arXiv subjects

Ziyue Yang

Publications and source records attributed to Ziyue Yang.

32 records · Page 2Linked to original sources

arXiv2025

Dynamic Programming: From Local Optimality to Global Optimality

In the theory of dynamic programming, an optimal policy is a policy whose lifetime value dominates that of all other policies from every possible initial condition in the state space. This raises a natural question: when does optimality from a single state imply optimality from every state? Working in a general setting, we provide sufficient conditions for this property that relate to reachability and irreducibility. Our results have significant implications for modern policy-based algorithms used to solve large-scale dynamic programs. We illustrate our findings by applying them to an optimal savings problem via an algorithm that implements gradient ascent in a policy space constructed from neural networks.

math.OC↗

arXiv2025

Uncertainty Aware Human-machine Collaboration in Camouflaged Object Detection

Camouflaged Object Detection (COD), the task of identifying objects concealed within their environments, has seen rapid growth due to its wide range of practical applications. A key step toward developing trustworthy COD systems is the estimation and effective utilization of uncertainty. In this work, we propose a human-machine collaboration framework for classifying the presence of camouflaged objects, leveraging the complementary strengths of computer vision (CV) models and noninvasive brain-computer interfaces (BCIs). Our approach introduces a multiview backbone to estimate uncertainty in CV model predictions, utilizes this uncertainty during training to improve efficiency, and defers low-confidence cases to human evaluation via RSVP-based BCIs during testing for more reliable decision-making. We evaluated the framework in the CAMO dataset, achieving state-of-the-art results with an average improvement of 4.56\% in balanced accuracy (BA) and 3.66\% in the F1 score compared to existing methods. For the best-performing participants, the improvements reached 7.6\% in BA and 6.66\% in the F1 score. Analysis of the training process revealed a strong correlation between our confidence measures and precision, while an ablation study confirmed the effectiveness of the proposed training policy and the human-machine collaboration strategy. In general, this work reduces human cognitive load, improves system reliability, and provides a strong foundation for advancements in real-world COD applications and human-computer interaction. Our code and data are available at: https://github.com/ziyuey/Uncertainty-aware-human-machine-collaboration-in-camouflaged-object-identification.

cs.CV↗

arXiv2025

CSSSTN: A Class-sensitive Subject-to-subject Semantic Style Transfer Network for EEG Classification in RSVP Tasks

The Rapid Serial Visual Presentation (RSVP) paradigm represents a promising application of electroencephalography (EEG) in Brain-Computer Interface (BCI) systems. However, cross-subject variability remains a critical challenge, particularly for BCI-illiterate users who struggle to effectively interact with these systems. To address this issue, we propose the Class-Sensitive Subject-to-Subject Semantic Style Transfer Network (CSSSTN), which incorporates a class-sensitive approach to align feature distributions between golden subjects (BCI experts) and target (BCI-illiterate) users on a class-by-class basis. Building on the SSSTN framework, CSSSTN incorporates three key components: (1) subject-specific classifier training, (2) a unique style loss to transfer class-discriminative features while preserving semantic information through a modified content loss, and (3) an ensemble approach to integrate predictions from both source and target domains. We evaluated CSSSTN using both a publicly available dataset and a self-collected dataset. Experimental results demonstrate that CSSSTN outperforms state-of-the-art methods, achieving mean balanced accuracy improvements of 6.4\% on the Tsinghua dataset and 3.5\% on the HDU dataset, with notable benefits for BCI-illiterate users. Ablation studies confirm the effectiveness of each component, particularly the class-sensitive transfer and the use of lower-layer features, which enhance transfer performance and mitigate negative transfer. Additionally, CSSSTN achieves competitive results with minimal target data, reducing calibration time and effort. These findings highlight the practical potential of CSSSTN for real-world BCI applications, offering a robust and scalable solution to improve the performance of BCI-illiterate users while minimizing reliance on extensive training data. Our code is available at https://github.com/ziyuey/CSSSTN.

eess.SP↗

arXiv2025

Sigma: Differential Rescaling of Query, Key and Value for Efficient Language Models

