• Publicationsclose
• Research softwareclose
• 2018-2022close
• Preprint close
• Article close
• AE close
• arXiv.org e-Print Archive close

Creating compelling captions for data visualizations has been a longstanding challenge. Visualization researchers are typically untrained in journalistic reporting and hence the captions that are placed below data visualizations tend to be not overly engaging and rather just stick to basic observations about the data. In this work we explore the opportunities offered by the newly emerging crop of large language models (LLM) which use sophisticated deep learning technology to produce human-like prose. We ask, can these powerful software devices be purposed to produce engaging captions for generic data visualizations like a scatterplot. It turns out that the key challenge lies in designing the most effective prompt for the LLM, a task called prompt engineering. We report on first experiments using the popular LLM GPT-3 and deliver some promising results.

We propose a 6D RGB-D odometry approach that finds the relative camera pose between consecutive RGB-D frames by keypoint extraction and feature matching both on the RGB and depth image planes. Furthermore, we feed the estimated pose to the highly accurate KinectFusion algorithm, which uses a fast ICP (Iterative Closest Point) to fine-tune the frame-to-frame relative pose and fuse the depth data into a global implicit surface. We evaluate our method on a publicly available RGB-D SLAM benchmark dataset by Sturm et al. The experimental results show that our proposed reconstruction method solely based on visual odometry and KinectFusion outperforms the state-of-the-art RGB-D SLAM system accuracy. Moreover, our algorithm outputs a ready-to-use polygon mesh (highly suitable for creating 3D virtual worlds) without any postprocessing steps.

Selected stories about the life of A. L. Onishchik, and a review of his contribution to the classification of non-split supermanifolds, in particular, supercurves a.k.a. superstrings; his editorial and educational work. A brief overview of his and his students' results in supersymmetry, and their impact on other researchers. Several open problems growing out of Onishchik's research are presented, some of them are related with odd parameters of deformations and non-holonomic structures of supermanifolds important in physical models, such as Minkowski superspaces and certain superstrings. Comment: 24pp

Existing deep learning-based 3D object detectors typically rely on the appearance of individual objects and do not explicitly pay attention to the rich contextual information of the scene. In this work, we propose Contextualized Multi-Stage Refinement for 3D Object Detection (CMR3D) framework, which takes a 3D scene as input and strives to explicitly integrate useful contextual information of the scene at multiple levels to predict a set of object bounding-boxes along with their corresponding semantic labels. To this end, we propose to utilize a context enhancement network that captures the contextual information at different levels of granularity followed by a multi-stage refinement module to progressively refine the box positions and class predictions. Extensive experiments on the large-scale ScanNetV2 benchmark reveal the benefits of our proposed method, leading to an absolute improvement of 2.0% over the baseline. In addition to 3D object detection, we investigate the effectiveness of our CMR3D framework for the problem of 3D object counting. Our source code will be publicly released. 5 figures, 10 pages including references

In its many variants, randomized benchmarking (RB) is a broadly used technique for assessing the quality of gate implementations on quantum computers. A detailed theoretical understanding and general guarantees exist for the functioning and interpretation of RB protocols if the gates under scrutiny are drawn uniformly at random from a compact group. In contrast, many practically attractive and scalable RB protocols implement random quantum circuits with local gates randomly drawn from some gate-set. Despite their abundance in practice, for those non-uniform RB protocols, general guarantees for gates from arbitrary compact groups under experimentally plausible assumptions are missing. In this work, we derive such guarantees for a large class of RB protocols for random circuits that we refer to as filtered RB. Prominent examples include linear cross-entropy benchmarking, character benchmarking, Pauli-noise tomography and variants of simultaneous RB. Building upon recent results for random circuits, we show that many relevant filtered RB schemes can be realized with random quantum circuits in linear depth, and we provide explicit small constants for common instances. We further derive general sample complexity bounds for filtered RB. We show filtered RB to be sample-efficient for several relevant groups, including protocols addressing higher-order cross-talk. Our theory for non-uniform filtered RB is, in principle, flexible enough to design new protocols for non-universal and analog quantum simulators. Comment: 59 pages, 5 figures. Accepted for a talk at QIP 2023. v3: more accessible overview of results and high-level discussion; improved structure and readability

