arXiv subjects

Yuan Cao

Publications and source records attributed to Yuan Cao.

At least 91 records · Page 5Linked to original sources

arXiv2024

Understanding the Benefits of SimCLR Pre-Training in Two-Layer Convolutional Neural Networks

SimCLR is one of the most popular contrastive learning methods for vision tasks. It pre-trains deep neural networks based on a large amount of unlabeled data by teaching the model to distinguish between positive and negative pairs of augmented images. It is believed that SimCLR can pre-train a deep neural network to learn efficient representations that can lead to a better performance of future supervised fine-tuning. Despite its effectiveness, our theoretical understanding of the underlying mechanisms of SimCLR is still limited. In this paper, we theoretically introduce a case study of the SimCLR method. Specifically, we consider training a two-layer convolutional neural network (CNN) to learn a toy image data model. We show that, under certain conditions on the number of labeled data, SimCLR pre-training combined with supervised fine-tuning achieves almost optimal test loss. Notably, the label complexity for SimCLR pre-training is far less demanding compared to direct training on supervised data. Our analysis sheds light on the benefits of SimCLR in learning with fewer labels.

cs.LG↗

arXiv2024

Improving Fast Adversarial Training via Self-Knowledge Guidance

Adversarial training has achieved remarkable advancements in defending against adversarial attacks. Among them, fast adversarial training (FAT) is gaining attention for its ability to achieve competitive robustness with fewer computing resources. Existing FAT methods typically employ a uniform strategy that optimizes all training data equally without considering the influence of different examples, which leads to an imbalanced optimization. However, this imbalance remains unexplored in the field of FAT. In this paper, we conduct a comprehensive study of the imbalance issue in FAT and observe an obvious class disparity regarding their performances. This disparity could be embodied from a perspective of alignment between clean and robust accuracy. Based on the analysis, we mainly attribute the observed misalignment and disparity to the imbalanced optimization in FAT, which motivates us to optimize different training data adaptively to enhance robustness. Specifically, we take disparity and misalignment into consideration. First, we introduce self-knowledge guided regularization, which assigns differentiated regularization weights to each class based on its training state, alleviating class disparity. Additionally, we propose self-knowledge guided label relaxation, which adjusts label relaxation according to the training accuracy, alleviating the misalignment and improving robustness. By combining these methods, we formulate the Self-Knowledge Guided FAT (SKG-FAT), leveraging naturally generated knowledge during training to enhance the adversarial robustness without compromising training efficiency. Extensive experiments on four standard datasets demonstrate that the SKG-FAT improves the robustness and preserves competitive clean accuracy, outperforming the state-of-the-art methods.

cs.CV↗

arXiv2024

Microsatellite-based real-time quantum key distribution

A quantum network provides an infrastructure connecting quantum devices with revolutionary computing, sensing, and communication capabilities. As the best-known application of a quantum network, quantum key distribution (QKD) shares secure keys guaranteed by the laws of quantum mechanics. A quantum satellite constellation offers a solution to facilitate the quantum network on a global scale. The Micius satellite has verified the feasibility of satellite quantum communications, however, scaling up quantum satellite constellations is challenging, requiring small lightweight satellites, portable ground stations and real-time secure key exchange. Here we tackle these challenges and report the development of a quantum microsatellite capable of performing space-to-ground QKD using portable ground stations. The quantum microsatellite features a payload weighing approximately 23 kg, while the portable ground station weighs about 100 kg. These weights represent reductions by more than an order and two orders of magnitude, respectively, compared to the Micius satellite. Additionally, we multiplex bidirectional satellite-ground optical communication with quantum communication, enabling key distillation and secure communication in real-time. Using the microsatellite and the portable ground stations, we demonstrate satellite-based QKD with multiple ground stations and achieve the sharing of up to 0.59 million bits of secure keys during a single satellite pass. The compact quantum payload can be readily assembled on existing space stations or small satellites, paving the way for a satellite-constellation-based quantum and classical network for widespread real-life applications.

quant-ph↗

arXiv2024

Characterization of Intensity Correlation via Single-photon Detection in Quantum Key Distribution