We introduce Sigma, an efficient large language model specialized for the system domain, empowered by a novel architecture including DiffQKV attention, and pre-trained on our meticulously collected system domain data. DiffQKV attention significantly enhances the inference efficiency of Sigma by optimizing the Query (Q), Key (K), and Value (V) components in the attention mechanism differentially, based on their varying impacts on the model performance and efficiency indicators. Specifically, we (1) conduct extensive experiments that demonstrate the model's varying sensitivity to the compression of K and V components, leading to the development of differentially compressed KV, and (2) propose augmented Q to expand the Q head dimension, which enhances the model's representation capacity with minimal impacts on the inference speed. Rigorous theoretical and empirical analyses reveal that DiffQKV attention significantly enhances efficiency, achieving up to a 33.36% improvement in inference speed over the conventional grouped-query attention (GQA) in long-context scenarios. We pre-train Sigma on 6T tokens from various sources, including 19.5B system domain data that we carefully collect and 1T tokens of synthesized and rewritten data. In general domains, Sigma achieves comparable performance to other state-of-arts models. In the system domain, we introduce the first comprehensive benchmark AIMicius, where Sigma demonstrates remarkable performance across all tasks, significantly outperforming GPT-4 with an absolute improvement up to 52.5%.

cs.CL↗

arXiv2024

SUANPAN: Scalable Photonic Linear Vector Machine

Photonic linear operation is a promising approach to handle the extensive vector multiplications in artificial intelligence techniques due to the natural bosonic parallelism and high-speed information transmission of photonics. Although it is believed that maximizing the interaction of the light beams is necessary to fully utilize the parallelism and tremendous efforts have been made in past decades, the achieved dimensionality of vector-matrix multiplication is very limited due to the difficulty of scaling up a tightly interconnected or highly coupled optical system. Additionally, there is still a lack of a universal photonic computing architecture that can be readily merged with existing computing system to meet the computing power demand of AI techniques. Here, we propose a programmable and reconfigurable photonic linear vector machine to perform only the inner product of two vectors, formed by a series of independent basic computing units, while each unit is just one pair of light-emitter and photodetector. Since there is no interaction among light beams inside, extreme scalability could be achieved by simply duplicating the independent basic computing unit while there is no requirement of large-scale analog-to-digital converter and digital-to-analog converter arrays. Our architecture is inspired by the traditional Chinese Suanpan or abacus and thus is denoted as photonic SUANPAN. As a proof of principle, SUANPAN architecture is implemented with an 8*8 vertical cavity surface emission laser array and an 8*8 MoTe2 two-dimensional material photodetector array. We believe that our proposed photonic SUANPAN is capable of serving as a fundamental linear vector machine that can be readily merged with existing electronic digital computing system and is potential to enhance the computing power for future various AI applications.

physics.optics↗

arXiv2024

SPFresh: Incremental In-Place Update for Billion-Scale Vector Search

Approximate Nearest Neighbor Search (ANNS) is now widely used in various applications, ranging from information retrieval, question answering, and recommendation, to search for similar high-dimensional vectors. As the amount of vector data grows continuously, it becomes important to support updates to vector index, the enabling technique that allows for efficient and accurate ANNS on vectors. Because of the curse of high dimensionality, it is often costly to identify the right neighbors of a single new vector, a necessary process for index update. To amortize update costs, existing systems maintain a secondary index to accumulate updates, which are merged by the main index by global rebuilding the entire index periodically. However, this approach has high fluctuations of search latency and accuracy, not even to mention that it requires substantial resources and is extremely time-consuming for rebuilds. We introduce SPFresh, a system that supports in-place vector updates. At the heart of SPFresh is LIRE, a lightweight incremental rebalancing protocol to split vector partitions and reassign vectors in the nearby partitions to adapt to data distribution shift. LIRE achieves low-overhead vector updates by only reassigning vectors at the boundary between partitions, where in a high-quality vector index the amount of such vectors are deemed small. With LIRE, SPFresh provides superior query latency and accuracy to solutions based on global rebuild, with only 1% of DRAM and less than 10% cores needed at the peak compared to the state-of-the-art, in a billion scale vector index with 1% of daily vector update rate.

cs.IR↗

arXiv2024

SuperBench: Improving Cloud AI Infrastructure Reliability with Proactive Validation

Reliability in cloud AI infrastructure is crucial for cloud service providers, prompting the widespread use of hardware redundancies. However, these redundancies can inadvertently lead to hidden degradation, so called "gray failure", for AI workloads, significantly affecting end-to-end performance and concealing performance issues, which complicates root cause analysis for failures and regressions. We introduce SuperBench, a proactive validation system for AI infrastructure that mitigates hidden degradation caused by hardware redundancies and enhances overall reliability. SuperBench features a comprehensive benchmark suite, capable of evaluating individual hardware components and representing most real AI workloads. It comprises a Validator which learns benchmark criteria to clearly pinpoint defective components. Additionally, SuperBench incorporates a Selector to balance validation time and issue-related penalties, enabling optimal timing for validation execution with a tailored subset of benchmarks. Through testbed evaluation and simulation, we demonstrate that SuperBench can increase the mean time between incidents by up to 22.61x. SuperBench has been successfully deployed in Azure production, validating hundreds of thousands of GPUs over the last two years.

