arXiv subjects

Qian Xu

Publications and source records attributed to Qian Xu.

At least 73 records · Page 4Linked to original sources

arXiv2022

An Efficient Industrial Federated Learning Framework for AIoT: A Face Recognition Application

Recently, the artificial intelligence of things (AIoT) has been gaining increasing attention, with an intriguing vision of providing highly intelligent services through the network connection of things, leading to an advanced AI-driven ecology. However, recent regulatory restrictions on data privacy preclude uploading sensitive local data to data centers and utilizing them in a centralized approach. Directly applying federated learning algorithms in this scenario could hardly meet the industrial requirements of both efficiency and accuracy. Therefore, we propose an efficient industrial federated learning framework for AIoT in terms of a face recognition application. Specifically, we propose to utilize the concept of transfer learning to speed up federated training on devices and further present a novel design of a private projector that helps protect shared gradients without incurring additional memory consumption or computational cost. Empirical studies on a private Asian face dataset show that our approach can achieve high recognition accuracy in only 20 communication rounds, demonstrating its effectiveness in prediction and its efficiency in training.

cs.CV↗

arXiv2022

Uncertainty-based Network for Few-shot Image Classification

The transductive inference is an effective technique in the few-shot learning task, where query sets update prototypes to improve themselves. However, these methods optimize the model by considering only the classification scores of the query instances as confidence while ignoring the uncertainty of these classification scores. In this paper, we propose a novel method called Uncertainty-Based Network, which models the uncertainty of classification results with the help of mutual information. Specifically, we first data augment and classify the query instance and calculate the mutual information of these classification scores. Then, mutual information is used as uncertainty to assign weights to classification scores, and the iterative update strategy based on classification scores and uncertainties assigns the optimal weights to query instances in prototype optimization. Extensive results on four benchmarks show that Uncertainty-Based Network achieves comparable performance in classification accuracy compared to state-of-the-art method.

cs.CV↗

arXiv2022

Distributed quantum error correction for chip-level catastrophic errors

Quantum error correction holds the key to scaling up quantum computers. Cosmic ray events severely impact the operation of a quantum computer by causing chip-level catastrophic errors, essentially erasing the information encoded in a chip. Here, we present a distributed error correction scheme to combat the devastating effect of such events by introducing an additional layer of quantum erasure error correcting code across separate chips. We show that our scheme is fault tolerant against chip-level catastrophic errors and discuss its experimental implementation using superconducting qubits with microwave links. Our analysis shows that in state-of-the-art experiments, it is possible to suppress the rate of these errors from 1 per 10 seconds to less than 1 per month.

quant-ph↗

arXiv2022

Tailored XZZX codes for biased noise

Quantum error correction (QEC) for generic errors is challenging due to the demanding threshold and resource requirements. Interestingly, when physical noise is biased, we can tailor our QEC schemes to the noise to improve performance. Here we study a family of codes having XZZX-type stabilizer generators, including a set of cyclic codes generalized from the five-qubit code and a set of topological codes that we call generalized toric codes (GTCs). We show that these XZZX codes are highly qubit efficient if tailored to biased noise. To characterize the code performance, we use the notion of effective distance, which generalizes code distance to the case of biased noise and constitutes a proxy for the logical failure rate. We find that the XZZX codes can achieve a favorable resource scaling by this metric under biased noise. We also show that the XZZX codes have remarkably high thresholds that reach what is achievable by random codes, and furthermore they can be efficiently decoded using matching decoders. Finally, by adding only one flag qubit, the XZZX codes can realize fault-tolerant QEC while preserving their large effective distance. In combination, our results show that tailored XZZX codes give a resource-efficient scheme for fault-tolerant QEC against biased noise.

quant-ph↗

arXiv2022

Stabilizing a Bosonic Qubit using Colored Dissipation

Protected qubits such as the 0-$π$ qubit, and bosonic qubits including cat qubits and GKP qubits offer advantages for fault-tolerance. Some of these protected qubits (e.g., 0-$π$ qubit and Kerr cat qubit) are stabilized by Hamiltonians which have (near-)degenerate ground state manifolds with large energy-gaps to the excited state manifolds. Without dissipative stabilization mechanisms the performance of such energy-gap-protected qubits can be limited by leakage to excited states. Here, we propose a scheme for dissipatively stabilizing an energy-gap-protected qubit using colored (i.e., frequency-selective) dissipation without inducing errors in the ground state manifold. Concretely we apply our colored dissipation technique to Kerr cat qubits and propose colored Kerr cat qubits which are protected by an engineered colored single-photon loss. When applied to the Kerr cat qubits our scheme significantly suppresses leakage-induced bit-flip errors (which we show are a limiting error mechanism) while only using linear interactions. Beyond the benefits to the Kerr cat qubit we also show that our frequency-selective loss technique can be applied to a broader class of protected qubits.

