Yao Chen

Enqiang Zhu, Yizi Liu, Yilong Luo et al.

Accurate prediction of protein-protein interaction sites (PPIS) is essential for understanding cellular processes, disease mechanisms, and therapeutic target discovery. Graph-based deep learning has advanced PPIS prediction by incorporating residue-level structural context. However, most graph-based models still rely on fixed propagation schemes that treat all residues similarly, despite the structural and functional heterogeneity of protein interfaces. Such propagation may limit the ability to adapt information diffusion to local geometric environments, making it difficult to distinguish true interaction sites from structurally similar non-interacting neighbors. We present SGAP-PPIS, a structure-guided adaptive propagation model for PPIS prediction. Rather than using a fixed propagation mechanism, SGAP-PPIS leverages multi-scale geometric states from an equivariant graph neural network to generate residue-wise propagation coefficients. This design allows each residue to adaptively balance local feature preservation and neighborhood diffusion according to its geometric microenvironment. Experimental results show that SGAP-PPIS achieves competitive performance among the state-of-the-art methods on Test\_60. Ablation studies show that geometry-conditioned adaptive propagation, scale-aligned geometric guidance, and multi-step propagation-state representation jointly drive these improvements.

Yao Chen, Samuel S. Streeter, Brady Hunt et al.

In this study, a radiomics approach was extended to optical fluorescence molecular imaging data for tissue classification, termed 'optomics'. Fluorescence molecular imaging is emerging for precise surgical guidance during head and neck squamous cell carcinoma (HNSCC) resection. However, the tumor-to-normal tissue contrast is confounded by intrinsic physiological limitations of heterogeneous expression of the target molecule, epidermal growth factor receptor (EGFR). Optomics seek to improve tumor identification by probing textural pattern differences in EGFR expression conveyed by fluorescence. A total of 1,472 standardized optomic features were extracted from fluorescence image samples. A supervised machine learning pipeline involving a support vector machine classifier was trained with 25 top-ranked features selected by minimum redundancy maximum relevance criterion. Model predictive performance was compared to fluorescence intensity thresholding method by classifying testing set image patches of resected tissue with histologically confirmed malignancy status. The optomics approach provided consistent improvement in prediction accuracy on all test set samples, irrespective of dose, compared to fluorescence intensity thresholding method (mean accuracies of 89% vs. 81%; P = 0.0072). The improved performance demonstrates that extending the radiomics approach to fluorescence molecular imaging data offers a promising image analysis technique for cancer detection in fluorescence-guided surgery.

Yao Chen, Wensheng Gan, Yongdong Wu et al.

Contrast pattern mining (CPM) is an important and popular subfield of data mining. Traditional sequential patterns cannot describe the contrast information between different classes of data, while contrast patterns involving the concept of contrast can describe the significant differences between datasets under different contrast conditions. Based on the number of papers published in this field, we find that researchers' interest in CPM is still active. Since CPM has many research questions and research methods. It is difficult for new researchers in the field to understand the general situation of the field in a short period of time. Therefore, the purpose of this article is to provide an up-to-date comprehensive and structured overview of the research direction of contrast pattern mining. First, we present an in-depth understanding of CPM, including basic concepts, types, mining strategies, and metrics for assessing discriminative ability. Then we classify CPM methods according to their characteristics into boundary-based algorithms, tree-based algorithms, evolutionary fuzzy system-based algorithms, decision tree-based algorithms, and other algorithms. In addition, we list the classical algorithms of these methods and discuss their advantages and disadvantages. Advanced topics in CPM are presented. Finally, we conclude our survey with a discussion of the challenges and opportunities in this field.

Yao Chen, Yijie Gui, Hong Lin et al.

