula:b07b1de4-80e4-430c-9d77-94833d4c55f7}} RuO{{formula:a631dbb4-be37-460c-9e06-f0e95024de4e}} . We have seen that the GGA overestimates equilibrium volume of the uniform phase on {{formula:b1295e70-5672-4d2e-a8c5-7f...
r
e8fabe3be158315871c2bcf9903b04db
Researchers have suggested many additive defense methods in response to such additive attacks. The current defenses against such attacks are training modified data focusing on gradient pathways{{cite:94263fdc488767cab547c8d7f7eacf6aff96593b}}, performing different filtering, or removing adversarial perturbation from th...
i
3008cc2a87d9e6fa05193864bc2da807
We chose to emphasize the authoring objective of storiability because of our focus on AI-augmented story writing. Storiability is not a one-size-fits-all metric for this research. The quality of a story can be judged on multiple dimensions that are often not consistently defined across different studies, as discussed i...
d
8e99912968bf0f78467556a2429d30c4
MIME, like ATISS, needs a pre-defined floor plan room layout as input. The resolution of the 2D floor plan is coarse; i.e., 1 pixel stands for around 10 centimeters, which is extracted as a 512 dimension feature by ResNet-18. Introducing a finer floor plan representation, such as dividing one floor plan into multiple p...
d
210465cfaf44ee411d3cdccb9b48ec73
In particular, we selected 15 of 66 regions in table A.2 of {{cite:f436c81c6a2bd9cb9fbffa98b5efd08dabc2f14a}}, based purely on the SDSS DR14 spectra, as satisfying our criteria. We added 71 data points for 43 regions from the literature and this work, which include the great majority of published direct O/H estimates w...
m
58f5250934fe50eefb3b9269b39f9f90
A sim-to-real strategy to utilize intermediate encoded features of a fully trained end-to-end DRL policy for perceiving regions that could cause instabilities for the robot during navigation in real world uneven outdoor environments. IMU and elevation gradient based rewards to identify critical elevations and rough s...
i
03638e5a0e5a7eeb90eded734c23d0f5
Generally speaking, loss functions can be divided into two categories: convex loss functions and nonconvex loss functions. Convex loss functions including Hinge loss, square loss are the most commonly used. If {{formula:1a86bab6-f699-4e21-9a3e-6d3a58cf312c}} is a convex loss function, then the classical representer th...
m
d10e6f62e23355011221fe8ddc14ca6d
The one-dimensional (1D) continuous fermionic gas with repulsive delta-function interaction, which in this paper we call 1D repulsive fermion model, was one of the first quantum problems solved by the Bethe ansatz (BA) {{cite:f6ec6c917f7366075744fc5f7c36aed0edd7652c}}. This was achieved by Yang {{cite:96aab1e8195891985...
i
b5cb65fa191439fd33d751bd0bcbd01e
Figures 5(a) and 5(b) show the band structure and spin density of the ferromagnetic Zr{{formula:1fca8ff9-c39b-4a0c-bbef-22ce3d0643f7}} S(001) surface, respectively, calculated without including SOC. We find that the {{formula:2811013e-8d18-4165-b42f-d0fbae29ad6f}} and {{formula:b40346f1-d11d-4251-9ba7-ba89e36337b8}} ...
r
0ef3229cf4a243c2552aeb6b9479ddcc
for all {{formula:2a55ecb6-5ca8-4b0a-b894-e5d9f31ce67b}} , i.e., the Arrow-Pratt measure of risk aversion is bounded below by {{formula:c6269ca1-7829-4992-9a86-a75999fe65d3}} . The Arrow-Pratt measure of risk aversion {{cite:90f62a293f8b6e3897a3b04e3bb27f7f00dec11c}} has many well-known desired theoretical properties a...
r
2277dac9c865f861ea9930b1fb7efef0
Datasets. We evaluated our method in the following two standard benchmarks for UDA. Office-Home {{cite:3c4f706abd8159653e61908d5c5c6c9099fab33f}} consists of images of everyday objects organized into four domains: Artistic (Ar), Clipart (Cl), Product (Pr), and Real-world (Rw). It contains 15,500 images of 65 classes. V...
