arXiv subjects

Yi Deng

Publications and source records attributed to Yi Deng.

17 recordsLinked to original sources

arXiv2025

Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod

Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentralized serving. This report presents xDeepServe, the production serving system behind Huawei Cloud's MaaS offering on CloudMatrix384, a 48-server SuperPod with 384 Ascend 910C chips connected by a high-bandwidth UB fabric and global shared memory. It serves models including DeepSeek, Kimi, GLM, Qwen, and MiniMax, among others. xDeepServe is built around Transformerless, a disaggregated execution architecture that decomposes transformer inference into modular units -- attention, feedforward, and MoE -- and supports disaggregated Prefill-Decode and MoE-Attention deployments. To enable disaggregation, we develop XCCL, a memory-semantic communication layer providing microsecond-level point-to-point and scalable all-to-all primitives, and we extend FlowServe with decentralized DP groups and techniques to mitigate stragglers and synchronization variance. In a peak decoding configuration, xDeepServe reaches 2400 tokens/s per Ascend 910C chip at ~50ms time-per-output-token (TPOT).

cs.DC↗

arXiv2025

Online survival analysis with quantile regression

We propose an online inference method for censored quantile regression with streaming data sets. A key strategy is to approximate the martingale-based unsmooth objective function with a quadratic loss function involving a well-justified second-order expansion. This enables us to derive a new online convex function based on the current data batch and summary statistics of historical data, thereby achieving online updating and occupying low storage space. To estimate the regression parameters, we design a novel majorize-minimize algorithm by reasonably constructing a quadratic surrogate objective function, which renders a closed-form parameter update and thus reduces the computational burden notably. Theoretically, compared to the oracle estimators derived from analyzing the entire raw data once, we posit a weaker assumption on the quantile grid size and show that the proposed online estimators can maintain the same convergence rate and statistical efficiency. Simulation studies and an application demonstrate the satisfactory empirical performance and practical utilities of the proposed online method.

math.ST↗

arXiv2025

NTIRE 2025 Challenge on Short-form UGC Video Quality Assessment and Enhancement: Methods and Results

This paper presents a review for the NTIRE 2025 Challenge on Short-form UGC Video Quality Assessment and Enhancement. The challenge comprises two tracks: (i) Efficient Video Quality Assessment (KVQ), and (ii) Diffusion-based Image Super-Resolution (KwaiSR). Track 1 aims to advance the development of lightweight and efficient video quality assessment (VQA) models, with an emphasis on eliminating reliance on model ensembles, redundant weights, and other computationally expensive components in the previous IQA/VQA competitions. Track 2 introduces a new short-form UGC dataset tailored for single image super-resolution, i.e., the KwaiSR dataset. It consists of 1,800 synthetically generated S-UGC image pairs and 1,900 real-world S-UGC images, which are split into training, validation, and test sets using a ratio of 8:1:1. The primary objective of the challenge is to drive research that benefits the user experience of short-form UGC platforms such as Kwai and TikTok. This challenge attracted 266 participants and received 18 valid final submissions with corresponding fact sheets, significantly contributing to the progress of short-form UGC VQA and image superresolution. The project is publicly available at https://github.com/lixinustc/KVQE- ChallengeCVPR-NTIRE2025.

eess.IV↗

arXiv2024

SynDCIM: A Performance-Aware Digital Computing-in-Memory Compiler with Multi-Spec-Oriented Subcircuit Synthesis

Digital Computing-in-Memory (DCIM) is an innovative technology that integrates multiply-accumulation (MAC) logic directly into memory arrays to enhance the performance of modern AI computing. However, the need for customized memory cells and logic components currently necessitates significant manual effort in DCIM design. Existing tools for facilitating DCIM macro designs struggle to optimize subcircuit synthesis to meet user-defined performance criteria, thereby limiting the potential system-level acceleration that DCIM can offer. To address these challenges and enable agile design of DCIM macros with optimal architectures, we present SynDCIM, a performance-aware DCIM compiler that employs multi-spec-oriented subcircuit synthesis. SynDCIM features an automated performance-to-layout generation process that aligns with user-defined performance expectations. This is supported by a scalable subcircuit library and a multi-spec-oriented searching algorithm for effective subcircuit synthesis. The effectiveness of SynDCIM is demonstrated through extensive experiments and validated with a test chip fabricated in a 40nm CMOS process. Testing results reveal that designs generated by SynDCIM exhibit competitive performance when compared to state-of-the-art manually designed DCIM macros.