We develop a high accuracy power series method for solving partial differential equations with emphasis on the nonlinear Schr\"odinger equations. The accuracy and computing speed can be systematically and arbitrarily increased to orders of magnitude larger than those of other methods. Machine precision accuracy can be easily reached and sustained for long evolution times within rather short computing time. In-depth analysis and characterisation for all sources of error are performed by comparing the numerical solutions with the exact analytical ones. Exact and approximate boundary conditions are considered and shown to minimise errors for solutions with finite background. The method is extended to cases with external potentials and coupled nonlinear Schr\"odinger equations. Comment: 39 pages, 20 figures

We present Camelira, a web-based Arabic multi-dialect morphological disambiguation tool that covers four major variants of Arabic: Modern Standard Arabic, Egyptian, Gulf, and Levantine. Camelira offers a user-friendly web interface that allows researchers and language learners to explore various linguistic information, such as part-of-speech, morphological features, and lemmas. Our system also provides an option to automatically choose an appropriate dialect-specific disambiguator based on the prediction of a dialect identification component. Camelira is publicly accessible at http://camelira.camel-lab.com.

We introduce a model for Non-Standard neutral current Interaction (NSI) between neutrinos and the matter fields, with an arbitrary coupling to the up and down quarks. The model is based on a new $U(1)$ gauge symmetry with a light gauge boson that mixes with the photon. We show that the couplings to the $u$ and $d$ quarks can have a ratio such that the contribution from NSI to the Coherent Elastic Neutrino-Nucleus Scattering (CE$\nu$NS) amplitude vanishes, relaxing the bound on the NSI from the CE$\nu$NS experiments. Additionally, the deviation of the measured value of the anomalous magnetic dipole moment of the muon from the standard-model prediction can be fitted. The most limiting constraints on our model come from the search for the decay of the new gauge boson to $e^-e^+$ and invisible particles, carried out by NA48/2 and NA64, respectively. We show that these bounds can be relaxed by opening up the decay of the new gauge boson to new light scalars that eventually decay into the $e^- e^+$ pairs. We show that there are ranges that can lead to both a solution to the $(g - 2)_\mu$ anomaly and values of $\epsilon_{\mu \mu} = \epsilon_{\tau \tau}$ large enough to be probed by future solar neutrino experiments. Comment: 13+4 pages, 4 figures. V2 matches the published version in PRD

Context. More than 10000 Jupiter Trojans have been detected so far. They are moving around the L4 and L5 triangular Lagrangian points of the Sun-Jupiter system and their distributions can provide important clues to the early evolution of the Solar System. Aims. The number asymmetry of the L4 and L5 Jupiter Trojans is a longstanding problem. We aim to test a new mechanism in order to explain this anomalous feature by invoking the jumping-Jupiter scenario. Methods. First, we introduce the orbital evolution of Jupiter caused by the giant planet instability in the early Solar System. In this scenario, Jupiter could undergo an outward migration at a very high speed. We then investigate how such a jump changes the numbers of the L4 (N4) and L5 (N5) Trojans. Results. The outward migration of Jupiter can distort the co-orbital orbits near the Lagrangian points, resulting in L4 Trojans being more stable than the L5 ones. We find that, this mechanism could potentially explain the unbiased number asymmetry of N4/N5~1.6 for the known Jupiter Trojans. The uncertainties of the system parameters, e.g. Jupiter's eccentricity and inclination, the inclination distribution of Jupiter Trojans, are also taken into account and our results about the L4/L5 asymmetry have been further validated. However, the resonant amplitudes of the simulated Trojans are excited to higher values compared to the current population. A possible solution is that collisions among the Trojans may reduce their resonant amplitudes. 11 pages, 5 figures, accepted for publication in A&A