r
f45e5706b5bb85b12444a8cc366e53b3
In the present paper we consider the modern experimental techniques of potential modulation for the two-dimensional electron gas and show that they permit to obtain the quasiperiodic potentials on the plane with different numbers of quasiperiods. Then we use the topological results concerning the geometry of the level ...
i
593b95aedd4f3a5745f5853916897b86
The software implementation of the networks was done in Tensorflow {{cite:2a2320a961e10157d9e3be1db1e0dc717cb25f07}}. We used the network models and the test set provided in {{cite:5665aa4cc4b29afafc22918d76d6e11b4e4809e6}}, {{cite:5feaadfd83fcbff55b4a074b30c8fcacd41a1c1b}}, {{cite:3b5b3db504fe534c389bbabb4559d2745b93e...
m
9a13a04ab6034ca2121d7f2be6ec4ed4
Most existing studies on representation learning in BMDPs {{cite:fbeeca39d0c419f58019d930d1ac811354651b94}}, {{cite:87e2b867d127b9238774b49a3c315d39e232158c}}, {{cite:b115a8ffc8a833d1758aac224a56799bf5a82c66}}, {{cite:93f399c53d5ee12bd15a7be0b88125b5b7114406}}, {{cite:a1cd1e040736daa9e3eefe2462c964632d77b4e1}}, {{cite:...
i
8855939aed8e037a3ce8aa7bd52dc5e2
In this section, we analyze performance of Mask R-CNN on filament detection, in juxtaposition with Bernasconi's segmentations reported to HEK, using the metrics and methodologies discussed in Sec. As the reader is looking at the results in this section, it is important to bear in mind a few key points in our experiment...
r
1e58c12f1ad62f04426f2778cb2ffc51
To compute the continual learning metrics {{cite:9eb5d1788471034f52f7b585f3cb7781ada306d7}}, each predicted trajectory needs to be marked as accurate or inaccurate based on its difference from the ground truth. Since there is no preexisting procedure for this, we follow the following steps. We set a threshold on the DT...
r
d94ef98c6a4004aca90d6101f40f8856
Deep learning methods have been shown to be applicable to robot manipulation {{cite:8bfea282ee9434cdf16e4b00ac7ec86860381347}}, {{cite:c335866fe5073e249b9ad1078e85faccb0a592e1}}. Grasping is a common topic, as it is often an essential step in many robotic tasks. Learning-based methods can generalize to grasp objects wi...
m
8424098ac3b762181edd88dfefaf594c
ReconFormer. As shown in Fig. REF (a), ReconFormer consists of three recurrent units and a refine module (RM). To maintain high-resolution information, ReconFormer employs globally columnar structure. In particular, recurrent units map the input degraded images to the original dimension. Meanwhile, across each recurre...
m
9ae521f754519b13487a1817f9e3a2b3
Moreover, we highlight some recent state-of-art work in this domain, which relates to our proposed methodology, showing promising results in application areas other than UAV aerial imagery. First, the work in {{cite:643a5759e5beb78d12b764b54173f6c91e8421e9}} incorporates two significant improvements: layered boosting (...
d
5233bb0cd7fab16753851b8c54705e7a
In previous Sections we used the Krylov–Fock theory of unstable states to find late time properties of the survival amplitude and instantaneous energy. Similar estimations of the late time behavior of the survival amplitude can be also found by means another method, e.g. methods of the quantum scattering theory. This m...
d
76b41442324ff333c0cf3bd65c6a0754
In future work, it would be interesting to consider many-body effects in Maryland-type NHQCs and investigate their dynamical properties. The interplay between non-Hermiticity and quantum chaos in the NRMM can also be studied following the mapping discussed in Appendix . Under open boundary conditions, the interaction b...
d
37542850ea00367d8b0c859cfc6f23b5
State-of-the-art secure aggregation protocols essentially rely on two main principles: (1) pairwise random-seed agreement between users in order to generate masks that hide users’ models while having an additive structure that allows their cancellation when added at the server; (2) secret sharing of the random-seeds, i...
i
49703922071a8e21c4ff9cac0a8eb7af
One type of meta-learning methods is called gradient-based meta-learning, and they include MAML {{cite:47d8ce168aa1e2698a7bb7a949a303e6c8294329}} and Reptile {{cite:1dbf8e5d0538d038e7a294c442dbd2d4d5cd07d0}}. These methods learn to reduce the loss of the model whose parameters (e.g., weights of neural networks) are upd...