In terms of artificial intelligence, there are several security and privacy deficiencies in the traditional centralized training methods of machine learning models by a server. To address this limitation, federated learning (FL) has been proposed and is known for breaking down ``data silos" and protecting the privacy of users. However, FL has not yet gained popularity in the industry, mainly due to its security, privacy, and high cost of communication. For the purpose of advancing the research in this field, building a robust FL system, and realizing the wide application of FL, this paper sorts out the possible attacks and corresponding defenses of the current FL system systematically. Firstly, this paper briefly introduces the basic workflow of FL and related knowledge of attacks and defenses. It reviews a great deal of research about privacy theft and malicious attacks that have been studied in recent years. Most importantly, in view of the current three classification criteria, namely the three stages of machine learning, the three different roles in federated learning, and the CIA (Confidentiality, Integrity, and Availability) guidelines on privacy protection, we divide attack approaches into two categories according to the training stage and the prediction stage in machine learning. Furthermore, we also identify the CIA property violated for each attack method and potential attack role. Various defense mechanisms are then analyzed separately from the level of privacy and security. Finally, we summarize the possible challenges in the application of FL from the aspect of attacks and defenses and discuss the future development direction of FL systems. In this way, the designed FL system has the ability to resist different attacks and is more secure and stable.

Yi Bai, Wenhao Zhang, Yao Chen et al.

Instruction fine-tuning is employed to enhance the instruction-following ability of large language models (LLMs). As the amount of instruction fine-tuning data increases, selecting the optimal core set becomes particularly important. However, ensuring the diversity of the core set remains a significant challenge. Existing methods predominantly distinguish different training data based on the text features themselves, decoupled from LLMs' own understanding and representation of the data. To address this issue, we propose a Model-Aware Diverse Core Set Selection method, which distinguishes data features based on the neural activation states during LLM inference. This approach serves as an efficient instantiation of coverage-based selection using model-intrinsic activation features to ensure the diversity in the core set. We extensively evaluate our method on six benchmarks that cover five distinct tasks. In our method, the core set selected by the 3B-parameter LLM performs effectively when utilized to fine-tune larger models with 7B, 8B, and 13B parameters. Experimental results on the Alpaca-GPT4 dataset, which comprises 52K instruction-response pairs, show that the core set, sized at 15\% of the original dataset and selected by Llama-3.2-3B-Instruct, achieves an average improvement of 2.5\% when fine-tuning four larger base models compared with training on the full dataset. The experimental results demonstrate that our method enhances model performance on multiple downstream tasks while reducing data requirements.

James Brannick, Yao Chen, Johannes Kraus et al.

This paper presents estimates of the convergence rate and complexity of an algebraic multilevel preconditioner based on piecewise constant coarse vector spaces applied to the graph Laplacian. A bound is derived on the energy norm of the projection operator onto any piecewise constant vector space, which results in an estimate of the two-level convergence rate where the coarse level graph is obtained by matching. The two-level convergence of the method is then used to establish the convergence of an Algebraic Multilevel Iteration that uses the two-level scheme recursively. On structured grids, the method is proven to have convergence rate $\approx (1-1/\log n)$ and $O(n\log n)$ complexity for each cycle, where $n$ denotes the number of unknowns in the given problem. Numerical results of the algorithm applied to various graph Laplacians are reported. It is also shown that all the theoretical estimates derived for matching can be generalized to the case of aggregates containing more than two vertices.

Yao Chen, Shan Huang, Wensheng Gan et al.

The metaverse, which is at the stage of innovation and exploration, faces the dilemma of data collection and the problem of private data leakage in the process of development. This can seriously hinder the widespread deployment of the metaverse. Fortunately, federated learning (FL) is a solution to the above problems. FL is a distributed machine learning paradigm with privacy-preserving features designed for a large number of edge devices. Federated learning for metaverse (FL4M) will be a powerful tool. Because FL allows edge devices to participate in training tasks locally using their own data, computational power, and model-building capabilities. Applying FL to the metaverse not only protects the data privacy of participants but also reduces the need for high computing power and high memory on servers. Until now, there have been many studies about FL and the metaverse, respectively. In this paper, we review some of the early advances of FL4M, which will be a research direction with unlimited development potential. We first introduce the concepts of metaverse and FL, respectively. Besides, we discuss the convergence of key metaverse technologies and FL in detail, such as big data, communication technology, the Internet of Things, edge computing, blockchain, and extended reality. Finally, we discuss some key challenges and promising directions of FL4M in detail. In summary, we hope that our up-to-date brief survey can help people better understand FL4M and build a fair, open, and secure metaverse.

James Brannick, Yao Chen, Xiaozhe Hu et al.

