arXiv subjects

Shuo Yang

Publications and source records attributed to Shuo Yang.

At least 307 records · Page 17Linked to original sources

arXiv2023

You Don't Know When I Will Arrive: Unpredictable Controller Synthesis for Temporal Logic Tasks

In this paper, we investigate the problem of synthesizing controllers for temporal logic specifications under security constraint. We assume that there exists a passive intruder (eavesdropper) that can partially observe the behavior of the system. For the purpose of security, we require that the system's behaviors are unpredictable in the sense that the intruder cannot determine for sure that the system will exactly accomplish the task in $K$ steps ahead. This problem is particularly challenging since future information is involved in the synthesis process. We propose a novel information structure that predicts the effect of control in the future. A sound and complete algorithm is developed to synthesize a controller which ensures both task completion and security guarantee. The proposed approach is illustrated by a case study of robot task planning.

eess.SY↗

arXiv2023

Construction of Non-Hermitian Parent Hamiltonian from Matrix Product States

There are various research strategies used for non-Hermitian systems, which typically involve introducing non-Hermitian terms to pre-existing Hermitian Hamiltonians. It can be challenging to directly design non-Hermitian many-body models that exhibit unique features not found in Hermitian systems. In this Letter, we propose a new method to construct non-Hermitian many-body systems by generalizing the parent Hamiltonian method into non-Hermitian regimes. This allows us to build a local Hamiltonian using given matrix product states as its left and right ground states. We demonstrate this method by constructing a non-Hermitian spin-$1$ model from the asymmetric Affleck-Kennedy-Lieb-Tasaki (AKLT) state, which preserves both chiral order and symmetry-protected topological order. Our approach opens up a new paradigm for systematically constructing and studying non-Hermitian many-body systems, providing guiding principles to explore new properties and phenomena in non-Hermitian physics.

quant-ph↗

arXiv2023

Rethink Long-tailed Recognition with Vision Transformers

In the real world, data tends to follow long-tailed distributions w.r.t. class or attribution, motivating the challenging Long-Tailed Recognition (LTR) problem. In this paper, we revisit recent LTR methods with promising Vision Transformers (ViT). We figure out that 1) ViT is hard to train with long-tailed data. 2) ViT learns generalized features in an unsupervised manner, like mask generative training, either on long-tailed or balanced datasets. Hence, we propose to adopt unsupervised learning to utilize long-tailed data. Furthermore, we propose the Predictive Distribution Calibration (PDC) as a novel metric for LTR, where the model tends to simply classify inputs into common classes. Our PDC can measure the model calibration of predictive preferences quantitatively. On this basis, we find many LTR approaches alleviate it slightly, despite the accuracy improvement. Extensive experiments on benchmark datasets validate that PDC reflects the model's predictive preference precisely, which is consistent with the visualization.

cs.CV↗

arXiv2023

Learning Imbalanced Data with Vision Transformers

The real-world data tends to be heavily imbalanced and severely skew the data-driven deep neural networks, which makes Long-Tailed Recognition (LTR) a massive challenging task. Existing LTR methods seldom train Vision Transformers (ViTs) with Long-Tailed (LT) data, while the off-the-shelf pretrain weight of ViTs always leads to unfair comparisons. In this paper, we systematically investigate the ViTs' performance in LTR and propose LiVT to train ViTs from scratch only with LT data. With the observation that ViTs suffer more severe LTR problems, we conduct Masked Generative Pretraining (MGP) to learn generalized features. With ample and solid evidence, we show that MGP is more robust than supervised manners. In addition, Binary Cross Entropy (BCE) loss, which shows conspicuous performance with ViTs, encounters predicaments in LTR. We further propose the balanced BCE to ameliorate it with strong theoretical groundings. Specially, we derive the unbiased extension of Sigmoid and compensate extra logit margins to deploy it. Our Bal-BCE contributes to the quick convergence of ViTs in just a few epochs. Extensive experiments demonstrate that with MGP and Bal-BCE, LiVT successfully trains ViTs well without any additional data and outperforms comparable state-of-the-art methods significantly, e.g., our ViT-B achieves 81.0% Top-1 accuracy in iNaturalist 2018 without bells and whistles. Code is available at https://github.com/XuZhengzhuo/LiVT.