i
9e5e6413239b8cfa4de78f2819e0e615
Dynamic Magnetic Resonance Imaging (MRI) is a powerful imaging modality to non-invasively capture time evolving phenomena in the human body, such as the beating heart, motion of vocal tract during speaking, or dynamics of contrast uptake in brain. A long standing challenge in MRI is its slow imaging speed which restric...
i
5101a5a534070e2a8150b179c7a29c6a
Five representative weak forms from Tab. REF were chosen for studying elapsed time and memory consumption of the einsum evaluation backends from Section REF , namely the vector dot product, weighted vector dot product, weak Laplacian, Navier Stokes convection and linear elasticity terms. For comparison, their counterp...
r
72f250f3a01d7fda6f17882d06f923f5
Unfortunately, it is not know whether or not the evolution system {{formula:7822269c-c55e-401a-89c1-f08ca93b7714}} leaves {{formula:fe9d96c1-b0be-495c-a9d8-b7f0e316ccdd}} invariant and is strongly continuous on {{formula:76022fae-3541-42db-be8a-be2c150c432f}} ; the only information in this direction stems from (REF )...
r
7e1e9c599548cfc0e4477726eda1f3b2
First, the codes are simulated over the i.i.d. {{formula:5f7616e6-881f-4567-a83c-0101c7bbf2b7}} channel model of {{cite:396a138353a0756743644499075a0de5dd84d45b}}, where {{formula:a2b3e53e-d6c2-47e5-b0e1-39fd124526c8}} and {{formula:5c1e29df-e74e-4788-8325-a413e513ea8c}} errors are modeled as independent events iden...
r
9734fd3506427a1e4fc7b31fcd0e8429
Figure REF shows the individual exchange paths for each term in Eq. REF . The Kitaev term is an indirect exchange interaction with hopping matrix elements {{formula:c4969446-cdee-41d9-ba37-17aacf97e5d6}} between the {{formula:37453f99-dada-4bb5-ba10-6ccb643de6a0}} , {{formula:60345b48-73c2-4188-aebc-e267122f6d60}} , ...
m
5b012dd33ca60d5885bb6b053ab82725
From the point of the quark model, some of them can be interpreted as mesons made of quark-anti-quark pairs and baryons constructed from three quarks. However, there is a plethora of states whose structure is puzzling, presenting properties that can be fitted in the hadronic molecular state picture if the mass is close...
i
c665b00cb39cff43550eaf6a38ecf676
Imbalanced CIFAR-10 and CIFAR-100. The original version of CIFAR-10 and CIFAR-100 contains {{formula:54558c22-c841-4d99-b237-ac9e233b6019}} training images and {{formula:cad1067e-bf55-46e0-8c40-6cd06f3ef8ed}} validation images of size {{formula:8b82583f-9729-4046-ad10-ab7dde4a4758}} with 10 and 100 classes, respecti...
r
0b04328912c1adb9503216817268b532
To test the generalization capability of U-S-VAE, we provide its performance on different datasets, such as the NTU RGB+D 60 {{cite:a72e80f988f97ff7b0f0b248665c04fcaa1d47f3}}, which is a large scale dataset commonly used for testing action/gesture models. NTU RGB+D contains 60 action categories collected from 40 subjec...
d
b71dee056f1edafb699922bfcad23b25
The DQN model's implementation was done in Tensorflow {{cite:846a7c6afca44b3722948c7692eb2418e0005f1b}} on a dual NVIDIA GeForce 1080Ti GPU. The DQN framework is trained with an RMSProp optimizer {{cite:03c001e3510108d021642115111679d32750d3cc}} with a learning rate of {{formula:fb77ae83-b493-4738-ac2f-5d7ab3997f36}} ...
r
a3c5619091b2fd7a733d27dc76550c78
We build a new recognition paradigm to improve the transferability using knowledge from the textual encoder of the well-pretrained vision-language model. We conduct extensive experiments on popular video and image datasets (i.e., Kinetics-400 {{cite:d70b1e287a11265b5f748d1fb0caf5ac2e0ec065}}, UCF-101 {{cite:38f3e70c3...