We design and implement a parallel algebraic multigrid method for isotropic graph Laplacian problems on multicore Graphical Processing Units (GPUs). The proposed AMG method is based on the aggregation framework. The setup phase of the algorithm uses a parallel maximal independent set algorithm in forming aggregates and the resulting coarse level hierarchy is then used in a K-cycle iteration solve phase with a $\ell^1$-Jacobi smoother. Numerical tests of a parallel implementation of the method for graphics processors are presented to demonstrate its effectiveness.

Silin Zhou, Yao Chen, Shuo Shang et al.

Trajectory representation learning (TRL) maps trajectories to vector embeddings and facilitates tasks such as trajectory classification and similarity search. State-of-the-art (SOTA) TRL methods transform raw GPS trajectories to grid or road trajectories to capture high-level travel semantics, i.e., regions and roads. However, they lose fine-grained spatial-temporal details as multiple GPS points are grouped into a single grid cell or road segment. To tackle this problem, we propose the BLUrred Encoding method, dubbed BLUE, which gradually reduces the precision of GPS coordinates to create hierarchical patches with multiple levels. The low-level patches are small and preserve fine-grained spatial-temporal details, while the high-level patches are large and capture overall travel patterns. To complement different patch levels with each other, our BLUE is an encoder-decoder model with a pyramid structure. At each patch level, a Transformer is used to learn the trajectory embedding at the current level, while pooling prepares inputs for the higher level in the encoder, and up-resolution provides guidance for the lower level in the decoder. BLUE is trained using the trajectory reconstruction task with the MSE loss. We compare BLUE with 8 SOTA TRL methods for 3 downstream tasks, the results show that BLUE consistently achieves higher accuracy than all baselines, outperforming the best-performing baselines by an average of 30.90%. Our code is available at https://github.com/slzhou-xy/BLUE.

Xiaofan Zhang, Yao Chen, Cong Hao et al.

Deep Neural Networks (DNNs) have achieved great success in a variety of machine learning (ML) applications, delivering high-quality inferencing solutions in computer vision, natural language processing, and virtual reality, etc. However, DNN-based ML applications also bring much increased computational and storage requirements, which are particularly challenging for embedded systems with limited compute/storage resources, tight power budgets, and small form factors. Challenges also come from the diverse application-specific requirements, including real-time responses, high-throughput performance, and reliable inference accuracy. To address these challenges, we introduce a series of effective design methodologies, including efficient ML model designs, customized hardware accelerator designs, and hardware/software co-design strategies to enable efficient ML applications on embedded systems.

Haifeng Wang, Hua Wu, Tian Wu et al.

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Cheng Gong, Yao Chen, Qiuyang Luo et al.

Multi-exit network is a promising architecture for efficient model inference by sharing backbone networks and weights among multiple exits. However, the gradient conflict of the shared weights results in sub-optimal accuracy. This paper introduces Deep Feature Surgery (\methodname), which consists of feature partitioning and feature referencing approaches to resolve gradient conflict issues during the training of multi-exit networks. The feature partitioning separates shared features along the depth axis among all exits to alleviate gradient conflict while simultaneously promoting joint optimization for each exit. Subsequently, feature referencing enhances multi-scale features for distinct exits across varying depths to improve the model accuracy. Furthermore, \methodname~reduces the training operations with the reduced complexity of backpropagation. Experimental results on Cifar100 and ImageNet datasets exhibit that \methodname~provides up to a \textbf{50.00\%} reduction in training time and attains up to a \textbf{6.94\%} enhancement in accuracy when contrasted with baseline methods across diverse models and tasks. Budgeted batch classification evaluation on MSDNet demonstrates that DFS uses about $\mathbf{2}\boldsymbol{\times}$ fewer average FLOPs per image to achieve the same classification accuracy as baseline methods on Cifar100. The code is available at https://github.com/GongCheng1919/dfs.

Sitian Chen, Yusen Li, Yao Chen et al.