cs.DC↗

arXiv2024

16-channel Photonic Solver for Optimization Problems on a Silicon Chip

In this article, we proposed a programmable 16-channel photonic solver for quadratic unconstrained binary optimization (QUBO) problems. The solver is based on a hybrid optoelectronic scheme including a photonic chip and the corresponding electronic driving circuit. The photonic chip is fabricated on silicon on insulator (SOI) substrate and integrates high-speed electro-optic modulators, thermo-optic phase shifters and photodetectors to conduct the 16-dimensional optical vector-matrix multiplication (OVMM). Due to the parallel and low latency propagation of lightwave, the calculation of the QUBO cost function can be accelerated. Besides, the electronic processor is employed to run the heuristic algorithm to search the optimal solution. In the experiment, two 16-dimensional randomly generated QUBO problems are solved with high successful probabilities. To our knowledge, it is the largest scale of programmable and on-chip photonic solver ever reported. Moreover, the computing speed of the OVMM on photonic chip is ~2 TFLOP/s. It shows the potential of fast solving such optimization problems with integrated photonic systems.

cs.ET↗

arXiv2023

FP8-LM: Training FP8 Large Language Models

In this paper, we explore FP8 low-bit data formats for efficient training of large language models (LLMs). Our key insight is that most variables, such as gradients and optimizer states, in LLM training can employ low-precision data formats without compromising model accuracy and requiring no changes to hyper-parameters. Specifically, we propose a new FP8 automatic mixed-precision framework for training LLMs. This framework offers three levels of FP8 utilization to streamline mixed-precision and distributed parallel training for LLMs. It gradually incorporates 8-bit gradients, optimizer states, and distributed learning in an incremental manner. Experiment results show that, during the training of GPT-175B model on H100 GPU platform, our FP8 mixed-precision training framework not only achieved a remarkable 39% reduction in real memory usage but also ran 75% faster than the widely adopted BF16 framework (i.e., Megatron-LM), surpassing the speed of Nvidia Transformer Engine by 37%. This largely reduces the training costs for large foundation models. Furthermore, our FP8 mixed-precision training methodology is generic. It can be seamlessly applied to other tasks such as LLM instruction tuning and reinforcement learning with human feedback, offering savings in fine-tuning expenses. Our FP8 low-precision training framework is open-sourced at {https://github.com/Azure/MS-AMP}{aka.ms/MS.AMP}.

cs.LG↗

arXiv2023

Tutel: Adaptive Mixture-of-Experts at Scale

Sparsely-gated mixture-of-experts (MoE) has been widely adopted to scale deep learning models to trillion-plus parameters with fixed computational cost. The algorithmic performance of MoE relies on its token routing mechanism that forwards each input token to the right sub-models or experts. While token routing dynamically determines the amount of expert workload at runtime, existing systems suffer inefficient computation due to their static execution, namely static parallelism and pipelining, which does not adapt to the dynamic workload. We present Flex, a highly scalable stack design and implementation for MoE with dynamically adaptive parallelism and pipelining. Flex designs an identical layout for distributing MoE model parameters and input data, which can be leveraged by all possible parallelism or pipelining methods without any mathematical inequivalence or tensor migration overhead. This enables adaptive parallelism/pipelining optimization at zero cost during runtime. Based on this key design, Flex also implements various MoE acceleration techniques. Aggregating all techniques, Flex finally delivers huge speedup at any scale -- 4.96x and 5.75x speedup of a single MoE layer over 16 and 2,048 A100 GPUs, respectively, over the previous state-of-the-art. Our evaluation shows that Flex efficiently and effectively runs a real-world MoE-based model named SwinV2-MoE, built upon Swin Transformer V2, a state-of-the-art computer vision architecture. On efficiency, Flex accelerates SwinV2-MoE, achieving up to 1.55x and 2.11x speedup in training and inference over Fairseq, respectively. On effectiveness, the SwinV2-MoE model achieves superior accuracy in both pre-training and down-stream computer vision tasks such as COCO object detection than the counterpart dense model, indicating the readiness of Flex for end-to-end real-world model training and inference.

