arXiv subjects

Shuo Yang

Publications and source records attributed to Shuo Yang.

At least 343 records · Page 19Linked to original sources

arXiv2021

Speeding up Deep Model Training by Sharing Weights and Then Unsharing

We propose a simple and efficient approach for training the BERT model. Our approach exploits the special structure of BERT that contains a stack of repeated modules (i.e., transformer encoders). Our proposed approach first trains BERT with the weights shared across all the repeated modules till some point. This is for learning the commonly shared component of weights across all repeated layers. We then stop weight sharing and continue training until convergence. We present theoretic insights for training by sharing weights then unsharing with analysis for simplified models. Empirical experiments on the BERT model show that our method yields better performance of trained models, and significantly reduces the number of training iterations.

cs.LG↗

arXiv2021

Equality Constrained Linear Optimal Control With Factor Graphs

This paper presents a novel factor graph-based approach to solve the discrete-time finite-horizon Linear Quadratic Regulator problem subject to auxiliary linear equality constraints within and across time steps. We represent such optimal control problems using constrained factor graphs and optimize the factor graphs to obtain the optimal trajectory and the feedback control policies using the variable elimination algorithm with a modified Gram-Schmidt process. We prove that our approach has the same order of computational complexity as the state-of-the-art dynamic programming approach. Furthermore, current dynamic programming approaches can only handle equality constraints between variables at the same time step, but ours can handle equality constraints among any combination of variables at any time step while maintaining linear complexity with respect to trajectory length. Our approach can be used to efficiently generate trajectories and feedback control policies to achieve periodic motion or repetitive manipulation.

cs.RO↗

arXiv2021

Tensor network representations of fermionic crystalline topological phases on two-dimensional lattices

We investigate the tensor network representations of fermionic crystalline symmetry-protected topological (SPT) phases on two-dimensional lattices. As a mapping from virtual indices to physical indices, projected entangled-pair state (PEPS) serves as a concrete way to construct the wavefunctions of 2D crystalline fermionic SPT (fSPT) phases protected by 17 wallpaper group symmetries, for both spinless and spin-1/2 fermions. Based on PEPS, the full classification of 2D crystalline fSPT phases with wallpaper groups can be obtained. Tensor network states provide a natural framework for studying 2D crystalline fSPT phases.

cond-mat.str-el↗

arXiv2021

Free Lunch for Few-shot Learning: Distribution Calibration

Learning from a limited number of samples is challenging since the learned model can easily become overfitted based on the biased distribution formed by only a few training examples. In this paper, we calibrate the distribution of these few-sample classes by transferring statistics from the classes with sufficient examples, then an adequate number of examples can be sampled from the calibrated distribution to expand the inputs to the classifier. We assume every dimension in the feature representation follows a Gaussian distribution so that the mean and the variance of the distribution can borrow from that of similar classes whose statistics are better estimated with an adequate number of samples. Our method can be built on top of off-the-shelf pretrained feature extractors and classification models without extra parameters. We show that a simple logistic regression classifier trained using the features sampled from our calibrated distribution can outperform the state-of-the-art accuracy on two datasets (~5% improvement on miniImageNet compared to the next best). The visualization of these generated features demonstrates that our calibrated distribution is an accurate estimation.

cs.LG↗

arXiv2021

Semi-TCL: Semi-Supervised Track Contrastive Representation Learning

Online tracking of multiple objects in videos requires strong capacity of modeling and matching object appearances. Previous methods for learning appearance embedding mostly rely on instance-level matching without considering the temporal continuity provided by videos. We design a new instance-to-track matching objective to learn appearance embedding that compares a candidate detection to the embedding of the tracks persisted in the tracker. It enables us to learn not only from videos labeled with complete tracks, but also unlabeled or partially labeled videos. We implement this learning objective in a unified form following the spirit of constrastive loss. Experiments on multiple object tracking datasets demonstrate that our method can effectively learning discriminative appearance embeddings in a semi-supervised fashion and outperform state of the art methods on representative benchmarks.