Approximate Nearest Neighbor Search (ANNS) is a core primitive in modern AI systems, and graph-based methods currently offer the best accuracy-efficiency trade-off at scale. The workload is fundamentally memory-bound: graph traversal produces frequent, irregular memory accesses that cap CPU throughput at main-memory bandwidth, while GPUs lack the high-bandwidth memory capacity to host billion-scale indexes. Processing-in-Memory (PIM) is a natural candidate, as placing computation next to data unlocks the abundant internal bandwidth that such bandwidth-starved workloads demand. Porting graph-based ANNS to PIM, however, exposes several architectural mismatches: each processing unit has only a small local memory, inter-unit communication is costly, host coordination adds overhead, and in-memory compute units are relatively weak -- limitations that have forced prior PIM-based ANNS designs to fall back on cluster-based indexing, whose recall ceiling is far below that of graph methods. This paper presents an algorithm-architecture co-design that overcomes these obstacles through three components: a compacted index layout that shrinks the PIM-resident memory footprint by 14.5x; an asynchronous pipelined scheduler that keeps the host-to-PIM interconnect saturated; and a multiplication-free distance kernel that loses under 0.08% recall. Across three billion-scale benchmarks, the proposed design achieves up to 20x and 17.1x higher throughput than CPU and GPU baselines, respectively, outperforms prior PIM accelerators by 129x in the high-recall regime, and scales gracefully across multi-node deployments and emerging PIM architecture.

Yao Chen, Jiawei Sheng, Wenyuan Zhang et al.

The significant computational demands of large language models have increased interest in distilling reasoning abilities into smaller models via Chain-of-Thought (CoT) distillation. Current CoT distillation methods mainly focus on transferring teacher-generated rationales for complex reasoning to student models. However, they do not adequately explore teachers' dynamic attention toward critical information during reasoning. We find that language models exhibit progressive attention shifts towards key information during reasoning, which implies essential clues for drawing conclusions. Building on this observation and analysis, we introduce a novel CoT distillation framework that transfers the teacher's stepwise attention on key information to the student model. This establishes structured guidance for the student's progressive concentration on key information during reasoning. More importantly, we develop a Mixture of Layers module enabling dynamic alignment that adapts to different layers between the teacher and student. Our method achieves consistent performance improvements across multiple mathematical and commonsense reasoning datasets. To our knowledge, it is the first method to leverage stepwise attention within CoT distillation to improve small model reasoning.

Yao Chen, Junhao Pan, Xinheng Liu et al.

Quantization for CNN has shown significant progress with the intention of reducing the cost of computation and storage with low-bitwidth data representations. There are, however, no systematic studies on how an existing full-bitwidth processing unit, such as ALU in CPUs and DSP in FPGAs, can be better utilized to deliver significantly higher computation throughput for convolution under various quantized bitwidths. In this study, we propose HiKonv, a unified solution that maximizes the throughput of convolution on a given underlying processing unit with low-bitwidth quantized data inputs through novel bit-wise management and parallel computation. We establish theoretical framework and performance models using a full-bitwidth multiplier for highly parallelized low-bitwidth convolution, and demonstrate new breakthroughs for high-performance computing in this critical domain. For example, a single 32-bit processing unit in CPU can deliver 128 binarized convolution operations (multiplications and additions) and 13 4-bit convolution operations with a single multiplication instruction, and a single 27x18 multiplier in the FPGA DSP can deliver 60, 8 or 2 convolution operations with 1, 4 or 8-bit inputs in one clock cycle. We demonstrate the effectiveness of HiKonv on both CPU and FPGA. On CPU, HiKonv outperforms the baseline implementation with 1 to 8-bit inputs and provides up to 7.6x and 1.4x performance improvements for 1-D convolution, and performs 2.74x and 3.19x over the baseline implementation for 4-bit signed and unsigned data inputs for 2-D convolution. On FPGA, HiKonv solution enables a single DSP to process multiple convolutions with a shorter processing latency. For binarized input, each DSP with HiKonv is equivalent up to 76.6 LUTs. Compared to the DAC-SDC 2020 champion model, HiKonv achieves a 2.37x throughput improvement and 2.61x DSP efficiency improvement, respectively.

Yao Chen, Savannah M. Decker, Petr Bruza et al.

