This paper introduces an unsupervised loss for training parametric deformation shape generators. The key idea is to enforce the preservation of local rigidity among the generated shapes. Our approach builds on an approximation of the as-rigid-as poss
ible (or ARAP) deformation energy. We show how to develop the unsupervised loss via a spectral decomposition of the Hessian of the ARAP energy. Our loss nicely decouples pose and shape variations through a robust norm. The loss admits simple closed-form expressions. It is easy to train and can be plugged into any standard generation models, e.g., variational auto-encoder (VAE) and auto-decoder (AD). Experimental results show that our approach outperforms existing shape generation approaches considerably on public benchmark datasets of various shape categories such as human, animal and bone.
In this paper, utilizing Wangs Harnack inequality with power and the Banach fixed point theorem, the weak well-posedness for distribution dependent SDEs with integrable drift is investigated. In addition, using a trick of decoupled method, some regul
arity such as relative entropy and Sobolevs estimate of invariant probability measure are proved. Furthermore, by comparing two stationary Fokker-Planck-Kolmogorov equations, the existence and uniqueness of invariant probability measure for McKean-Vlasov SDEs are obtained by log-Sobolevs inequality and Banachs fixed theorem. Finally, some examples are presented.
The propagation of light in moving media is dragged by atomic motion. The light-drag effect can be dramatically enhanced by reducing the group velocity with electro-magnetically induced transparency. We demonstrate a systematic procedure to estimate
the velocity field of the moving atoms, by holographically reconstructing the complex wavefront of the slow light and to simultaneously retrieve the absorption and phase shift. This large-NA, photon-shot-noise-limited inline coherent imaging technique may assist a wide range of cold atom experiments to access phase space information with in situ and minimally destructive measurements. By faithfully expanding the imaging data from real to complex numbers, the holographic technique also paves a way toward single shot spectroscopic imaging of atomic ensembles, even in presence of atomic density fluctuations.
We demonstrate that the discrepancy between the anomalous magnetic moment measured at BNL and Fermilab and the Standard Model prediction could be explained within the context of low-scale gravity and large extra-dimensions. The dominant contribution
to $(g-2)_mu$ originates in Kaluza-Klein (KK) excitations (of the lepton gauge boson) which do not mix with quarks (to lowest order) and therefore can be quite light avoiding LHC constraints. We show that the KK contribution to $(g-2)_mu$ is universal with the string scale entering as an effective cutoff. The KK tower provides a unequivocal distinctive signal which will be within reach of the future muon smasher.
Very recently, the Muon $g-2$ experiment at Fermilab has confirmed the E821 Brookhaven result, which hinted at a deviation of the muon anomalous magnetic moment from the Standard Model (SM) expectation. The combined results from Brookhaven and Fermil
ab show a difference with the SM prediction $delta a_mu = (251 pm 59) times 10^{-11}$ at a significance of $4.2sigma$, strongly indicating the presence of new physics. Motivated by this new result we reexamine the contributions to the muon anomalous magnetic moment from both: (i)~the ubiquitous $U(1)$ gauge bosons of D-brane string theory constructions and (ii)~the Regge excitations of the string. We show that, for a string scale ${cal O} ({rm PeV})$, the contribution from anomalous $U(1)$ gauge bosons which couple to hadrons could help to reduce (though not fully eliminate) the discrepancy reported by the Muon $g-2$ Collaboration. Consistency with null results from LHC searches of new heavy vector bosons imparts the dominant constraint. We demonstrate that the contribution from Regge excitations is strongly suppressed as it was previously conjectured. We also comment on contributions from Kaluza-Klein (KK) modes, which could help resolve the $delta a_mu$ discrepancy. In particular, we argue that for 4-stack intersecting D-brane models, the KK excitations of the $U(1)$ boson living on the lepton brane would not couple to hadrons and therefore can evade the LHC bounds while fully bridging the $delta a_mu$ gap observed at Brookhaven and Fermilab.
In this paper, the discrete parameter expansion is adopted to investigate the estimation of heat kernel for Euler-Maruyama scheme of SDEs driven by {alpha}-stable noise, which implies krylovs estimate and khasminskiis estimate. As an application, the
convergence rate of Euler-Maruyama scheme of a class of multidimensional SDEs with singular drift( in aid of Zvonkins transformation) is obtained.
