arXiv subjects

Yuan Wang

Publications and source records attributed to Yuan Wang.

At least 145 records · Page 8Linked to original sources

arXiv2024

Advertiser Content Understanding via LLMs for Google Ads Safety

Ads Content Safety at Google requires classifying billions of ads for Google Ads content policies. Consistent and accurate policy enforcement is important for advertiser experience and user safety and it is a challenging problem, so there is a lot of value for improving it for advertisers and users. Inconsistent policy enforcement causes increased policy friction and poor experience with good advertisers, and bad advertisers exploit the inconsistency by creating multiple similar ads in the hope that some will get through our defenses. This study proposes a method to understand advertiser's intent for content policy violations, using Large Language Models (LLMs). We focus on identifying good advertisers to reduce content over-flagging and improve advertiser experience, though the approach can easily be extended to classify bad advertisers too. We generate advertiser's content profile based on multiple signals from their ads, domains, targeting info, etc. We then use LLMs to classify the advertiser content profile, along with relying on any knowledge the LLM has of the advertiser, their products or brand, to understand whether they are likely to violate a certain policy or not. After minimal prompt tuning our method was able to reach 95\% accuracy on a small test set.

cs.IR↗

arXiv2024

Localization and Expansion: A Decoupled Framework for Point Cloud Few-shot Semantic Segmentation

Point cloud few-shot semantic segmentation (PC-FSS) aims to segment targets of novel categories in a given query point cloud with only a few annotated support samples. The current top-performing prototypical learning methods employ prototypes originating from support samples to direct the classification of query points. However, the inherent fragility of point-level matching and the prevalent intra-class diversity pose great challenges to this cross-instance matching paradigm, leading to erroneous background activations or incomplete target excavation. In this work, we propose a simple yet effective framework in the spirit of Decoupled Localization and Expansion (DLE). The proposed DLE, including a structural localization module (SLM) and a self-expansion module (SEM), enjoys several merits. First, structural information is injected into the matching process through the agent-level correlation in SLM, and the confident target region can thus be precisely located. Second, more reliable intra-object similarity is harnessed in SEM to derive the complete target, and the conservative expansion strategy is introduced to reasonably constrain the expansion. Extensive experiments on two challenging benchmarks under different settings demonstrate that DLE outperforms previous state-of-the-art approaches by large margins.

cs.CV↗

arXiv2024

Photonic time-delayed reservoir computing based on lithium niobate microring resonators

On-chip micro-ring resonators (MRRs) have been proposed for constructing delay reservoir computing (RC) systems, offering a highly scalable, high-density computational architecture that is easy to manufacture. However, most proposed RC schemes have utilized passive integrated optical components based on silicon-on-insulator (SOI), and RC systems based on lithium niobate on insulator (LNOI) have not yet been reported. The nonlinear optical effects exhibited by lithium niobate microphotonic devices introduce new possibilities for RC design. In this work, we design an RC scheme based on a series-coupled MRR array, leveraging the unique interplay between thermo-optic nonlinearity and photorefractive effects in lithium niobate. We first demonstrate the existence of three regions defined by wavelength detuning between the primary LNOI micro-ring resonator and the coupled micro-ring array, where one region achieves an optimal balance between nonlinearity and high memory capacity at extremely low input energy, leading to superior computational performance. We then discuss in detail the impact of each ring's nonlinearity and the system's symbol duration on performance. Finally, we design a wavelength-division multiplexing (WDM) based multi-task parallel computing scheme, showing that the computational performance for multiple tasks matches that of single-task computations.

physics.optics↗

arXiv2024

AdapMoE: Adaptive Sensitivity-based Expert Gating and Management for Efficient MoE Inference

Mixture-of-Experts (MoE) models are designed to enhance the efficiency of large language models (LLMs) without proportionally increasing the computational demands. However, their deployment on edge devices still faces significant challenges due to high on-demand loading overheads from managing sparsely activated experts. This paper introduces AdapMoE, an algorithm-system co-design framework for efficient MoE inference. AdapMoE features adaptive expert gating and management to reduce the on-demand loading overheads. We observe the heterogeneity of experts loading across layers and tokens, based on which we propose a sensitivity-based strategy to adjust the number of activated experts dynamically. Meanwhile, we also integrate advanced prefetching and cache management techniques to further reduce the loading latency. Through comprehensive evaluations on various platforms, we demonstrate AdapMoE consistently outperforms existing techniques, reducing the average number of activated experts by 25% and achieving a 1.35x speedup without accuracy degradation. Code is available at: https://github.com/PKU-SEC-Lab/AdapMoE.