One of the most significant vulnerabilities in the source unit of quantum key distribution (QKD) is the correlation between quantum states after modulation, which shall be characterized and evaluated for its practical security performance. In this work, we propose a methodology to characterize the intensity correlation according to the single-photon detection results in the measurement unit without modifying the configuration of the QKD system. In contrast to the previous research that employs extra classical optical detector to measure the correlation, our method can directly analyse the detection data generated during the raw key exchange, enabling to characterize the feature of correlation in real-time system operation. The basic method is applied to a BB84 QKD system and the characterized correlation decreases the secure key rate shown by the security proof. Furthermore, the method is extended and applied to characterize the correlation from the result of Bell-state measurement, which demonstrates its applicability to a running full-scheme MDI QKD system. This study provides an approach for standard certification of a QKD system.

quant-ph↗

arXiv2024

Single-photon interference over 8.4 km urban atmosphere: towards testing quantum effects in curved spacetime with photons

The emergence of quantum mechanics and general relativity has transformed our understanding of the natural world significantly. However, integrating these two theories presents immense challenges, and their interplay remains untested. Recent theoretical studies suggest that the single-photon interference covering huge space can effectively probe the interface between quantum mechanics and general relativity. We developed an alternative design using unbalanced Michelson interferometers to address this and validated its feasibility over an 8.4 km free-space channel. Using a high-brightness single-photon source based on quantum dots, we demonstrated single-photon interference along this long-distance baseline. We achieved a phase measurement precision of 16.2 mrad, which satisfied the measurement requirements for a gravitational redshift at the geosynchronous orbit by five times the standard deviation. Our results confirm the feasibility of the single-photon version of the Colella-Overhauser-Werner experiment for testing the quantum effects in curved spacetime.

quant-ph↗

arXiv2024

Realization of Topology-controlled Photonic Cavities in a Valley Photonic Crystal

We report an experimental realization of a new type of topology-controlled photonic cavities in valley photonic crystals by adopting judiciously oriented mirrors to localize the valley-polarized edge states along their propagation path. By using microwave frequency- and time-domain measurements, we directly observe the strong confinement of electromagnetic energy at the mirror surface due to the extended time delay required for the valley index flipping. Moreover, we experimentally demonstrate that both the degree of energy localization and quality factors of the topology-controlled photonic cavities are determined by the valley-flipping time which is controlled by the topology of the mirror. These results extend and complement the current design paradigm of topological photonic cavities.

physics.optics↗

arXiv2024

IG Captioner: Information Gain Captioners are Strong Zero-shot Classifiers

Generative training has been demonstrated to be powerful for building visual-language models. However, on zero-shot discriminative benchmarks, there is still a performance gap between models trained with generative and discriminative objectives. In this paper, we aim to narrow this gap by improving the efficacy of generative training on classification tasks, without any finetuning processes or additional modules. Specifically, we focus on narrowing the gap between the generative captioner and the CLIP classifier. We begin by analysing the predictions made by the captioner and classifier and observe that the caption generation inherits the distribution bias from the language model trained with pure text modality, making it less grounded on the visual signal. To tackle this problem, we redesign the scoring objective for the captioner to alleviate the distributional bias and focus on measuring the gain of information brought by the visual inputs. We further design a generative training objective to match the evaluation objective. We name our model trained and evaluated from the novel procedures as Information Gain (IG) captioner. We pretrain the models on the public Laion-5B dataset and perform a series of discriminative evaluations. For the zero-shot classification on ImageNet, IG captioner achieves $> 18\%$ improvements over the standard captioner, achieving comparable performances with the CLIP classifier. IG captioner also demonstrated strong performance on zero-shot image-text retrieval tasks on MSCOCO and Flickr30K. We hope this paper inspires further research towards unifying generative and discriminative training procedures for visual-language models.