cs.CV↗

arXiv2021

Positive-Congruent Training: Towards Regression-Free Model Updates

Reducing inconsistencies in the behavior of different versions of an AI system can be as important in practice as reducing its overall error. In image classification, sample-wise inconsistencies appear as "negative flips": A new model incorrectly predicts the output for a test sample that was correctly classified by the old (reference) model. Positive-congruent (PC) training aims at reducing error rate while at the same time reducing negative flips, thus maximizing congruency with the reference model only on positive predictions, unlike model distillation. We propose a simple approach for PC training, Focal Distillation, which enforces congruence with the reference model by giving more weights to samples that were correctly classified. We also found that, if the reference model itself can be chosen as an ensemble of multiple deep neural networks, negative flips can be further reduced without affecting the new model's accuracy.

cs.CV↗

arXiv2021

Compatibility-aware Heterogeneous Visual Search

We tackle the problem of visual search under resource constraints. Existing systems use the same embedding model to compute representations (embeddings) for the query and gallery images. Such systems inherently face a hard accuracy-efficiency trade-off: the embedding model needs to be large enough to ensure high accuracy, yet small enough to enable query-embedding computation on resource-constrained platforms. This trade-off could be mitigated if gallery embeddings are generated from a large model and query embeddings are extracted using a compact model. The key to building such a system is to ensure representation compatibility between the query and gallery models. In this paper, we address two forms of compatibility: One enforced by modifying the parameters of each model that computes the embeddings. The other by modifying the architectures that compute the embeddings, leading to compatibility-aware neural architecture search (CMP-NAS). We test CMP-NAS on challenging retrieval tasks for fashion images (DeepFashion2), and face images (IJB-C). Compared to ordinary (homogeneous) visual search using the largest embedding model (paragon), CMP-NAS achieves 80-fold and 23-fold cost reduction while maintaining accuracy within 0.3% and 1.6% of the paragon on DeepFashion2 and IJB-C respectively.

cs.CV↗

arXiv2021

Detecting anomaly in vector boson scattering

Measuring the vector boson scattering (VBS) precisely is an important step towards understanding the electroweak symmetry breaking of the standard model (SM) and detecting new physics beyond the SM. We propose a neural network which compress the features of the VBS into three dimensional latent space. The consistency of the SM prediction and the experimental data is tested by the binned log-likelihood analysis in the latent space. We will show that the network is capable of distinguish different polarization modes of $WWjj$ production in both dileptonic channel and semi-leptonic channel. The method is also applied to constrain the effective field theory and two Higgs Doublet Model. The results demonstrate that the method is sensitive to generic new physics contributing to the VBS.

hep-ph↗

arXiv2021

Optimal Control for Structurally Sparse Systems using Graphical Inference

Dynamical systems with a distributed yet interconnected structure, like multi-rigid-body robots or large-scale multi-agent systems, introduce valuable sparsity into the system dynamics that can be exploited in an optimal control setting for speeding up computation and improving numerical conditioning. Conventional approaches for solving the Optimal Control Problem (OCP) rarely capitalize on such structural sparsity, and hence suffer from a cubic computational complexity growth as the dimensionality of the system scales. In this paper, we present an OCP formulation that relies on graphical models to capture the sparsely-interconnected nature of the system dynamics. Such a representational choice allows the use of contemporary graphical inference algorithms that enable our solver to achieve a linear time complexity in the state and control dimensions as well as the time horizon. We demonstrate the numerical and computational advantages of our approach on a canonical dynamical system in simulation.

cs.RO↗

arXiv2021

Gamma-convergent projection-free finite element methods for nematic liquid crystals: The Ericksen model