cs.AR↗

arXiv2024

Large Model Based Agents: State-of-the-Art, Cooperation Paradigms, Security and Privacy, and Future Trends

With the rapid advancement of large models (LMs), the development of general-purpose intelligent agents powered by LMs has become a reality. It is foreseeable that in the near future, LM-driven general AI agents will serve as essential tools in production tasks, capable of autonomous communication and collaboration without human intervention. This paper investigates scenarios involving the autonomous collaboration of future LM agents. We review the current state of LM agents, the key technologies enabling LM agent collaboration, and the security and privacy challenges they face during cooperative operations. To this end, we first explore the foundational principles of LM agents, including their general architecture, key components, enabling technologies, and modern applications. We then discuss practical collaboration paradigms from data, computation, and knowledge perspectives to achieve connected intelligence among LM agents. After that, we analyze the security vulnerabilities and privacy risks associated with LM agents, particularly in multi-agent settings, examining underlying mechanisms and reviewing current and potential countermeasures. Lastly, we propose future research directions for building robust and secure LM agent ecosystems.

cs.AI↗

arXiv2024

Seed-Music: A Unified Framework for High Quality and Controlled Music Generation

We introduce Seed-Music, a suite of music generation systems capable of producing high-quality music with fine-grained style control. Our unified framework leverages both auto-regressive language modeling and diffusion approaches to support two key music creation workflows: controlled music generation and post-production editing. For controlled music generation, our system enables vocal music generation with performance controls from multi-modal inputs, including style descriptions, audio references, musical scores, and voice prompts. For post-production editing, it offers interactive tools for editing lyrics and vocal melodies directly in the generated audio. We encourage readers to listen to demo audio examples at https://team.doubao.com/seed-music "https://team.doubao.com/seed-music".

cs.SD↗

arXiv2022

MIDI-DDSP: Detailed Control of Musical Performance via Hierarchical Modeling

Musical expression requires control of both what notes are played, and how they are performed. Conventional audio synthesizers provide detailed expressive controls, but at the cost of realism. Black-box neural audio synthesis and concatenative samplers can produce realistic audio, but have few mechanisms for control. In this work, we introduce MIDI-DDSP a hierarchical model of musical instruments that enables both realistic neural audio synthesis and detailed user control. Starting from interpretable Differentiable Digital Signal Processing (DDSP) synthesis parameters, we infer musical notes and high-level properties of their expressive performance (such as timbre, vibrato, dynamics, and articulation). This creates a 3-level hierarchy (notes, performance, synthesis) that affords individuals the option to intervene at each level, or utilize trained priors (performance given notes, synthesis given performance) for creative assistance. Through quantitative experiments and listening tests, we demonstrate that this hierarchy can reconstruct high-fidelity audio, accurately predict performance attributes for a note sequence, independently manipulate the attributes of a given performance, and as a complete system, generate realistic audio from a novel note sequence. By utilizing an interpretable hierarchy, with multiple levels of granularity, MIDI-DDSP opens the door to assistive tools to empower individuals across a diverse range of musical experience.

cs.SD↗

arXiv2020

Privacy-Preserving Methods for Vertically Partitioned Incomplete Data

Distributed health data networks that use information from multiple sources have drawn substantial interest in recent years. However, missing data are prevalent in such networks and present significant analytical challenges. The current state-of-the-art methods for handling missing data require pooling data into a central repository before analysis, which may not be possible in a distributed health data network. In this paper, we propose a privacy-preserving distributed analysis framework for handling missing data when data are vertically partitioned. In this framework, each institution with a particular data source utilizes the local private data to calculate necessary intermediate aggregated statistics, which are then shared to build a global model for handling missing data. To evaluate our proposed methods, we conduct simulation studies that clearly demonstrate that the proposed privacy-preserving methods perform as well as the methods using the pooled data and outperform several naïve methods. We further illustrate the proposed methods through the analysis of a real dataset. The proposed framework for handling vertically partitioned incomplete data is substantially more privacy-preserving than methods that require pooling of the data, since no individual-level data are shared, which can lower hurdles for collaboration across multiple institutions and build stronger public trust.

cs.CR↗

arXiv2020

Rule-embedded network for audio-visual voice activity detection in live musical video streams