cs.LG↗

arXiv2024

Occupancy-driven Zeeman suppression and inversion in trapped polariton condensates

We study the magneto-photoluminescence of an optically trapped exciton-polariton condensate in a planar semiconductor microcavity with multiple In0.08Ga0.92As quantum wells. Extremely high condensate coherence time and continuous control over the polariton confinement are among the advantages provided by optical trapping. This allows us to resolve magnetically induced μeV fine-energy shifts in the condensate and identify unusual dynamical regions in its parameter space. We observe polariton Zeeman splitting and, in small traps with tight confinement, demonstrate its full parametric screening when the condensate density exceeds a critical value, reminiscent of the spin-Meissner effect. For larger optical traps, we observe a complete inversion in the Zeeman splitting as a function of power, underlining the importance of condensate confinement and interactions with its background reservoir excitons.

cond-mat.mes-hall↗

arXiv2024

HG-PIPE: Vision Transformer Acceleration with Hybrid-Grained Pipeline

Vision Transformer (ViT) acceleration with field programmable gate array (FPGA) is promising but challenging. Existing FPGA-based ViT accelerators mainly rely on temporal architectures, which process different operators by reusing the same hardware blocks and suffer from extensive memory access overhead. Pipelined architectures, either coarse-grained or fine-grained, unroll the ViT computation spatially for memory access efficiency. However, they usually suffer from significant hardware resource constraints and pipeline bubbles induced by the global computation dependency of ViT. In this paper, we introduce HG-PIPE, a pipelined FPGA accelerator for high-throughput and low-latency ViT processing. HG-PIPE features a hybrid-grained pipeline architecture to reduce on-chip buffer cost and couples the computation dataflow and parallelism design to eliminate the pipeline bubbles. HG-PIPE further introduces careful approximations to implement both linear and non-linear operators with abundant Lookup Tables (LUTs), thus alleviating resource constraints. On a ZCU102 FPGA, HG-PIPE achieves 2.78 times better throughput and 2.52 times better resource efficiency than the prior-art accelerators, e.g., AutoViTAcc. With a VCK190 FPGA, HG-PIPE realizes end-to-end ViT acceleration on a single device and achieves 7118 images/s, which is 2.81 times faster than a V100 GPU.

cs.AR↗

arXiv2024

Quantifying predictive uncertainty of aphasia severity in stroke patients with sparse heteroscedastic Bayesian high-dimensional regression

Sparse linear regression methods for high-dimensional data commonly assume that residuals have constant variance, which can be violated in practice. For example, Aphasia Quotient (AQ) is a critical measure of language impairment and informs treatment decisions, but it is challenging to measure in stroke patients. It is of interest to use high-resolution T2 neuroimages of brain damage to predict AQ. However, sparse regression models show marked evidence of heteroscedastic error even after transformations are applied. This violation of the homoscedasticity assumption can lead to bias in estimated coefficients, prediction intervals (PI) with improper length, and increased type I errors. Bayesian heteroscedastic linear regression models relax the homoscedastic error assumption but can enforce restrictive prior assumptions on parameters, and many are computationally infeasible in the high-dimensional setting. This paper proposes estimating high-dimensional heteroscedastic linear regression models using a heteroscedastic partitioned empirical Bayes Expectation Conditional Maximization (H-PROBE) algorithm. H-PROBE is a computationally efficient maximum a posteriori estimation approach that requires minimal prior assumptions and can incorporate covariates hypothesized to impact heterogeneity. We apply this method by using high-dimensional neuroimages to predict and provide PIs for AQ that accurately quantify predictive uncertainty. Our analysis demonstrates that H-PROBE can provide narrower PI widths than standard methods without sacrificing coverage. Narrower PIs are clinically important for determining the risk of moderate to severe aphasia. Additionally, through extensive simulation studies, we exhibit that H-PROBE results in superior prediction, variable selection, and predictive inference compared to alternative methods.

