6D pose estimation of rigid objects from a single RGB image has seen tremendous improvements recently by using deep learning to combat complex real-world variations, but a majority of methods build models on the per-object level, failing to scale to
multiple objects simultaneously. In this paper, we present a novel approach for scalable 6D pose estimation, by self-supervised learning on synthetic data of multiple objects using a single autoencoder. To handle multiple objects and generalize to unseen objects, we disentangle the latent object shape and pose representations, so that the latent shape space models shape similarities, and the latent pose code is used for rotation retrieval by comparison with canonical rotations. To encourage shape space construction, we apply contrastive metric learning and enable the processing of unseen objects by referring to similar training objects. The different symmetries across objects induce inconsistent latent pose spaces, which we capture with a conditioned block producing shape-dependent pose codebooks by re-entangling shape and pose representations. We test our method on two multi-object benchmarks with real data, T-LESS and NOCS REAL275, and show it outperforms existing RGB-based methods in terms of pose estimation accuracy and generalization.
This paper presents a novel mutual information (MI) matrix based method for fault detection. Given a $m$-dimensional fault process, the MI matrix is a $m times m$ matrix in which the $(i,j)$-th entry measures the MI values between the $i$-th dimensio
n and the $j$-th dimension variables. We introduce the recently proposed matrix-based Renyis $alpha$-entropy functional to estimate MI values in each entry of the MI matrix. The new estimator avoids density estimation and it operates on the eigenspectrum of a (normalized) symmetric positive definite (SPD) matrix, which makes it well suited for industrial process. We combine different orders of statistics of the transformed components (TCs) extracted from the MI matrix to constitute the detection index, and derive a simple similarity index to monitor the changes of characteristics of the underlying process in consecutive windows. We term the overall methodology projections of mutual information matrix (PMIM). Experiments on both synthetic data and the benchmark Tennessee Eastman process demonstrate the interpretability of PMIM in identifying the root variables that cause the faults, and its superiority in detecting the occurrence of faults in terms of the improved fault detection rate (FDR) and the lowest false alarm rate (FAR). The advantages of PMIM is also less sensitive to hyper-parameters. The advantages of PMIM is also less sensitive to hyper-parameters. Code of PMIM is available at https://github.com/SJYuCNEL/Fault_detection_PMIM
We consider a homogeneous mixture of bosons and polarized fermions. We find that long-range and attractive fermion-mediated interactions between bosons have dramatic effects on the properties of the bosons. We construct the phase diagram spanned by b
oson-fermion mass ratio and boson-fermion scattering parameter. It consists of stable region of mixing and unstable region toward phase separation. In stable mixing phase, the collective long-wavelength excitations can either be well-behaved with infinite lifetime or be finite in lifetime suffered from the Landau damping. We examine the effects of the induced interaction on the properties of weakly interacting bosons. It turns out that the induced interaction not only enhances the repulsion between the bosons against collapse but also enhances the stability of the superfluid state by suppressing quantum depletion.
With the rise of third parties in the machine learning pipeline, the service provider in Machine Learning as a Service (MLaaS), or external data contributors in online learning, or the retraining of existing models, the need to ensure the security of
the resulting machine learning models has become an increasingly important topic. The security community has demonstrated that without transparency of the data and the resulting model, there exist many potential security risks, with new risks constantly being discovered. In this paper, we focus on one of these security risks -- poisoning attacks. Specifically, we analyze how attackers may interfere with the results of regression learning by poisoning the training datasets. To this end, we analyze and develop a new poisoning attack algorithm. Our attack, termed Nopt, in contrast with previous poisoning attack algorithms, can produce larger errors with the same proportion of poisoning data-points. Furthermore, we also significantly improve the state-of-the-art defense algorithm, termed TRIM, proposed by Jagielsk et al. (IEEE S&P 2018), by incorporating the concept of probability estimation of clean data-points into the algorithm. Our new defense algorithm, termed Proda, demonstrates an increased effectiveness in reducing errors arising from the poisoning dataset through optimizing ensemble models. We highlight that the time complexity of TRIM had not been estimated; however, we deduce from their work that TRIM can take exponential time complexity in the worst-case scenario, in excess of Prodas logarithmic time. The performance of both our proposed attack and defense algorithms is extensively evaluated on four real-world datasets of housing prices, loans, health care, and bike sharing services. We hope that our work will inspire future research to develop more robust learning algorithms immune to poisoning attacks.