cs.CV↗

arXiv2024

On the Convergence of Adaptive Gradient Methods for Nonconvex Optimization

Adaptive gradient methods are workhorses in deep learning. However, the convergence guarantees of adaptive gradient methods for nonconvex optimization have not been thoroughly studied. In this paper, we provide a fine-grained convergence analysis for a general class of adaptive gradient methods including AMSGrad, RMSProp and AdaGrad. For smooth nonconvex functions, we prove that adaptive gradient methods in expectation converge to a first-order stationary point. Our convergence rate is better than existing results for adaptive gradient methods in terms of dimension. In addition, we also prove high probability bounds on the convergence rates of AMSGrad, RMSProp as well as AdaGrad, which have not been established before. Our analyses shed light on better understanding the mechanism behind adaptive gradient methods in optimizing nonconvex objectives.

cs.LG↗

arXiv2024

The Implicit Bias of Adam on Separable Data

Adam has become one of the most favored optimizers in deep learning problems. Despite its success in practice, numerous mysteries persist regarding its theoretical understanding. In this paper, we study the implicit bias of Adam in linear logistic regression. Specifically, we show that when the training data are linearly separable, Adam converges towards a linear classifier that achieves the maximum $\ell_\infty$-margin. Notably, for a general class of diminishing learning rates, this convergence occurs within polynomial time. Our result shed light on the difference between Adam and (stochastic) gradient descent from a theoretical perspective.

stat.ML↗

arXiv2024

On-chip multi-degree-of-freedom control of two-dimensional quantum and nonlinear materials

Two-dimensional materials (2DM) and their derived heterostructures have electrical and optical properties that are widely tunable via several approaches, most notably electrostatic gating and interfacial engineering such as twisting. While electrostatic gating is simple and has been ubiquitously employed on 2DM, being able to tailor the interfacial properties in a similar real-time manner represents the next leap in our ability to modulate the underlying physics and build exotic devices with 2DM. However, all existing approaches rely on external machinery such as scanning microscopes, which often limit their scope of applications, and there is currently no means of tuning a 2D interface that has the same accessibility and scalability as electrostatic gating. Here, we demonstrate the first on-chip platform designed for 2D materials with in situ tunable interfacial properties, utilizing a microelectromechanical system (MEMS). Each compact, cost-effective, and versatile device is a standalone micromachine that allows voltage-controlled approaching, twisting, and pressurizing of 2DM with high accuracy. As a demonstration, we engineer synthetic topological singularities, known as merons, in the nonlinear optical susceptibility of twisted hexagonal boron nitride (h-BN), via simultaneous control of twist angle and interlayer separation. The chirality of the resulting moire pattern further induces a strong circular dichroism in the second-harmonic generation. A potential application of this topological nonlinear susceptibility is to create integrated classical and quantum light sources that have widely and real-time tunable polarization. Our invention pushes the boundary of available technologies for manipulating low-dimensional quantum materials, which in turn opens up the gateway for designing future hybrid 2D-3D devices for condensed-matter physics, quantum optics, and beyond.

cond-mat.mes-hall↗

arXiv2024

Experimental single-photon quantum key distribution surpassing the fundamental coherent-state rate limit

Single-photon sources are essential for quantum networks, enabling applications ranging from quantum key distribution (QKD) to the burgeoning quantum internet. Despite the remarkable advancements, the current reliance of QKD on attenuated coherent (laser) light sources has imposed a fundamental limit on the secret key rate (SKR). This constraint is primarily attributable to the scarcity of single-photon components within coherent light, confined by an inherent upper bound of 1/e. Here, we report high-rate QKD using a high-efficiency single-photon source, enabling an SKR transcending the fundamental rate limit of coherent light. We developed an on-demand, bright semiconductor quantum-dot single-photon source with an efficiency of 0.71(2), exceeding the inherent bound of coherent light by approximately 2.87 dB. Implementing narrow-bandwidth filtering and random polarization modulation, we conducted a field QKD trial over a 14.6(1.1)-dB-loss free-space urban channel, achieving an SKR of 0.00108 bits per pulse. This surpasses the practical limit of coherent-light-based QKD by 2.53 dB. Our findings conclusively demonstrate the superior performance of nanotechnology-based single-photon sources over coherent light for QKD applications, marking a pivotal stride towards the realization of a global quantum internet.