i
f9cceb9f3e1c70a6817ed28c43f1720b
The dust is also destroyed by collisions with the nuclei in the hot plasma. The thermo-kinetic sputtering of the dust, is dominated by collisions with H and He nuclei. In a gas that is shocked to temperatures between (1–10){{formula:fd63cb27-5dcb-4457-836a-1a3ba56bab5c}}  K, the lifetime of silicate dust ({{formula:6a7...
d
53689364a62cbb69743991e014a7de60
In general relativity there are covariant quantum fields defined on a curved manifold {{formula:11800532-577f-46be-ae54-5a2bbfcc6c27}} which transform under isometries according to CRs induced by finite-dimensional representations (reps.) of the universal covering group {{formula:7bb1594f-5e35-45a7-95dc-9f7bf895f785}}...
i
e49e3dd65c58bde9b308bdb56c1e4c13
This relation also holds for classical mechanics, the classical limit of the above, given the correspondence between Poisson brackets and commutators, {{cite:25d2e2bd536797ffc1de013620222fb7085d6cd2}}, {{cite:edc4a960501cf923718b4fa9d2999bfa967f3efc}}, {{cite:e8776c41fd0e27592ad6636b492a354751940a61}}, {{cite:dd63c5182...
m
8e9186d42e289c983a6256a3f40aa4bc
Although network reports are designed to be flexible across different domains and use cases, the usefulness of network reports still relies on the integrity of the creators. In the near term, it is unlikely that the full contents of network reports could be automatically created and standardized to prevent all inapprop...
d
e60664aa1b3ebdcbd132e552da9a7cfa
Despite remark REF , we note that assumption REF is strict and difficult to prove that happens with high probability in arbitrary MDPs, as this depends on the markov chain underlying the decision process as well as the Q-values themselves during learning. Thus, we provide a probabilistic relaxation of this assumption ...
r
9825f6fd83670f763d353ccff7236f9e
Web APIs consist of programmable endpoints to interact with software systems. They can be implemented in different ways, including the well known REST {{cite:4c10d7adb57c8fb67ba0e9d0d42521cd3aa71d45}} and SOAP {{cite:8dcfb12193042f244d9c4d8bc94cb1c571bb227d}}, {{cite:e878b32c591a2847a4f590093e0b19409507fc07}} paradigms...
i
e55d1621e61a4eecb4e34c860a48b9c9
In a semiclassical treatment, a classical orbit {{formula:b243d998-bf47-4258-a1e5-b9d905418eb8}} satisfying the following Einstein-Brillouin-Keller (EBK) quantization condition {{cite:4def2d0f7d6179211cb07de303b5639257915f4a}}: {{formula:7e26c3ea-5a33-4b93-8795-5dd9cf1d0318}}
m
bc37b35718fa0ad5436c6f6633c81807
Much research is conducted on CT images, but the 3D information of CT images is under-explored, such as the work by {{cite:bc09a7b23c3d27eada1cba131cf6c2eb6406c695}}, {{cite:fdd862ecaf1e678c2588560ad4eb5ab9fe657671}}, {{cite:3c4402b97d45bae6562046ce0c6a232c0065db24}}. These work mainly propose 2D DL models for COVID-19...
m
967ccbb5280c8ac4f85e76829603cb8f
Finally, in fig. REF we have plotted the spatial diffusion coefficient as a function of temperature scaled with critical temperature. Expressing the transport coefficients in the units of thermal wavelength is very convenient; thus, we have multiplied a factor of {{formula:4f270fde-f488-470b-a595-b11de846f346}} , maki...
r
5d941e6dc9c7c9561c64032f0ae7bba2
Our conclusion is that the effective theory breaks down for any physical process involving outgoing particles with {{formula:89f046ce-0a47-41e2-8415-ddf320bf5743}} (including Hawking radiation) when the collapsing matter is around a Planck length from the horizon. This is reminiscent of the brick-wall model {{cite:5f5...
d
1202b0a9c9acb3fa999def0534bd2737
The second lemma is the Pólya theorem, see Theorem 9.1.4 in {{cite:d6140798325ea250364facbe2b221dfc12949ff2}}.
r
f864eba58ca89c085e1bdcd8c20fdef3
We computed the QR probability for two atomic species of experimental relevance (Rb {{cite:7f589817bcea119da93bea1f6934eaf1d87ec2eb}} and Na {{cite:58d3b494b690221a9fc0fd529882fd723fc84bcd}}, {{cite:3603f170b3169333a9e0f24baf5828acabbff269}}), and for three graphene family materials already synthesized (stanene {{cite:...