cs.DC↗

arXiv2023

Copyright Protection and Accountability of Generative AI:Attack, Watermarking and Attribution

Generative AI (e.g., Generative Adversarial Networks - GANs) has become increasingly popular in recent years. However, Generative AI introduces significant concerns regarding the protection of Intellectual Property Rights (IPR) (resp. model accountability) pertaining to images (resp. toxic images) and models (resp. poisoned models) generated. In this paper, we propose an evaluation framework to provide a comprehensive overview of the current state of the copyright protection measures for GANs, evaluate their performance across a diverse range of GAN architectures, and identify the factors that affect their performance and future research directions. Our findings indicate that the current IPR protection methods for input images, model watermarking, and attribution networks are largely satisfactory for a wide range of GANs. We highlight that further attention must be directed towards protecting training sets, as the current approaches fail to provide robust IPR protection and provenance tracing on training sets.

cs.LG↗

arXiv2021

ModelDiff: Testing-Based DNN Similarity Comparison for Model Reuse Detection

The knowledge of a deep learning model may be transferred to a student model, leading to intellectual property infringement or vulnerability propagation. Detecting such knowledge reuse is nontrivial because the suspect models may not be white-box accessible and/or may serve different tasks. In this paper, we propose ModelDiff, a testing-based approach to deep learning model similarity comparison. Instead of directly comparing the weights, activations, or outputs of two models, we compare their behavioral patterns on the same set of test inputs. Specifically, the behavioral pattern of a model is represented as a decision distance vector (DDV), in which each element is the distance between the model's reactions to a pair of inputs. The knowledge similarity between two models is measured with the cosine similarity between their DDVs. To evaluate ModelDiff, we created a benchmark that contains 144 pairs of models that cover most popular model reuse methods, including transfer learning, model compression, and model stealing. Our method achieved 91.7% correctness on the benchmark, which demonstrates the effectiveness of using ModelDiff for model reuse detection. A study on mobile deep learning apps has shown the feasibility of ModelDiff on real-world models.

cs.LG↗

arXiv2020

Humanoid: A Deep Learning-based Approach to Automated Black-box Android App Testing

Automated input generators are widely used for large-scale dynamic analysis of mobile apps. Such input generators must constantly choose which UI element to interact with and how to interact with it, in order to achieve high coverage with a limited time budget. Currently, most input generators adopt pseudo-random or brute-force searching strategies, which may take very long to find the correct combination of inputs that can drive the app into new and important states. In this paper, we propose Humanoid, a deep learning-based approach to GUI test input generation by learning from human interactions. Our insight is that if we can learn from human-generated interaction traces, it is possible to automatically prioritize test inputs based on their importance as perceived by users. We design and implement a deep neural network model to learn how end-users would interact with an app (specifically, which UI elements to interact with and how). Our experiments showed that the interaction model can successfully prioritize user-preferred inputs for any new UI (with a top-1 accuracy of 51.2% and a top-10 accuracy of 85.2%). We implemented an input generator for Android apps based on the learned model and evaluated it on both open-source apps and market apps. The results indicated that Humanoid was able to achieve higher coverage than six state-of-the-art test generators. However, further analysis showed that the learned model was not the main reason of coverage improvement. Although the learned interaction pattern could drive the app into some important GUI states with higher probabilities, it had limited effect on the width and depth of GUI state search, which is the key to improve test coverage in the long term. Whether and how human interaction patterns can be used to improve coverage is still an unknown and challenging problem.

cs.SE↗

arXiv2018

Automated Extraction of Personal Knowledge from Smartphone Push Notifications

Personalized services are in need of a rich and powerful personal knowledge base, i.e. a knowledge base containing information about the user. This paper proposes an approach to extracting personal knowledge from smartphone push notifications, which are used by mobile systems and apps to inform users of a rich range of information. Our solution is based on the insight that most notifications are formatted using templates, while knowledge entities can be usually found within the parameters to the templates. As defining all the notification templates and their semantic rules are impractical due to the huge number of notification templates used by potentially millions of apps, we propose an automated approach for personal knowledge extraction from push notifications. We first discover notification templates through pattern mining, then use machine learning to understand the template semantics. Based on the templates and their semantics, we are able to translate notification text into knowledge facts automatically. Users' privacy is preserved as we only need to upload the templates to the server for model training, which do not contain any personal information. According to our experiments with about 120 million push notifications from 100,000 smartphone users, our system is able to extract personal knowledge accurately and efficiently.

cs.IR↗