quant-ph↗

arXiv2024

SIAM: A Simple Alternating Mixer for Video Prediction

Video prediction, predicting future frames from the previous ones, has broad applications such as autonomous driving and weather forecasting. Existing state-of-the-art methods typically focus on extracting either spatial, temporal, or spatiotemporal features from videos. Different feature focuses, resulting from different network architectures, may make the resultant models excel at some video prediction tasks but perform poorly on others. Towards a more generic video prediction solution, we explicitly model these features in a unified encoder-decoder framework and propose a novel simple alternating Mixer (SIAM). The novelty of SIAM lies in the design of dimension alternating mixing (DaMi) blocks, which can model spatial, temporal, and spatiotemporal features through alternating the dimensions of the feature maps. Extensive experimental results demonstrate the superior performance of the proposed SIAM on four benchmark video datasets covering both synthetic and real-world scenarios.

cs.CV↗

arXiv2024

Understanding the performance gap between online and offline alignment algorithms

Reinforcement learning from human feedback (RLHF) is the canonical framework for large language model alignment. However, rising popularity in offline alignment algorithms challenge the need for on-policy sampling in RLHF. Within the context of reward over-optimization, we start with an opening set of experiments that demonstrate the clear advantage of online methods over offline methods. This prompts us to investigate the causes to the performance discrepancy through a series of carefully designed experimental ablations. We show empirically that hypotheses such as offline data coverage and data quality by itself cannot convincingly explain the performance difference. We also find that while offline algorithms train policy to become good at pairwise classification, it is worse at generations; in the meantime the policies trained by online algorithms are good at generations while worse at pairwise classification. This hints at a unique interplay between discriminative and generative capabilities, which is greatly impacted by the sampling process. Lastly, we observe that the performance discrepancy persists for both contrastive and non-contrastive loss functions, and appears not to be addressed by simply scaling up policy networks. Taken together, our study sheds light on the pivotal role of on-policy sampling in AI alignment, and hints at certain fundamental challenges of offline alignment algorithms.

cs.LG↗

arXiv2024

Can Public Large Language Models Help Private Cross-device Federated Learning?

We study (differentially) private federated learning (FL) of language models. The language models in cross-device FL are relatively small, which can be trained with meaningful formal user-level differential privacy (DP) guarantees when massive parallelism in training is enabled by the participation of a moderate size of users. Recently, public data has been used to improve privacy-utility trade-offs for both large and small language models. In this work, we provide a systematic study of using large-scale public data and LLMs to help differentially private training of on-device FL models, and further improve the privacy-utility tradeoff by techniques of distillation. Moreover, we propose a novel distribution matching algorithm with theoretical grounding to sample public data close to private data distribution, which significantly improves the sample efficiency of (pre-)training on public data. The proposed method is efficient and effective for training private models by taking advantage of public data, especially for customized on-device architectures that do not have ready-to-use pre-trained models.

cs.LG↗

arXiv2024

Retrieval Augmented End-to-End Spoken Dialog Models

We recently developed SLM, a joint speech and language model, which fuses a pretrained foundational speech model and a large language model (LLM), while preserving the in-context learning capability intrinsic to the pretrained LLM. In this paper, we apply SLM to speech dialog applications where the dialog states are inferred directly from the audio signal. Task-oriented dialogs often contain domain-specific entities, i.e., restaurants, hotels, train stations, and city names, which are difficult to recognize, however, critical for the downstream applications. Inspired by the RAG (retrieval-augmented generation) paradigm, we propose a retrieval augmented SLM (ReSLM) that overcomes this weakness. We first train a speech retriever to retrieve text entities mentioned in the audio. The retrieved entities are then added as text inputs to the underlying SLM to bias model predictions. We evaluated ReSLM on speech MultiWoz task (DSTC-11 challenge), and found that this retrieval augmentation boosts model performance, achieving joint goal accuracy (38.6% vs 32.7%), slot error rate (20.6% vs 24.8%) and ASR word error rate (5.5% vs 6.7%). While demonstrated on dialog state tracking, our approach is broadly applicable to other speech tasks requiring contextual information or domain-specific entities, such as contextual ASR with biasing capability.