Detecting anchor's voice in live musical streams is an important preprocessing for music and speech signal processing. Existing approaches to voice activity detection (VAD) primarily rely on audio, however, audio-based VAD is difficult to effectively focus on the target voice in noisy environments. With the help of visual information, this paper proposes a rule-embedded network to fuse the audio-visual (A-V) inputs to help the model better detect target voice. The core role of the rule in the model is to coordinate the relation between the bi-modal information and use visual representations as the mask to filter out the information of non-target sound. Experiments show that: 1) with the help of cross-modal fusion by the proposed rule, the detection result of A-V branch outperforms that of audio branch; 2) the performance of bi-modal model far outperforms that of audio-only models, indicating that the incorporation of both audio and visual signals is highly beneficial for VAD. To attract more attention to the cross-modal music and audio signal processing, a new live musical video corpus with frame-level label is introduced.

cs.SD↗

arXiv2020

Temporal Logic Inference for Hybrid System Observation with Spatial and Temporal Uncertainties

In this paper, we present a mechanism for building hybrid system observers to differentiate between specific positions of the hybrid system. The mechanism is designed through inferring metric temporal logic (MTL) formulae from simulated trajectories from the hybrid system. We first approximate the system behavior by simulating finitely many trajectories with timerobust tube segments around them. These time-robust tube segments account for both spatial and temporal uncertainties that exist in the hybrid system with initial state variations. The inferred MTL formulae classify different time-robust tube segments and thus can be used for classifying the hybrid system behaviors in a provably correct fashion. We implement our approach on a model of a smart building testbed to distinguish two cases of room occupancy.

eess.SY↗

arXiv2017

High-Dimensional Dependency Structure Learning for Physical Processes

In this paper, we consider the use of structure learning methods for probabilistic graphical models to identify statistical dependencies in high-dimensional physical processes. Such processes are often synthetically characterized using PDEs (partial differential equations) and are observed in a variety of natural phenomena, including geoscience data capturing atmospheric and hydrological phenomena. Classical structure learning approaches such as the PC algorithm and variants are challenging to apply due to their high computational and sample requirements. Modern approaches, often based on sparse regression and variants, do come with finite sample guarantees, but are usually highly sensitive to the choice of hyper-parameters, e.g., parameter $λ$ for sparsity inducing constraint or regularization. In this paper, we present ACLIME-ADMM, an efficient two-step algorithm for adaptive structure learning, which estimates an edge specific parameter $λ_{ij}$ in the first step, and uses these parameters to learn the structure in the second step. Both steps of our algorithm use (inexact) ADMM to solve suitable linear programs, and all iterations can be done in closed form in an efficient block parallel manner. We compare ACLIME-ADMM with baselines on both synthetic data simulated by partial differential equations (PDEs) that model advection-diffusion processes, and real data (50 years) of daily global geopotential heights to study information flow in the atmosphere. ACLIME-ADMM is shown to be efficient, stable, and competitive, usually better than the baselines especially on difficult problems. On real data, ACLIME-ADMM recovers the underlying structure of global atmospheric circulation, including switches in wind directions at the equator and tropics entirely from the data.

cs.LG↗

arXiv2015

Using Causal Discovery to Track Information Flow in Spatio-Temporal Data - A Testbed and Experimental Results Using Advection-Diffusion Simulations

Causal discovery algorithms based on probabilistic graphical models have emerged in geoscience applications for the identification and visualization of dynamical processes. The key idea is to learn the structure of a graphical model from observed spatio-temporal data, which indicates information flow, thus pathways of interactions, in the observed physical system. Studying those pathways allows geoscientists to learn subtle details about the underlying dynamical mechanisms governing our planet. Initial studies using this approach on real-world atmospheric data have shown great potential for scientific discovery. However, in these initial studies no ground truth was available, so that the resulting graphs have been evaluated only by whether a domain expert thinks they seemed physically plausible. This paper seeks to fill this gap. We develop a testbed that emulates two dynamical processes dominant in many geoscience applications, namely advection and diffusion, in a 2D grid. Then we apply the causal discovery based information tracking algorithms to the simulation data to study how well the algorithms work for different scenarios and to gain a better understanding of the physical meaning of the graph results, in particular of instantaneous connections. We make all data sets used in this study available to the community as a benchmark. Keywords: Information flow, graphical model, structure learning, causal discovery, geoscience.