stat.ME↗

arXiv2024

Topological Persistence Guided Knowledge Distillation for Wearable Sensor Data

Deep learning methods have achieved a lot of success in various applications involving converting wearable sensor data to actionable health insights. A common application areas is activity recognition, where deep-learning methods still suffer from limitations such as sensitivity to signal quality, sensor characteristic variations, and variability between subjects. To mitigate these issues, robust features obtained by topological data analysis (TDA) have been suggested as a potential solution. However, there are two significant obstacles to using topological features in deep learning: (1) large computational load to extract topological features using TDA, and (2) different signal representations obtained from deep learning and TDA which makes fusion difficult. In this paper, to enable integration of the strengths of topological methods in deep-learning for time-series data, we propose to use two teacher networks, one trained on the raw time-series data, and another trained on persistence images generated by TDA methods. The distilled student model utilizes only the raw time-series data at test-time. This approach addresses both issues. The use of KD with multiple teachers utilizes complementary information, and results in a compact model with strong supervisory features and an integrated richer representation. To assimilate desirable information from different modalities, we design new constraints, including orthogonality imposed on feature correlation maps for improving feature expressiveness and allowing the student to easily learn from the teacher. Also, we apply an annealing strategy in KD for fast saturation and better accommodation from different features, while the knowledge gap between the teachers and student is reduced. Finally, a robust student model is distilled, which uses only the time-series data as an input, while implicitly preserving topological features.

eess.SP↗

arXiv2024

Do Large Language Models Rank Fairly? An Empirical Study on the Fairness of LLMs as Rankers

The integration of Large Language Models (LLMs) in information retrieval has raised a critical reevaluation of fairness in the text-ranking models. LLMs, such as GPT models and Llama2, have shown effectiveness in natural language understanding tasks, and prior works (e.g., RankGPT) have also demonstrated that the LLMs exhibit better performance than the traditional ranking models in the ranking task. However, their fairness remains largely unexplored. This paper presents an empirical study evaluating these LLMs using the TREC Fair Ranking dataset, focusing on the representation of binary protected attributes such as gender and geographic location, which are historically underrepresented in search outcomes. Our analysis delves into how these LLMs handle queries and documents related to these attributes, aiming to uncover biases in their ranking algorithms. We assess fairness from both user and content perspectives, contributing an empirical benchmark for evaluating LLMs as the fair ranker.

cs.IR↗

arXiv2024

Do Not Wait: Learning Re-Ranking Model Without User Feedback At Serving Time in E-Commerce

Recommender systems have been widely used in e-commerce, and re-ranking models are playing an increasingly significant role in the domain, which leverages the inter-item influence and determines the final recommendation lists. Online learning methods keep updating a deployed model with the latest available samples to capture the shifting of the underlying data distribution in e-commerce. However, they depend on the availability of real user feedback, which may be delayed by hours or even days, such as item purchases, leading to a lag in model enhancement. In this paper, we propose a novel extension of online learning methods for re-ranking modeling, which we term LAST, an acronym for Learning At Serving Time. It circumvents the requirement of user feedback by using a surrogate model to provide the instructional signal needed to steer model improvement. Upon receiving an online request, LAST finds and applies a model modification on the fly before generating a recommendation result for the request. The modification is request-specific and transient. It means the modification is tailored to and only to the current request to capture the specific context of the request. After a request, the modification is discarded, which helps to prevent error propagation and stabilizes the online learning procedure since the predictions of the surrogate model may be inaccurate. Most importantly, as a complement to feedback-based online learning methods, LAST can be seamlessly integrated into existing online learning systems to create a more adaptive and responsive recommendation experience. Comprehensive experiments, both offline and online, affirm that LAST outperforms state-of-the-art re-ranking models.

cs.IR↗

arXiv2024

Reparameterizable Dual-Resolution Network for Real-time Semantic Segmentation

Semantic segmentation plays a key role in applications such as autonomous driving and medical image. Although existing real-time semantic segmentation models achieve a commendable balance between accuracy and speed, their multi-path blocks still affect overall speed. To address this issue, this study proposes a Reparameterizable Dual-Resolution Network (RDRNet) dedicated to real-time semantic segmentation. Specifically, RDRNet employs a two-branch architecture, utilizing multi-path blocks during training and reparameterizing them into single-path blocks during inference, thereby enhancing both accuracy and inference speed simultaneously. Furthermore, we propose the Reparameterizable Pyramid Pooling Module (RPPM) to enhance the feature representation of the pyramid pooling module without increasing its inference time. Experimental results on the Cityscapes, CamVid, and Pascal VOC 2012 datasets demonstrate that RDRNet outperforms existing state-of-the-art models in terms of both performance and speed. The code is available at https://github.com/gyyang23/RDRNet.