To solve large-scale or high-resolution topology optimization problem, a novel algorithm is developed based on modified bi-directional evolutionary structure optimization (BESO) and extended finite element method (XFEM). Within XFEM, a set of enriche
d nodes are defined to divide the finite element into several uniform sub-regions, i.e. sub-triangles and sub-tetrahedrons. The material grid and shape functions are defined on each sub-region to improve the computational accuracy, whereas the equilibrium equation is established on the level of coarse finite elements to increase the computational efficiency. We set all the standard FE nodes and the enriched nodes as the design variables, and a modified material interpolation model is introduced to calculate the material properties for sub-regions. An enrichment function originating from modeling voids scheme is adopted to character the discontinuity between solid material to void material. To efficiently use the gradient-based algorithm, BESO, sensitivity analysis is performed with the aid of adjoint method. Typical numerical examples, involving millions of design variables, are carried to verify the effectiveness of the proposed method.
For a split reductive group defined over a number field, we first introduce the notations of arithmetic torsors and arithmetic Higgs torsors. Then we construct arithmetic characteristic curves associated to arithmetic Higgs torsors, based on the Chev
alley characteristic morphism and the existence of Chevalley basis for the associated Lie algebra. As to be expected, this work is motivated by the works of Beauville-Narasimhan on spectral curves and Donagi-Gaistgory on cameral curves in algebraic geometry. In the forthcoming papers, we will use arithmetic characteristic curves to construct arithmetic Hitchin fibrations and study the intersection homologies and perverse sheaves for the associated structures, following Ngos approach to the fundamental lemma.
We study the ground-state behavior of a Bose-Einstein Condensate (BEC) in a Raman-laser-assisted one-dimensional (1D) optical lattice potential forming a multilayer system. We find that, such system can be described by an effective model with spin-or
bit coupling (SOC) of pseudospin $(N-1)/2$, where $N$ is the number of layers. Due to the intricate interplay between atomic interactions, SOC and laser-assisted tunnelings, the ground-state phase diagrams generally consist of three phases -- a stripe, a plane wave and a normal phase with zero-momentum, touching at a quantum tricritical point. More important, even though the single-particle states only minimize at zero-momentum for odd $N$, the many-body ground states may still develop finite momenta. The underlying mechanisms are elucidated. Our results provide an alternative way to realize an effective spin-orbit coupling of Bose gas with the Raman-laser-assisted optical lattice, and would also be beneficial to the studies on SOC effects in spinor Bose systems with large spin.
Solitons play a fundamental role in dynamics of nonlinear excitations. Here we explore the motion of solitons in one-dimensional uniform Bose-Einstein condensates subjected to a spin-orbit coupling (SOC). We demonstrate that the spin dynamics of soli
tons is governed by a nonlinear Bloch equation. The spin dynamics influences the orbital motion of the solitons leading to the spin-orbit effects in the dynamics of the macroscopic quantum objects (mean-field solitons). The latter perform oscillations with a frequency determined by the SOC, Raman coupling, and intrinsic nonlinearity. These findings reveal unique features of solitons affected by the SOC, which is confirmed by analytical considerations and numerical simulations of the underlying Gross-Pitaevskii equations.
Motivated by a goal of realizing spin-orbit coupling (SOC) beyond one-dimension (1D), we propose and analyze a method to generate an effective 2D SOC in bilayer BECs with laser-assisted inter-layer tunneling. We show that an interplay between the int
er-layer tunneling, SOC and intra-layer atomic interaction can give rise to diverse ground state configurations. In particular, the system undergoes a transition to a new type of stripe phase which spontaneously breaks the time-reversal symmetry. Different from the ordinary Rashba-type SOC, a fractionalized skyrmion lattice emerges spontaneously in the bilayer system without external traps. Furthermore, we predict the occurrence of a tetracritical point in the phase diagram of the bilayer BECs, where four different phases merge together. The origin of the emerging different phases is elucidated.
We show two kinds of inhomogeneous spin domain possessing N{e}el-like domain walls in spin-1 Bose-Einstein condensate, which are induced by the positive and negative quadratic Zeeman effect (QZE) respectively. In both cases, the spin density distribu
tion is inhomogeneous and has zeros where the magnetization vanishes. For positive and negative QZE, the spin patterns and topological structures are remarkably different. Such phenomena are due to the pointwise different axisymmetry-breaking caused by the pointwise different population exchange between the sublevels, arising uniquely from the QZE. We analyze in detail the inhomogeneous domain formation and related experimental observations for the spin-1 $^{87}$Rb and $^{23}$Na condensate.