cs.LG↗

arXiv2015

Safe Neighborhood Computation for Hybrid System Verification

For the design and implementation of engineering systems, performing model-based analysis can disclose potential safety issues at an early stage. The analysis of hybrid system models is in general difficult due to the intrinsic complexity of hybrid dynamics. In this paper, a simulation-based approach to formal verification of hybrid systems is presented.

eess.SY↗

arXiv2012

Efficient Multiple Object Tracking Using Mutually Repulsive Active Membranes

Studies of social and group behavior in interacting organisms require high-throughput analysis of the motion of a large number of individual subjects. Computer vision techniques offer solutions to specific tracking problems, and allow automated and efficient tracking with minimal human intervention. In this work, we adopt the open active contour model to track the trajectories of moving objects at high density. We add repulsive interactions between open contours to the original model, treat the trajectories as an extrusion in the temporal dimension, and show applications to two tracking problems. The walking behavior of Drosophila is studied at different population density and gender composition. We demonstrate that individual male flies have distinct walking signatures, and that the social interaction between flies in a mixed gender arena is gender specific. We also apply our model to studies of trajectories of gliding Myxococcus xanthus bacteria at high density. We examine the individual gliding behavioral statistics in terms of the gliding speed distribution. Using these two examples at very distinctive spatial scales, we illustrate the use of our algorithm on tracking both short rigid bodies (Drosophila) and long flexible objects (Myxococcus xanthus). Our repulsive active membrane model reaches error rates better than $5\times 10^{-6}$ per fly per second for Drosophila tracking and comparable results for Myxococcus xanthus.

q-bio.QM↗

arXiv2011

Direct measurement of cell wall stress-stiffening and turgor pressure in live bacterial cells

We study intact and bulging Escherichia coli cells using atomic force microscopy to separate the contributions of the cell wall and turgor pressure to the overall cell stiffness. We find strong evidence of power-law stress-stiffening in the E. coli cell wall, with an exponent of 1.22 \pm 0.12, such that the wall is significantly stiffer in intact cells (E = 23 \pm 8 MPa and 49 \pm 20 MPa in the axial and circumferential directions) than in unpressurized sacculi. These measurements also indicate that the turgor pressure in living cells E. coli is 29 \pm 3 kPa.

physics.bio-ph↗

arXiv2006

Concurrently Non-Malleable Zero Knowledge in the Authenticated Public-Key Model

We consider a type of zero-knowledge protocols that are of interest for their practical applications within networks like the Internet: efficient zero-knowledge arguments of knowledge that remain secure against concurrent man-in-the-middle attacks. In an effort to reduce the setup assumptions required for efficient zero-knowledge arguments of knowledge that remain secure against concurrent man-in-the-middle attacks, we consider a model, which we call the Authenticated Public-Key (APK) model. The APK model seems to significantly reduce the setup assumptions made by the CRS model (as no trusted party or honest execution of a centralized algorithm are required), and can be seen as a slightly stronger variation of the Bare Public-Key (BPK) model from \cite{CGGM,MR}, and a weaker variation of the registered public-key model used in \cite{BCNP}. We then define and study man-in-the-middle attacks in the APK model. Our main result is a constant-round concurrent non-malleable zero-knowledge argument of knowledge for any polynomial-time relation (associated to a language in $\mathcal{NP}$), under the (minimal) assumption of the existence of a one-way function family. Furthermore,We show time-efficient instantiations of our protocol based on known number-theoretic assumptions. We also note a negative result with respect to further reducing the setup assumptions of our protocol to those in the (unauthenticated) BPK model, by showing that concurrently non-malleable zero-knowledge arguments of knowledge in the BPK model are only possible for trivial languages.

cs.CR↗

arXiv2006

Resettable Zero Knowledge in the Bare Public-Key Model under Standard Assumption

In this paper we resolve an open problem regarding resettable zero knowledge in the bare public-key (BPK for short) model: Does there exist constant round resettable zero knowledge argument with concurrent soundness for $\mathcal{NP}$ in BPK model without assuming \emph{sub-exponential hardness}? We give a positive answer to this question by presenting such a protocol for any language in $\mathcal{NP}$ in the bare public-key model assuming only collision-resistant hash functions against \emph{polynomial-time} adversaries.

cs.CR↗