cs.CV↗

arXiv2024

Holistic-Motion2D: Scalable Whole-body Human Motion Generation in 2D Space

In this paper, we introduce a novel path to $\textit{general}$ human motion generation by focusing on 2D space. Traditional methods have primarily generated human motions in 3D, which, while detailed and realistic, are often limited by the scope of available 3D motion data in terms of both the size and the diversity. To address these limitations, we exploit extensive availability of 2D motion data. We present $\textbf{Holistic-Motion2D}$, the first comprehensive and large-scale benchmark for 2D whole-body motion generation, which includes over 1M in-the-wild motion sequences, each paired with high-quality whole-body/partial pose annotations and textual descriptions. Notably, Holistic-Motion2D is ten times larger than the previously largest 3D motion dataset. We also introduce a baseline method, featuring innovative $\textit{whole-body part-aware attention}$ and $\textit{confidence-aware modeling}$ techniques, tailored for 2D $\underline{\text T}$ext-driv$\underline{\text{EN}}$ whole-bo$\underline{\text D}$y motion gen$\underline{\text{ER}}$ation, namely $\textbf{Tender}$. Extensive experiments demonstrate the effectiveness of $\textbf{Holistic-Motion2D}$ and $\textbf{Tender}$ in generating expressive, diverse, and realistic human motions. We also highlight the utility of 2D motion for various downstream applications and its potential for lifting to 3D motion. The page link is: https://holistic-motion2d.github.io.

cs.CV↗

arXiv2024

Audio Matters Too! Enhancing Markerless Motion Capture with Audio Signals for String Performance Capture

In this paper, we touch on the problem of markerless multi-modal human motion capture especially for string performance capture which involves inherently subtle hand-string contacts and intricate movements. To fulfill this goal, we first collect a dataset, named String Performance Dataset (SPD), featuring cello and violin performances. The dataset includes videos captured from up to 23 different views, audio signals, and detailed 3D motion annotations of the body, hands, instrument, and bow. Moreover, to acquire the detailed motion annotations, we propose an audio-guided multi-modal motion capture framework that explicitly incorporates hand-string contacts detected from the audio signals for solving detailed hand poses. This framework serves as a baseline for string performance capture in a completely markerless manner without imposing any external devices on performers, eliminating the potential of introducing distortion in such delicate movements. We argue that the movements of performers, particularly the sound-producing gestures, contain subtle information often elusive to visual methods but can be inferred and retrieved from audio cues. Consequently, we refine the vision-based motion capture results through our innovative audio-guided approach, simultaneously clarifying the contact relationship between the performer and the instrument, as deduced from the audio. We validate the proposed framework and conduct ablation studies to demonstrate its efficacy. Our results outperform current state-of-the-art vision-based algorithms, underscoring the feasibility of augmenting visual motion capture with audio modality. To the best of our knowledge, SPD is the first dataset for musical instrument performance, covering fine-grained hand motion details in a multi-modal, large-scale collection.

cs.MM↗

arXiv2024

Mathematical Foundation and Corrections for Full Range Head Pose Estimation

Numerous works concerning head pose estimation (HPE) offer algorithms or proposed neural network-based approaches for extracting Euler angles from either facial key points or directly from images of the head region. However, many works failed to provide clear definitions of the coordinate systems and Euler or Tait-Bryan angles orders in use. It is a well-known fact that rotation matrices depend on coordinate systems, and yaw, roll, and pitch angles are sensitive to their application order. Without precise definitions, it becomes challenging to validate the correctness of the output head pose and drawing routines employed in prior works. In this paper, we thoroughly examined the Euler angles defined in the 300W-LP dataset, head pose estimation such as 3DDFA-v2, 6D-RepNet, WHENet, etc, and the validity of their drawing routines of the Euler angles. When necessary, we infer their coordinate system and sequence of yaw, roll, pitch from provided code. This paper presents (1) code and algorithms for inferring coordinate system from provided source code, code for Euler angle application order and extracting precise rotation matrices and the Euler angles, (2) code and algorithms for converting poses from one rotation system to another, (3) novel formulae for 2D augmentations of the rotation matrices, and (4) derivations and code for the correct drawing routines for rotation matrices and poses. This paper also addresses the feasibility of defining rotations with right-handed coordinate system in Wikipedia and SciPy, which makes the Euler angle extraction much easier for full-range head pose research.