The Ericksen model for nematic liquid crystals couples a director field with a scalar degree of orientation variable, and allows the formation of various defects with finite energy. We propose a simple but novel finite element approximation of the problem that can be implemented easily within standard finite element packages. Our scheme is projection-free and thus circumvents the use of weakly acute meshes, which are quite restrictive in 3D but are required by recent algorithms for convergence. We prove stability and $Γ$-convergence properties of the new method in the presence of defects. We also design an effective nested gradient flow algorithm for computing minimizers that controls the violation of the unit-length constraint of the director. We present several simulations in 2D and 3D that document the performance of the proposed scheme and its ability to capture quite intriguing defects.

math.NA↗

arXiv2021

Single-View 3D Object Reconstruction from Shape Priors in Memory

Existing methods for single-view 3D object reconstruction directly learn to transform image features into 3D representations. However, these methods are vulnerable to images containing noisy backgrounds and heavy occlusions because the extracted image features do not contain enough information to reconstruct high-quality 3D shapes. Humans routinely use incomplete or noisy visual cues from an image to retrieve similar 3D shapes from their memory and reconstruct the 3D shape of an object. Inspired by this, we propose a novel method, named Mem3D, that explicitly constructs shape priors to supplement the missing information in the image. Specifically, the shape priors are in the forms of "image-voxel" pairs in the memory network, which is stored by a well-designed writing strategy during training. We also propose a voxel triplet loss function that helps to retrieve the precise 3D shapes that are highly related to the input image from shape priors. The LSTM-based shape encoder is introduced to extract information from the retrieved 3D shapes, which are useful in recovering the 3D shape of an object that is heavily occluded or in complex environments. Experimental results demonstrate that Mem3D significantly improves reconstruction quality and performs favorably against state-of-the-art methods on the ShapeNet and Pix3D datasets.

cs.CV↗

arXiv2021

Combinatorial Bandits without Total Order for Arms

We consider the combinatorial bandits problem, where at each time step, the online learner selects a size-$k$ subset $s$ from the arms set $\mathcal{A}$, where $\left|\mathcal{A}\right| = n$, and observes a stochastic reward of each arm in the selected set $s$. The goal of the online learner is to minimize the regret, induced by not selecting $s^*$ which maximizes the expected total reward. Specifically, we focus on a challenging setting where 1) the reward distribution of an arm depends on the set $s$ it is part of, and crucially 2) there is \textit{no total order} for the arms in $\mathcal{A}$. In this paper, we formally present a reward model that captures set-dependent reward distribution and assumes no total order for arms. Correspondingly, we propose an Upper Confidence Bound (UCB) algorithm that maintains UCB for each individual arm and selects the arms with top-$k$ UCB. We develop a novel regret analysis and show an $O\left(\frac{k^2 n \log T}ε\right)$ gap-dependent regret bound as well as an $O\left(k^2\sqrt{n T \log T}\right)$ gap-independent regret bound. We also provide a lower bound for the proposed reward model, which shows our proposed algorithm is near-optimal for any constant $k$. Empirical results on various reward models demonstrate the broad applicability of our algorithm.

cs.LG↗

arXiv2021

CelebA-Spoof Challenge 2020 on Face Anti-Spoofing: Methods and Results

As facial interaction systems are prevalently deployed, security and reliability of these systems become a critical issue, with substantial research efforts devoted. Among them, face anti-spoofing emerges as an important area, whose objective is to identify whether a presented face is live or spoof. Recently, a large-scale face anti-spoofing dataset, CelebA-Spoof which comprised of 625,537 pictures of 10,177 subjects has been released. It is the largest face anti-spoofing dataset in terms of the numbers of the data and the subjects. This paper reports methods and results in the CelebA-Spoof Challenge 2020 on Face AntiSpoofing which employs the CelebA-Spoof dataset. The model evaluation is conducted online on the hidden test set. A total of 134 participants registered for the competition, and 19 teams made valid submissions. We will analyze the top ranked solutions and present some discussion on future work directions.

cs.CV↗

arXiv2021

DeeperForensics Challenge 2020 on Real-World Face Forgery Detection: Methods and Results