quant-ph↗

arXiv2021

Band-Notched Frequency-Selective Absorber with Polarization Rotation Function

This paper presents the theoretical analysis, design, simulations, and experimental verification of a novel band-notched frequency-selective absorber (BFSA) with polarization rotation function and absorption out-of-band. The BFSA consists of multiple layers including one lossy layer, one polarization-rotary layer and an air layer. The lossy layer is loaded with lumped resistances for obtaining a wide absorption band. A new four-port equivalent circuit model (ECM) with lossy layer is developed for providing theoretical analysis. The BFSA realizes -0.41dB cross-polarization reflection at 4.5GHz. In the upper and lower bands, the BFSA realizes a broad absorption function from 2.2GHz to 4.1GHz and 5.03GHz to 6.5GHz. The fractional bandwidth of co-polarization reflection is 98.8% with 10dB reflection reduction. The full-wave simulation, ECM, and experimental measurements are conducted to validate the polarization conversion of the band-notched absorbers, and good agreement between theory and measurement results is observed.

physics.app-ph↗

arXiv2021

Graph Random Neural Network for Semi-Supervised Learning on Graphs

We study the problem of semi-supervised learning on graphs, for which graph neural networks (GNNs) have been extensively explored. However, most existing GNNs inherently suffer from the limitations of over-smoothing, non-robustness, and weak-generalization when labeled nodes are scarce. In this paper, we propose a simple yet effective framework -- GRAPH RANDOM NEURAL NETWORKS (GRAND) -- to address these issues. In GRAND, we first design a random propagation strategy to perform graph data augmentation. Then we leverage consistency regularization to optimize the prediction consistency of unlabeled nodes across different data augmentations. Extensive experiments on graph benchmark datasets suggest that GRAND significantly outperforms state-of-the-art GNN baselines on semi-supervised node classification. Finally, we show that GRAND mitigates the issues of over-smoothing and non-robustness, exhibiting better generalization behavior than existing GNNs. The source code of GRAND is publicly available at https://github.com/Grand20/grand.

cs.LG↗

arXiv2021

Learning Class-level Prototypes for Few-shot Learning

Few-shot learning aims to recognize new categories using very few labeled samples. Although few-shot learning has witnessed promising development in recent years, most existing methods adopt an average operation to calculate prototypes, thus limited by the outlier samples. In this work, we propose a simple yet effective framework for few-shot classification, which can learn to generate preferable prototypes from few support data, with the help of an episodic prototype generator module. The generated prototype is meant to be close to a certain \textit{\targetproto{}} and is less influenced by outlier samples. Extensive experiments demonstrate the effectiveness of this module, and our approach gets a significant raise over baseline models, and get a competitive result compared to previous methods on \textit{mini}ImageNet, \textit{tiered}ImageNet, and cross-domain (\textit{mini}ImageNet $\rightarrow$ CUB-200-2011) datasets.

cs.CV↗

arXiv2021

Deep Generative Learning via Schrödinger Bridge

We propose to learn a generative model via entropy interpolation with a Schrödinger Bridge. The generative learning task can be formulated as interpolating between a reference distribution and a target distribution based on the Kullback-Leibler divergence. At the population level, this entropy interpolation is characterized via an SDE on $[0,1]$ with a time-varying drift term. At the sample level, we derive our Schrödinger Bridge algorithm by plugging the drift term estimated by a deep score estimator and a deep density ratio estimator into the Euler-Maruyama method. Under some mild smoothness assumptions of the target distribution, we prove the consistency of both the score estimator and the density ratio estimator, and then establish the consistency of the proposed Schrödinger Bridge approach. Our theoretical results guarantee that the distribution learned by our approach converges to the target distribution. Experimental results on multimodal synthetic data and benchmark data support our theoretical findings and indicate that the generative model via Schrödinger Bridge is comparable with state-of-the-art GANs, suggesting a new formulation of generative learning. We demonstrate its usefulness in image interpolation and image inpainting.

cs.LG↗

arXiv2021

Quantum repeaters based on concatenated bosonic and discrete-variable quantum codes