cs.CV↗

arXiv2023

Dataset Pruning: Reducing Training Data by Examining Generalization Influence

The great success of deep learning heavily relies on increasingly larger training data, which comes at a price of huge computational and infrastructural costs. This poses crucial questions that, do all training data contribute to model's performance? How much does each individual training sample or a sub-training-set affect the model's generalization, and how to construct the smallest subset from the entire training data as a proxy training set without significantly sacrificing the model's performance? To answer these, we propose dataset pruning, an optimization-based sample selection method that can (1) examine the influence of removing a particular set of training samples on model's generalization ability with theoretical guarantee, and (2) construct the smallest subset of training data that yields strictly constrained generalization gap. The empirically observed generalization gap of dataset pruning is substantially consistent with our theoretical expectations. Furthermore, the proposed method prunes 40% training examples on the CIFAR-10 dataset, halves the convergence time with only 1.3% test accuracy decrease, which is superior to previous score-based sample selection methods.

cs.LG↗

arXiv2023

Chebyshev pseudosite matrix product state approach for the spectral functions of electron-phonon coupling systems

The electron-phonon ($e$-ph) coupling system often has a large number of phonon degrees of freedom, whose spectral functions are numerically difficult to compute using matrix product state (MPS) formalisms. To solve this problem, we propose a new and practical method that combines the Chebyshev MPS and the pseudosite density matrix renormalization group (DMRG) algorithm. The Chebyshev vector is represented by a pseudosite MPS with global $U(1)$ fermion symmetry, which maps $2^{N_p}$ bosonic degrees of freedom onto $N_p$ pseudosites, each with two states. This approach can handle arbitrary $e$-ph coupling Hamiltonians where pseudosite DMRG performs efficiently. We use this method to study the spectral functions of the doped extended Hubbard-Holstein model in a regime of strong Coulomb repulsion, which has not been studied extensively before. Key features of the excitation spectra are captured at a modest computational cost. Our results show that weak extended $e$-ph couplings can increase the spectral weight of the holon-folding branch at low phonon frequencies, in agreement with angle-resolved photoemission observations on one-dimensional (1D) cuprates.

cond-mat.str-el↗

arXiv2023

Differentiable Safe Controller Design through Control Barrier Functions

Learning-based controllers, such as neural network (NN) controllers, can show high empirical performance but lack formal safety guarantees. To address this issue, control barrier functions (CBFs) have been applied as a safety filter to monitor and modify the outputs of learning-based controllers in order to guarantee the safety of the closed-loop system. However, such modification can be myopic with unpredictable long-term effects. In this work, we propose a safe-by-construction NN controller which employs differentiable CBF-based safety layers, and investigate the performance of safe-by-construction NN controllers in learning-based control. Specifically, two formulations of controllers are compared: one is projection-based and the other relies on our proposed set-theoretic parameterization. Both methods demonstrate improved closed-loop performance over using CBF as a separate safety filter in numerical experiments.

eess.SY↗

arXiv2023

Fast Contact-Implicit Model-Predictive Control

We present a general approach for controlling robotic systems that make and break contact with their environments. Contact-implicit model predictive control (CI-MPC) generalizes linear MPC to contact-rich settings by utilizing a bi-level planning formulation with lower-level contact dynamics formulated as time-varying linear complementarity problems (LCPs) computed using strategic Taylor approximations about a reference trajectory. These dynamics enable the upper-level planning problem to reason about contact timing and forces, and generate entirely new contact-mode sequences online. To achieve reliable and fast numerical convergence, we devise a structure-exploiting interior-point solver for these LCP contact dynamics and a custom trajectory optimizer for the tracking problem. We demonstrate real-time solution rates for CI-MPC and the ability to generate and track non-periodic behaviours in hardware experiments on a quadrupedal robot. We also show that the controller is robust to model mismatch and can respond to disturbances by discovering and exploiting new contact modes across a variety of robotic systems in simulation, including a pushbot, planar hopper, planar quadruped, and planar biped.

cs.RO↗

arXiv2023

Noise effects on purity and quantum entanglement in terms of physical implementability

Quantum decoherence due to imperfect manipulation of quantum devices is a key issue in the noisy intermediate-scale quantum (NISQ) era. Standard analyses in quantum information and quantum computation use error rates to parameterize quantum noise channels. However, there is no explicit relation between the decoherence effect induced by a noise channel and its error rate. In this work, we propose to characterize the decoherence effect of a noise channel by the physical implementability of its inverse, which is a universal parameter quantifying the difficulty to simulate the noise inverse with accessible quantum channels. We establish two concise inequalities connecting the decrease of the state purity and logarithmic negativity after a noise channel to the physical implementability of the noise inverse, which is required to be decomposed as mutually orthogonal unitaries or product channels respectively. Our results are numerically demonstrated on several commonly adopted two-qubit noise models. We believe that these relations contribute to the theoretical research on the entanglement properties of noise channels and provide guiding principles for quantum circuit design.