Accurate patient positioning is critical for precise radiotherapy dose delivery, as positioning errors can significantly affect treatment outcomes. This study introduces a novel method for tracking loco-regional tissue deformation through Cherenkov image analysis during fractionated breast cancer radiotherapy. The primary goal was to develop and test an algorithm for Cherenkov-based regional position accuracy quantification, specifically for loco-regional deformations, which lack ideal quantification methods in radiotherapy. Blood vessel detection and segmentation were developed in Cherenkov images using a tissue phantom with incremental movements, and later applied to images from fractionated whole breast radiotherapy in human patients (n=10). A combined rigid and non-rigid registration technique was used to detect inter- and intra-fractional positioning variations. This approach quantified positioning variations in two parts: a global shift from rigid registration and a two-dimensional variation map of loco-regional deformation from non-rigid registration. The methodology was validated using an anthropomorphic chest phantom experiment, where known treatment couch translations and respiratory motion were simulated to assess inter- and intra-fractional uncertainties, yielding an average accuracy of 0.83 mm for couch translations up to 20 mm. Analysis of clinical Cherenkov data from ten breast cancer patients showed an inter-fraction setup variation of 3.7 plus minus 2.4 mm relative to the first fraction and loco-regional deformations (95th percentile) of up to 3.3 plus minus 1.9 mm. This study presents a Cherenkov-based approach to quantify global and local positioning variations, demonstrating feasibility in addressing loco-regional deformations that conventional imaging techniques fail to capture.

Shiru Wang, Yao Chen, Lesley A. Jarvis et al.

Cherenkov imaging enables real-time visualization of megavoltage X-ray or electron beam delivery to the patient during Radiation Therapy (RT). Bio-morphological features, such as vasculature, seen in these images are patient-specific signatures that can be used for verification of positioning and motion management that are essential to precise RT treatment. However until now, no concerted analysis of this biological feature-based tracking was utilized because of the slow speed and accuracy of conventional image processing for feature segmentation. This study demonstrated the first deep learning framework for such an application, achieving video frame rate processing. To address the challenge of limited annotation of these features in Cherenkov images, a transfer learning strategy was applied. A fundus photography dataset including 20,529 patch retina images with ground-truth vessel annotation was used to pre-train a ResNet segmentation framework. Subsequently, a small Cherenkov dataset (1,483 images from 212 treatment fractions of 19 breast cancer patients) with known annotated vasculature masks was used to fine-tune the model for accurate segmentation prediction. This deep learning framework achieved consistent and rapid segmentation of Cherenkov-imaged bio-morphological features on another 19 patients, including subcutaneous veins, scars, and pigmented skin. Average segmentation by the model achieved Dice score of 0.85 and required less than 0.7 milliseconds processing time per instance. The model demonstrated outstanding consistency against input image variances and speed compared to conventional manual segmentation methods, laying the foundation for online segmentation in real-time monitoring in a prospective setting.

Feng Yu, Hongshi Tan, Yao Chen et al.

The widespread adoption of mixed-precision quantization in large language models (LLMs) has created demand for hardware that can efficiently perform multiply-accumulate (MAC) operations across mixed datatypes and switch datatypes at runtime. Existing FPGA-based MAC solutions fall short due to limitations in fixed-datatype design, inefficient spatial or temporal resource sharing, and poor support for mixed-precision execution. These limitations collectively lead to under-utilization of DSP resources, limiting achievable parallelism and throughput. In this work, we present XtraMAC, a novel MAC architecture that unifies integer, floating-point, and mixed-precision operations within a single, datatype-adaptive microarchitecture. XtraMAC decomposes all supported MAC formats into a shared integer mantissa product with lightweight sign and exponent handling, enabling dynamic operand packing and efficient DSP resource sharing with constant latency and initiation interval of one across all datatypes. Evaluated on an AMD Xilinx U55c FPGA, XtraMAC achieves 1.4-2.0x higher compute density, reduces per-operation LUT, FF, and DSP consumption by 27-51%, and delivers up to 1.9x greater energy efficiency and 1.2x speedup on representative mixed-precision LLM workloads. The implementation of XtraMAC is open-sourced at https://github.com/Xtra-Computing/XtraMAC.

Yao Chen, Yilong Chen, Yinqi Yang et al.