m
def5e53f7712b2b136b8c75a660ce309
We remark that the above statements for the two and three-dimensional cases have been already proved by Bourgain–Li in {{cite:48c4b3fe4246a5047ee2e7f3d6567ce98cb67f78}} and {{cite:f2e6b3d41ed1adfa155eb06320355f7633cca9c8}}, respectively. We refer the interested readers to these works for an extensive list of literature...
r
30f065a2c4d28201b6dfd2861f0937f9
In addition to the ResNet50-based models discussed in sec:experiments,sec:results, we also present results from two models with transformer architectures {{cite:eb1e0eca0506a299242ce6822a0950d120a2fa0b}}. MAE uses masked autoencoders {{cite:8f5ae443b905d2b0bb94d4cc4a53e5e1d564f4b4}} and we use its ViT-Large variant. MO...
r
c67c60c85a154cf2c73c8bd1e8437115
Table REF reports the intent classification accuracy for different KD strategies in combination with 4 CL approaches, i.e. a rehearsal approach with 3 different sample selection strategies (random, iCaRL {{cite:bcd37c14273c1c04dc9680e45eff321f59271d0c}}, and “closest_to_mean”, where the samples which are closest to t...
r
76c257f5999d0f1378e29b2850755bf6
Categorical logic unifies languages: virtually any formalism, from a simple heap to the calculus of constructions, can be modelled as a structured category {{cite:3221941769a19a18a043023cfc69c72c6619b288}}. In doing so, we inherit a wealth of tools from category theory. In particular, we can generate expressive type sy...
i
36bd1dcf1d7516708b808d402303951d
During this presentation of tappings, a few additional issues came up that still need to be discussed. Models with memory like recurrent neural networks or dynamic Bayesian networks need special consideration with respect to tappings. Such models naturally retain an internal memory of past input values. Because of thi...
d
1a0880d9add8e3515d26cca18273cf2f
ViT: is an architecture based on transformer which is used in the field of Natural Language Processing (NLP) {{cite:524c9445c759c5903db1b54341d2130d6968218a}}. Internally, the transformer learns a relationship between input token pairs, and uses 16x16 patches of images as input tokens {{cite:1f707f04b09c6f577d577128bf1...
m
bbe7a2ba895daccfc49a14bd9f66fd7a
Cosmic Microwave Background: We consider the CMB temperature, polarization, and lensing reconstruction angular power spectra as measured by 2018 Planck legacy release  {{cite:74b6ef59c5105a294f9cef8cf08a1fc78ba9d266}}, {{cite:bb9f68b115c035ef1a36d260ea97c35341a7c1a0}}. We denote them as "Planck," and they include the...
m
d6e32409c16012e748cc3a113a771388
To best of our knowledge, we are first to observe that adopting existing approaches for fairness-aware FL is not feasible under data heterogeneity. This is because the client specific fairness losses will become unbounded (Section ). We argue that by altering the aggregation heuristics may ensure fairness while simul...
i
b0cd41f761bea3552f40487066b0ab41
Finally, Figure REF illustartes the class-wise vulnerability of the ICARL {{cite:468e6aa9c8d84e3641d5542f1abfbb8a0164df1c}} against the FGSM, PGD, and CW {{cite:cceff7ffb4d3b29c215d08507950b6a5e69f61c6}}, {{cite:c3fccf0ff9253b92c42d2bf0bfc9f109ecfd1a1a}}, {{cite:3519d381d3816b9d84b2d65e2bb04bb34bae094c}} adversarial a...
r
0c4ae849d083568c29f8d375d5e6943c
Moreover, in contrast with conventional multi-modal approaches that assume paired input, our proposed method is more flexible since it can process samples with missing chest X-ray images. There is a rising interest in learning cross-modal interactions between modalities during training time and in reconstructing missin...
d
e7355e8f54ffdbb5b62c9281776c2100
Compared with the passive RIS, a noticeable feature of active RIS is that it needs additional power to supply the amplifiers. Therefore, the active RIS requires a larger power budget than the passive RIS given the same number of reflecting elements. Even though the superiority of the active RIS over the passive RIS has...
i
17930d8f9f3fe990be8a972285447421