Unsupervised domain adaptive object detection aims to adapt detectors from a labelled source domain to an unlabelled target domain. Most existing works take a two-stage strategy that first generates region proposals and then detects objects of intere
st, where adversarial learning is widely adopted to mitigate the inter-domain discrepancy in both stages. However, adversarial learning may impair the alignment of well-aligned samples as it merely aligns the global distributions across domains. To address this issue, we design an uncertainty-aware domain adaptation network (UaDAN) that introduces conditional adversarial learning to align well-aligned and poorly-aligned samples separately in different manners. Specifically, we design an uncertainty metric that assesses the alignment of each sample and adjusts the strength of adversarial learning for well-aligned and poorly-aligned samples adaptively. In addition, we exploit the uncertainty metric to achieve curriculum learning that first performs easier image-level alignment and then more difficult instance-level alignment progressively. Extensive experiments over four challenging domain adaptive object detection datasets show that UaDAN achieves superior performance as compared with state-of-the-art methods.
Due to their intrinsic link with nonlinear Fokker-Planck equations and many other applications, distribution dependent stochastic differential equations (DDSDEs for short) have been intensively investigated. In this paper we summarize some recent pro
gresses in the study of DDSDEs, which include the correspondence of weak solutions and nonlinear Fokker-Planck equations, the well-posedness, regularity estimates, exponential ergodicity, long time large deviations, and comparison theorems.
We first derive the boundary theory from the U(1) Chern-Simons theory. We then introduce the Wilson line and discuss the effective action on an $n$-sheet manifold from the back-reaction of the Wilson line. The reason is that the U(1) Chern-Simons the
ory can provide an exact effective action when introducing the Wilson line. This study cannot be done in the SL(2) Chern-Simons formulation of pure AdS$_3$ Einstein gravity theory. It is known that the expectation value of the Wilson line in the pure AdS$_3$ Einstein gravity is equivalent to entanglement entropy in the boundary theory up to classical gravity. We show that the boundary theory of the U(1) Chern-Simons theory deviates by a self-interaction term from the boundary theory of the AdS$_3$ Einstein gravity theory. It provides a convenient path to the building of minimum surface=entanglement entropy in the SL(2) Chern-Simons formulation. We also discuss the Hayward term in the SL(2) Chern-Simons formulation to compare with the Wilson line approach. To reproduce the entanglement entropy for a single interval at the classical level, we introduce two wedges under a regularization scheme. We propose the quantum generalization by combining the bulk and Hayward terms. The quantum correction of the partition function vanishes. In the end, we exactly calculate the entanglement entropy for a single interval. The pure AdS$_3$ Einstein gravity theory shows a shift of central charge by 26 at the one-loop level. The U(1) Chern-Simons theory does not have such a shift from the quantum effect, and the result is the same in the weak gravitational constant limit. The non-vanishing quantum correction shows the naive quantum generalization of the Hayward term is incorrect.
We present the design and test results of a Drivers and Limiting AmplifierS ASIC operating at 10 Gbps (DLAS10) and three Miniature Optical Transmitter/Receiver/Transceiver modules (MTx+, MRx+, and MTRx+) based on DLAS10. DLAS10 can drive two Transmit
ter Optical Sub-Assemblies (TOSAs) of Vertical Cavity Surface Emitting Lasers (VCSELs), receive the signals from two Receiver Optical Sub-Assemblies (ROSAs) that have no embedded limiting amplifiers, or drive a VCSEL TOSA and receive the signal from a ROSA, respectively. Each channel of DLAS10 consists of an input Continuous Time Linear Equalizer (CTLE), a four-stage limiting amplifier (LA), and an output driver. The LA amplifies the signals of variable levels to a stable swing. The output driver drives VCSELs or impedance-controlled traces. DLAS10 is fabricated in a 65 nm CMOS technology. The die is 1 mm x 1 mm. DLAS10 is packaged in a 4 mm x 4 mm 24-pin quad-flat no-leads (QFN) package. DLAS10 has been tested in MTx+, MRx+, and MTRx+ modules. Both measured optical and electrical eye diagrams pass the 10 Gbps eye mask test. The input electrical sensitivity is 40 mVp-p, while the input optical sensitivity is -12 dBm. The total jitter of MRx+ is 29 ps (P-P) with a random jitter of 1.6 ps (RMS) and a deterministic jitter of 9.9 ps. Each MTx+/MTRx+ module consumes 82 mW/ch and 174 mW/ch, respectively.