cs.CV↗

arXiv2024

An Aggregation-Free Federated Learning for Tackling Data Heterogeneity

The performance of Federated Learning (FL) hinges on the effectiveness of utilizing knowledge from distributed datasets. Traditional FL methods adopt an aggregate-then-adapt framework, where clients update local models based on a global model aggregated by the server from the previous training round. This process can cause client drift, especially with significant cross-client data heterogeneity, impacting model performance and convergence of the FL algorithm. To address these challenges, we introduce FedAF, a novel aggregation-free FL algorithm. In this framework, clients collaboratively learn condensed data by leveraging peer knowledge, the server subsequently trains the global model using the condensed data and soft labels received from the clients. FedAF inherently avoids the issue of client drift, enhances the quality of condensed data amid notable data heterogeneity, and improves the global model performance. Extensive numerical studies on several popular benchmark datasets show FedAF surpasses various state-of-the-art FL algorithms in handling label-skew and feature-skew data heterogeneity, leading to superior global model accuracy and faster convergence.

cs.CV↗

arXiv2024

Monad: Towards Cost-effective Specialization for Chiplet-based Spatial Accelerators

Advanced packaging offers a new design paradigm in the post-Moore era, where many small chiplets can be assembled into a large system. Based on heterogeneous integration, a chiplet-based accelerator can be highly specialized for a specific workload, demonstrating extreme efficiency and cost reduction. To fully leverage this potential, it is critical to explore both the architectural design space for individual chiplets and different integration options to assemble these chiplets, which have yet to be fully exploited by existing proposals. This paper proposes Monad, a cost-aware specialization approach for chiplet-based spatial accelerators that explores the tradeoffs between PPA and fabrication costs. To evaluate a specialized system, we introduce a modeling framework considering the non-uniformity in dataflow, pipelining, and communications when executing multiple tensor workloads on different chiplets. We propose to combine the architecture and integration design space by uniformly encoding the design aspects for both spaces and exploring them with a systematic ML-based approach. The experiments demonstrate that Monad can achieve an average of 16% and 30% EDP reduction compared with the state-of-the-art chiplet-based accelerators, Simba and NN-Baton, respectively.

cs.AR↗

arXiv2024

Using Neural Networks to Model Hysteretic Kinematics in Tendon-Actuated Continuum Robots

The ability to accurately model mechanical hysteretic behavior in tendon-actuated continuum robots using deep learning approaches is a growing area of interest. In this paper, we investigate the hysteretic response of two types of tendon-actuated continuum robots and, ultimately, compare three types of neural network modeling approaches with both forward and inverse kinematic mappings: feedforward neural network (FNN), FNN with a history input buffer, and long short-term memory (LSTM) network. We seek to determine which model best captures temporal dependent behavior. We find that, depending on the robot's design, choosing different kinematic inputs can alter whether hysteresis is exhibited by the system. Furthermore, we present the results of the model fittings, revealing that, in contrast to the standard FNN, both FNN with a history input buffer and the LSTM model exhibit the capacity to model historical dependence with comparable performance in capturing rate-dependent hysteresis.

cs.RO↗

arXiv2024

Observation of dichotomic field-tunable electronic structure in twisted monolayer-bilayer graphene

Twisted bilayer graphene (tBLG) provides a fascinating platform for engineering flat bands and inducing correlated phenomena. By designing the stacking architecture of graphene layers, twisted multilayer graphene can exhibit different symmetries with rich tunability. For example, in twisted monolayer-bilayer graphene (tMBG) which breaks the C2z symmetry, transport measurements reveal an asymmetric phase diagram under an out-of-plane electric field, exhibiting correlated insulating state and ferromagnetic state respectively when reversing the field direction. Revealing how the electronic structure evolves with electric field is critical for providing a better understanding of such asymmetric field-tunable properties. Here we report the experimental observation of field-tunable dichotomic electronic structure of tMBG by nanospot angle-resolved photoemission spectroscopy (NanoARPES) with operando gating. Interestingly, selective enhancement of the relative spectral weight contributions from monolayer and bilayer graphene is observed when switching the polarity of the bias voltage. Combining experimental results with theoretical calculations, the origin of such field-tunable electronic structure, resembling either tBLG or twisted double-bilayer graphene (tDBG), is attributed to the selectively enhanced contribution from different stacking graphene layers with a strong electron-hole asymmetry. Our work provides electronic structure insights for understanding the rich field-tunable physics of tMBG.

cond-mat.mes-hall↗