quant-ph↗

arXiv2022

PartImageNet: A Large, High-Quality Dataset of Parts

It is natural to represent objects in terms of their parts. This has the potential to improve the performance of algorithms for object recognition and segmentation but can also help for downstream tasks like activity recognition. Research on part-based models, however, is hindered by the lack of datasets with per-pixel part annotations. This is partly due to the difficulty and high cost of annotating object parts so it has rarely been done except for humans (where there exists a big literature on part-based models). To help address this problem, we propose PartImageNet, a large, high-quality dataset with part segmentation annotations. It consists of $158$ classes from ImageNet with approximately $24,000$ images. PartImageNet is unique because it offers part-level annotations on a general set of classes including non-rigid, articulated objects, while having an order of magnitude larger size compared to existing part datasets (excluding datasets of humans). It can be utilized for many vision tasks including Object Segmentation, Semantic Part Segmentation, Few-shot Learning and Part Discovery. We conduct comprehensive experiments which study these tasks and set up a set of baselines. The dataset and scripts are released at https://github.com/TACJu/PartImageNet.

cs.CV↗

arXiv2022

CPMLHO:Hyperparameter Tuning via Cutting Plane and Mixed-Level Optimization

The hyperparameter optimization of neural network can be expressed as a bilevel optimization problem. The bilevel optimization is used to automatically update the hyperparameter, and the gradient of the hyperparameter is the approximate gradient based on the best response function. Finding the best response function is very time consuming. In this paper we propose CPMLHO, a new hyperparameter optimization method using cutting plane method and mixed-level objective function.The cutting plane is added to the inner layer to constrain the space of the response function. To obtain more accurate hypergradient,the mixed-level can flexibly adjust the loss function by using the loss of the training set and the verification set. Compared to existing methods, the experimental results show that our method can automatically update the hyperparameters in the training process, and can find more superior hyperparameters with higher accuracy and faster convergence.

cs.LG↗

arXiv2022

Towards Explainability in Modular Autonomous Vehicle Software

Safety-critical Autonomous Systems require trustworthy and transparent decision-making process to be deployable in the real world. The advancement of Machine Learning introduces high performance but largely through black-box algorithms. We focus the discussion of explainability specifically with Autonomous Vehicles (AVs). As a safety-critical system, AVs provide the unique opportunity to utilize cutting-edge Machine Learning techniques while requiring transparency in decision making. Interpretability in every action the AV takes becomes crucial in post-hoc analysis where blame assignment might be necessary. In this paper, we provide positioning on how researchers could consider incorporating explainability and interpretability into design and optimization of separate Autonomous Vehicle modules including Perception, Planning, and Control.

cs.RO↗

arXiv2022

Variational Matrix Product State Approach for Non-Hermitian System Based on a Companion Hermitian Hamiltonian

Non-Hermitian systems exhibiting topological properties are attracting growing interest. In this work, we propose an algorithm for solving the ground state of a non-Hermitian system in the matrix product state (MPS) formalism based on a companion Hermitian Hamiltonian. If the eigenvalues of the non-Hermitian system are known, the companion Hermitian Hamiltonian can be directly constructed and solved using Hermitian variational methods. When the eigenvalues are unknown, a gradient descent along with the companion Hermitian Hamiltonian yields both the ground state eigenenergy and the eigenstate. With the variational principle as a solid foundation, our algorithm ensures convergence and provides results in excellent agreement with the exact solutions of the non-Hermitian Su-Schrieffer-Heeger (nH-SSH) model as well as its interacting extension. The approach we present avoids solving any non-Hermitian matrix and overcomes numerical instabilities commonly encountered in large non-Hermitian systems.

quant-ph↗

arXiv2022

Social norms of fairness with reputation-based role assignment in the dictator game

A vast body of experiments share the view that social norms are major factors for the emergence of fairness in a population of individuals playing the dictator game (DG). Recently, to explore which social norms are conducive to sustaining cooperation has obtained considerable concern. However, thus far few studies have investigated how social norms influence the evolution of fairness by means of indirect reciprocity. In this study, we propose an indirect reciprocal model of the DG and consider that an individual can be assigned as the dictator due to its good reputation. We investigate the `leading eight' norms and all second-order social norms by a two-timescale theoretical analysis. We show that when role assignment is based on reputation, four of the `leading eight' norms, including stern judging and simple standing, lead to a high level of fairness, which increases with the selection intensity. Our work also reveals that not only the correct treatment of making a fair split with good recipients but also distinguishing unjustified unfair split from justified unfair split matters in elevating the level of fairness.