cs.CL↗

arXiv2024

Can overfitted deep neural networks in adversarial training generalize? -- An approximation viewpoint

Adversarial training is a widely used method to improve the robustness of deep neural networks (DNNs) over adversarial perturbations. However, it is empirically observed that adversarial training on over-parameterized networks often suffers from the \textit{robust overfitting}: it can achieve almost zero adversarial training error while the robust generalization performance is not promising. In this paper, we provide a theoretical understanding of the question of whether overfitted DNNs in adversarial training can generalize from an approximation viewpoint. Specifically, our main results are summarized into three folds: i) For classification, we prove by construction the existence of infinitely many adversarial training classifiers on over-parameterized DNNs that obtain arbitrarily small adversarial training error (overfitting), whereas achieving good robust generalization error under certain conditions concerning the data quality, well separated, and perturbation level. ii) Linear over-parameterization (meaning that the number of parameters is only slightly larger than the sample size) is enough to ensure such existence if the target function is smooth enough. iii) For regression, our results demonstrate that there also exist infinitely many overfitted DNNs with linear over-parameterization in adversarial training that can achieve almost optimal rates of convergence for the standard generalization error. Overall, our analysis points out that robust overfitting can be avoided but the required model capacity will depend on the smoothness of the target function, while a robust generalization gap is inevitable. We hope our analysis will give a better understanding of the mathematical foundations of robustness in DNNs from an approximation view.

stat.ML↗

arXiv2024

Generative AI-enabled Quantum Computing Networks and Intelligent Resource Allocation

Quantum computing networks enable scalable collaboration and secure information exchange among multiple classical and quantum computing nodes while executing large-scale generative AI computation tasks and advanced quantum algorithms. Quantum computing networks overcome limitations such as the number of qubits and coherence time of entangled pairs and offer advantages for generative AI infrastructure, including enhanced noise reduction through distributed processing and improved scalability by connecting multiple quantum devices. However, efficient resource allocation in quantum computing networks is a critical challenge due to factors including qubit variability and network complexity. In this article, we propose an intelligent resource allocation framework for quantum computing networks to improve network scalability with minimized resource costs. To achieve scalability in quantum computing networks, we formulate the resource allocation problem as stochastic programming, accounting for the uncertain fidelities of qubits and entangled pairs. Furthermore, we introduce state-of-the-art reinforcement learning (RL) algorithms, from generative learning to quantum machine learning for optimal quantum resource allocation to resolve the proposed stochastic resource allocation problem efficiently. Finally, we optimize the resource allocation in heterogeneous quantum computing networks supporting quantum generative learning applications and propose a multi-agent RL-based algorithm to learn the optimal resource allocation policies without prior knowledge.

cs.NI↗

arXiv2023

On-Chip Multidimensional Dynamic Control of Twisted Moiré Photonic Crystal for Smart Sensing and Imaging

Reconfigurable optics, optical systems that have a dynamically tunable configuration, are emerging as a new frontier in photonics research. Recently, twisted moiré photonic crystal has become a competitive candidate for implementing reconfigurable optics because of its high degree of tunability. However, despite its great potential as versatile optics components, simultaneous and dynamic modulation of multiple degrees of freedom in twisted moiré photonic crystal has remained out of reach, severely limiting its area of application. In this paper, we present a MEMS-integrated twisted moiré photonic crystal sensor that offers precise control over the interlayer gap and twist angle between two photonic crystal layers, and demonstrate an active twisted moiré photonic crystal-based optical sensor that can simultaneously resolve wavelength and polarization. Leveraging twist- and gap-tuned resonance modes, we achieve high-accuracy spectropolarimetric reconstruction of light using an adaptive sensing algorithm over a broad operational bandwidth in the telecom range and full Poincaré sphere. Our research showcases the remarkable capabilities of multidimensional control over emergent degrees of freedom in reconfigurable nanophotonics platforms and establishes a scalable pathway towards creating comprehensive flat-optics devices suitable for versatile light manipulation and information processing tasks.

physics.optics↗