We propose an architecture of quantum-error-correction-based quantum repeaters that combines techniques used in discrete- and continuous-variable quantum information. Specifically, we propose to encode the transmitted qubits in a concatenated code consisting of two levels. On the first level we use a continuous-variable GKP code encoding the qubit in a single bosonic mode. On the second level we use a small discrete-variable code. Such an architecture has two important features. Firstly, errors on each of the two levels are corrected in repeaters of two different types. This enables for achieving performance needed in practical scenarios with a reduced cost with respect to an architecture for which all repeaters are the same. Secondly, the use of continuous-variable GKP code on the lower level generates additional analog information which enhances the error-correcting capabilities of the second-level code such that long-distance communication becomes possible with encodings consisting of only four or seven optical modes.

quant-ph↗

arXiv2021

Engineering fast bias-preserving gates on stabilized cat qubits

Stabilized cat codes can provide a biased noise channel with a set of bias-preserving (BP) gates, which can significantly reduce the resource overhead for fault-tolerant quantum computing. All existing schemes of BP gates, however, require adiabatic quantum evolution, with performance limited by excitation loss and non-adiabatic errors during the adiabatic gates. In this work, we apply a derivative-based leakage suppression technique to overcome non-adiabatic errors, so that we can implement fast BP gates on Kerr-cat qubits with improved gate fidelity while maintaining high noise bias. When applied to concatenated quantum error correction, the fast BP gates can not only improve the logical error rate but also reduce resource overhead, which enables more efficient implementation of fault-tolerant quantum computing.

quant-ph↗

arXiv2020

A Learning-based Discretionary Lane-Change Decision-Making Model with Driving Style Awareness

Discretionary lane change (DLC) is a basic but complex maneuver in driving, which aims at reaching a faster speed or better driving conditions, e.g., further line of sight or better ride quality. Although many DLC decision-making models have been studied in traffic engineering and autonomous driving, the impact of human factors, which is an integral part of current and future traffic flow, is largely ignored in the existing literature. In autonomous driving, the ignorance of human factors of surrounding vehicles will lead to poor interaction between the ego vehicle and the surrounding vehicles, thus, a high risk of accidents. The human factors are also a crucial part to simulate a human-like traffic flow in the traffic engineering area. In this paper, we integrate the human factors that are represented by driving styles to design a new DLC decision-making model. Specifically, our proposed model takes not only the contextual traffic information but also the driving styles of surrounding vehicles into consideration and makes lane-change/keep decisions. Moreover, the model can imitate human drivers' decision-making maneuvers to the greatest extent by learning the driving style of the ego vehicle. Our evaluation results show that the proposed model almost follows the human decision-making maneuvers, which can achieve 98.66% prediction accuracy with respect to human drivers' decisions against the ground truth. Besides, the lane-change impact analysis results demonstrate that our model even performs better than human drivers in terms of improving the safety and speed of traffic.

cs.RO↗

arXiv2020

Quantum wave-particle superposition in a delayed-choice experiment

Wave-particle duality epitomizes the counterintuitive character of quantum physics. A striking illustration is the quantum delay-choice experiment, which is based on Wheeler's classic delayed-choice gedanken experiment, but with the addition of a quantum-controlled device enabling wave-to-particle transitions. Here we realize a quantum delayed-choice experiment in which we control the wave and the particle states of photons and in particular the phase between them, thus directly establishing the created quantum nature of the wave-particle. We generate three-photon entangled states and inject one photon into a Mach--Zehnder interferometer embedded in a 187-m-long two-photon Hong-Ou-Mandel interferometer. The third photon is sent 141m away from the interferometers and remotely prepares a two-photon quantum gate according to independent active choices under Einstein locality conditions. We have realized transitions between wave and particle states in both classical and quantum scenarios, and therefore tests of the complementarity principle that go fundamentally beyond earlier implementations.

quant-ph↗

arXiv2020

MxPool: Multiplex Pooling for Hierarchical Graph Representation Learning

How to utilize deep learning methods for graph classification tasks has attracted considerable research attention in the past few years. Regarding graph classification tasks, the graphs to be classified may have various graph sizes (i.e., different number of nodes and edges) and have various graph properties (e.g., average node degree, diameter, and clustering coefficient). The diverse property of graphs has imposed significant challenges on existing graph learning techniques since diverse graphs have different best-fit hyperparameters. It is difficult to learn graph features from a set of diverse graphs by a unified graph neural network. This motivates us to use a multiplex structure in a diverse way and utilize a priori properties of graphs to guide the learning. In this paper, we propose MxPool, which concurrently uses multiple graph convolution/pooling networks to build a hierarchical learning structure for graph representation learning tasks. Our experiments on numerous graph classification benchmarks show that our MxPool has superiority over other state-of-the-art graph representation learning methods.