cs.GT↗

arXiv2022

Quantum error mitigation via matrix product operators

In the era of noisy intermediate-scale quantum (NISQ) devices, the number of controllable hardware qubits is insufficient to implement quantum error correction (QEC). As an alternative, quantum error mitigation (QEM) can suppress errors in measurement results via repeated experiments and postprocessing of data. Typical techniques for error mitigation, e.g., the quasi-probability decomposition method, ignore correlated errors between different gates. Here, we introduce a QEM method based on the matrix product operator (MPO) representation of a quantum circuit that can characterize the noise channel with polynomial complexity. Our technique is demonstrated on a $\rm{depth}=20$ fully parallel quantum circuit of up to $N_q=20$ qubits undergoing local and global noise. The circuit error is reduced by several times with only a small bond dimension $D^{\prime} = 1$ for the noise channel. The MPO representation increases the accuracy of modeling noise without consuming more experimental resources, which improves the QEM performance and broadens its scope of application. Our method is hopeful of being applied to circuits in higher dimensions with more qubits and deeper depth.

quant-ph↗

arXiv2022

Manipulating synthetic gauge fluxes via multicolor dressing of Rydberg-atom arrays

Arrays of highly excited Rydberg atoms can be used as powerful quantum simulation platforms. Here, we introduce an approach that makes it possible to implement fully controllable effective spin interactions in such systems. We show that optical Rydberg dressing with multicolor laser fields opens up distinct interaction channels that enable complete site-selective control of the induced interactions and favorable scaling with respect to decoherence. We apply this method to generate synthetic gauge fields for Rydberg excitations where the effective magnetic flux can be manipulated at the single-plaquette level by simply varying the phase of the local dressing field. The system can be mapped to a highly anisotropic Heisenberg model, and the resulting spin interaction opens the door for explorations of topological phenomena with nonlocal density interactions. A remarkable consequence of the interaction is the emergence of topologically protected long-range doublons, which exhibit strongly correlated motion in a chiral and robust manner.

quant-ph↗

arXiv2022

Towards Regression-Free Neural Networks for Diverse Compute Platforms

With the shift towards on-device deep learning, ensuring a consistent behavior of an AI service across diverse compute platforms becomes tremendously important. Our work tackles the emergent problem of reducing predictive inconsistencies arising as negative flips: test samples that are correctly predicted by a less accurate model, but incorrectly by a more accurate one. We introduce REGression constrained Neural Architecture Search (REG-NAS) to design a family of highly accurate models that engender fewer negative flips. REG-NAS consists of two components: (1) A novel architecture constraint that enables a larger model to contain all the weights of the smaller one thus maximizing weight sharing. This idea stems from our observation that larger weight sharing among networks leads to similar sample-wise predictions and results in fewer negative flips; (2) A novel search reward that incorporates both Top-1 accuracy and negative flips in the architecture search metric. We demonstrate that \regnas can successfully find desirable architectures with few negative flips in three popular architecture search spaces. Compared to the existing state-of-the-art approach, REG-NAS enables 33-48% relative reduction of negative flips.

cs.CV↗

arXiv2022

Toward Understanding Privileged Features Distillation in Learning-to-Rank

In learning-to-rank problems, a privileged feature is one that is available during model training, but not available at test time. Such features naturally arise in merchandised recommendation systems; for instance, "user clicked this item" as a feature is predictive of "user purchased this item" in the offline data, but is clearly not available during online serving. Another source of privileged features is those that are too expensive to compute online but feasible to be added offline. Privileged features distillation (PFD) refers to a natural idea: train a "teacher" model using all features (including privileged ones) and then use it to train a "student" model that does not use the privileged features. In this paper, we first study PFD empirically on three public ranking datasets and an industrial-scale ranking problem derived from Amazon's logs. We show that PFD outperforms several baselines (no-distillation, pretraining-finetuning, self-distillation, and generalized distillation) on all these datasets. Next, we analyze why and when PFD performs well via both empirical ablation studies and theoretical analysis for linear models. Both investigations uncover an interesting non-monotone behavior: as the predictive power of a privileged feature increases, the performance of the resulting student model initially increases but then decreases. We show the reason for the later decreasing performance is that a very predictive privileged teacher produces predictions with high variance, which lead to high variance student estimates and inferior testing performance.

cs.LG↗