This paper reports methods and results in the DeeperForensics Challenge 2020 on real-world face forgery detection. The challenge employs the DeeperForensics-1.0 dataset, one of the most extensive publicly available real-world face forgery detection datasets, with 60,000 videos constituted by a total of 17.6 million frames. The model evaluation is conducted online on a high-quality hidden test set with multiple sources and diverse distortions. A total of 115 participants registered for the competition, and 25 teams made valid submissions. We will summarize the winning solutions and present some discussions on potential research directions.

cs.CV↗

arXiv2021

The ANTARES Astronomical Time-Domain Event Broker

We describe the Arizona-NOIRLab Temporal Analysis and Response to Events System (ANTARES), a software instrument designed to process large-scale streams of astronomical time-domain alerts. With the advent of large-format CCDs on wide-field imaging telescopes, time-domain surveys now routinely discover tens of thousands of new events each night, more than can be evaluated by astronomers alone. The ANTARES event broker will process alerts, annotating them with catalog associations and filtering them to distinguish customizable subsets of events. We describe the data model of the system, the overall architecture, annotation, implementation of filters, system outputs, provenance tracking, system performance, and the user interface.

astro-ph.IM↗

arXiv2020

LDG approximation of large deformations of prestrained plates

A reduced model for large deformations of prestrained plates consists of minimizing a second order bending energy subject to a nonconvex metric constraint. The former involves the second fundamental form of the middle plate and the later is a restriction on its first fundamental form. We discuss a formal derivation of this reduced model along with an equivalent formulation that makes it amenable computationally. We propose a local discontinuous Galerkin (LDG) finite element approach that hinges on the notion of reconstructed Hessian. We design discrete gradient flows to minimize the ensuing nonconvex problem and to find a suitable initial deformation. We present several insightful numerical experiments, some of practical interest, and assess various computational aspects of the approximation process.

math.NA↗

arXiv2020

SMOT: Single-Shot Multi Object Tracking

We present single-shot multi-object tracker (SMOT), a new tracking framework that converts any single-shot detector (SSD) model into an online multiple object tracker, which emphasizes simultaneously detecting and tracking of the object paths. Contrary to the existing tracking by detection approaches which suffer from errors made by the object detectors, SMOT adopts the recently proposed scheme of tracking by re-detection. We combine this scheme with SSD detectors by proposing a novel tracking anchor assignment module. With this design SMOT is able to generate tracklets with a constant per-frame runtime. A light-weighted linkage algorithm is then used for online tracklet linking. On three benchmarks of object tracking: Hannah, Music Videos, and MOT17, the proposed SMOT achieves state-of-the-art performance.

cs.CV↗

arXiv2020

Secure Your Intention: On Notions of Pre-Opacity in Discrete-Event Systems

This paper investigates an important informationflow security property called opacity in partially-observed discrete-event systems. We consider the presence of a passive intruder (eavesdropper) that knows the dynamic model of the system and can use the generated information-flow to infer some "secret" of the system. A system is said to be opaque if it always holds the plausible deniability for its secret. Existing notions of opacity only consider secret either as currently visiting some secret states or as having visited some secret states in the past. In this paper, we investigate information-flow security from a new angle by considering the secret of the system as the intention to execute some particular behavior of importance in the future. To this end, we propose a new class of opacity called pre-opacity that characterizes whether or not the intruder can predict the visit of secret states a certain number of steps ahead before the system actually does so. Depending the prediction task of the intruder, we propose two specific kinds of pre-opacity called K-step instant pre-opacity and K-step trajectory pre-opacity to specify this concept. For each notion of pre-opacity, we provide a necessary and sufficient condition as well as an effective verification algorithm. The complexity for the verification of pre-opacity is exponential in the size of the system as we show that pre-opacity is inherently PSPACE-hard. Finally, we generalize our setting to the case where the secret intention of the system is modeled as executing a particular sequence of events rather than visiting a secret state.

eess.SY↗