cs.LG↗

arXiv2020

Federated Deep Reinforcement Learning

In deep reinforcement learning, building policies of high-quality is challenging when the feature space of states is small and the training data is limited. Despite the success of previous transfer learning approaches in deep reinforcement learning, directly transferring data or models from an agent to another agent is often not allowed due to the privacy of data and/or models in many privacy-aware applications. In this paper, we propose a novel deep reinforcement learning framework to federatively build models of high-quality for agents with consideration of their privacies, namely Federated deep Reinforcement Learning (FedRL). To protect the privacy of data and models, we exploit Gausian differentials on the information shared with each other when updating their local models. In the experiment, we evaluate our FedRL framework in two diverse domains, Grid-world and Text2Action domains, by comparing to various baselines.

cs.LG↗

arXiv2019

L2RS: A Learning-to-Rescore Mechanism for Automatic Speech Recognition

Modern Automatic Speech Recognition (ASR) systems primarily rely on scores from an Acoustic Model (AM) and a Language Model (LM) to rescore the N-best lists. With the abundance of recent natural language processing advances, the information utilized by current ASR for evaluating the linguistic and semantic legitimacy of the N-best hypotheses is rather limited. In this paper, we propose a novel Learning-to-Rescore (L2RS) mechanism, which is specialized for utilizing a wide range of textual information from the state-of-the-art NLP models and automatically deciding their weights to rescore the N-best lists for ASR systems. Specifically, we incorporate features including BERT sentence embedding, topic vector, and perplexity scores produced by n-gram LM, topic modeling LM, BERT LM and RNNLM to train a rescoring model. We conduct extensive experiments based on a public dataset, and experimental results show that L2RS outperforms not only traditional rescoring methods but also its deep neural network counterparts by a substantial improvement of 20.67% in terms of NDCG@10. L2RS paves the way for developing more effective rescoring models for ASR.

cs.CL↗

arXiv2019

Rethinking Secure Precoding via Interference Exploitation: A Smart Eavesdropper Perspective

Based on the concept of constructive interference (CI), multiuser interference (MUI) has recently been shown to be beneficial for communication secrecy. A few CI-based secure precoding algorithms have been proposed that use both the channel state information (CSI) and knowledge of the instantaneous transmit symbols. In this paper, we examine the CI-based secure precoding problem with a focus on smart eavesdroppers that exploit statistical information gleaned from the precoded data for symbol detection. Moreover, the impact of correlation between the main and eavesdropper channels is taken into account. We first modify an existing CI-based preocding scheme to better utilize the destructive impact of the interference. Then, we point out the drawback of both the existing and the new modified CI-based precoders when faced with a smart eavesdropper. To address this deficiency, we provide a general principle for precoder design and then give two specific design examples. Finally, the scenario where the eavesdropper's CSI is unavailable is studied. Numerical results show that although our modified CI-based precoder can achieve a better energy-secrecy trade-off than the existing approach, both have a limited secrecy benefit. On the contrary, the precoders developed using the new CI-design principle can achieve a much improved trade-off and significantly degrade the eavesdropper's performance.

cs.IT↗

arXiv2019

A Fully-Automatic Framework for Parkinson's Disease Diagnosis by Multi-Modality Images

Background: Parkinson's disease (PD) is a prevalent long-term neurodegenerative disease. Though the diagnostic criteria of PD are relatively well defined, the current medical imaging diagnostic procedures are expertise-demanding, and thus call for a higher-integrated AI-based diagnostic algorithm. Methods: In this paper, we proposed an automatic, end-to-end, multi-modality diagnosis framework, including segmentation, registration, feature generation and machine learning, to process the information of the striatum for the diagnosis of PD. Multiple modalities, including T1- weighted MRI and 11C-CFT PET, were used in the proposed framework. The reliability of this framework was then validated on a dataset from the PET center of Huashan Hospital, as the dataset contains paired T1-MRI and CFT-PET images of 18 Normal (NL) subjects and 49 PD subjects. Results: We obtained an accuracy of 100% for the PD/NL classification task, besides, we conducted several comparative experiments to validate the diagnosis ability of our framework. Conclusion: Through experiment we illustrate that (1) automatic segmentation has the same classification effect as the manual segmentation, (2) the multi-modality images generates a better prediction than single modality images, and (3) volume feature is shown to be irrelevant to PD diagnosis.

q-bio.QM↗