Existing approaches to increasing the effective depth of Transformers predominantly rely on parameter reuse, extending computation through recursive execution. Under this paradigm, the network structure remains static along the training timeline, and additional computational depth is uniformly assigned to entire blocks at the parameter level. This rigidity across training time and parameter space leads to substantial computational redundancy during training. In contrast, we argue that depth allocation during training should not be a static preset, but rather a progressively growing structural process. Our systematic analysis reveals a deep-to-shallow maturation trajectory across layers, where high-entropy attention heads play a crucial role in semantic integration. Motivated by this observation, we introduce the Sparse Growing Transformer (SGT). SGT is a training-time sparse depth allocation framework that progressively extends recurrence from deeper to shallower layers via targeted attention looping on informative heads. This mechanism induces structural sparsity by selectively increasing depth only for a small subset of parameters as training evolves. Extensive experiments across multiple parameter scales demonstrate that SGT consistently outperforms training-time static block-level looping baselines under comparable settings, while reducing the additional training FLOPs overhead from approximately 16--20% to only 1--3% relative to a standard Transformer backbone.

Yao Chen, Yuanhan Mo, Aimee Readie et al.

Manual annotation of vertebrae on spinal X-ray imaging is costly and time-consuming due to bone shape complexity and image quality variations. In this study, we address this challenge by proposing an ensemble method called VertXNet, to automatically segment and label vertebrae in X-ray spinal images. VertXNet combines two state-of-the-art segmentation models, namely U-Net and Mask R-CNN to improve vertebrae segmentation. A main feature of VertXNet is to also infer vertebrae labels thanks to its Mask R-CNN component (trained to detect 'reference' vertebrae) on a given spinal X-ray image. VertXNet was evaluated on an in-house dataset of lateral cervical and lumbar X-ray imaging for ankylosing spondylitis (AS) patients. Our results show that VertXNet can accurately label spinal X-rays (mean Dice of 0.9). It can be used to circumvent the lack of annotated vertebrae without requiring human expert review. This step is crucial to investigate clinical associations by solving the lack of segmentation, a common bottleneck for most computational imaging projects.

Yang Yang, Wen Wang, Liang Peng et al.

Customization generation techniques have significantly advanced the synthesis of specific concepts across varied contexts. Multi-concept customization emerges as the challenging task within this domain. Existing approaches often rely on training a fusion matrix of multiple Low-Rank Adaptations (LoRAs) to merge various concepts into a single image. However, we identify this straightforward method faces two major challenges: 1) concept confusion, where the model struggles to preserve distinct individual characteristics, and 2) concept vanishing, where the model fails to generate the intended subjects. To address these issues, we introduce LoRA-Composer, a training-free framework designed for seamlessly integrating multiple LoRAs, thereby enhancing the harmony among different concepts within generated images. LoRA-Composer addresses concept vanishing through concept injection constraints, enhancing concept visibility via an expanded cross-attention mechanism. To combat concept confusion, concept isolation constraints are introduced, refining the self-attention computation. Furthermore, latent re-initialization is proposed to effectively stimulate concept-specific latent within designated regions. Our extensive testing showcases a notable enhancement in LoRA-Composer's performance compared to standard baselines, especially when eliminating the image-based conditions like canny edge or pose estimations. Code is released at \url{https://github.com/Young98CN/LoRA_Composer}

Yao Chen, Yuanhan Mo, Aimee Readie et al.

Reliable vertebrae annotations are key to perform analysis of spinal X-ray images. However, obtaining annotation of vertebrae from those images is usually carried out manually due to its complexity (i.e. small structures with varying shape), making it a costly and tedious process. To accelerate this process, we proposed an ensemble pipeline, VertXNet, that combines two state-of-the-art (SOTA) segmentation models (respectively U-Net and Mask R-CNN) to automatically segment and label vertebrae in X-ray spinal images. Moreover, VertXNet introduces a rule-based approach that allows to robustly infer vertebrae labels (by locating the 'reference' vertebrae which are easier to segment than others) for a given spinal X-ray image. We evaluated the proposed pipeline on three spinal X-ray datasets (two internal and one publicly available), and compared against vertebrae annotated by radiologists. Our experimental results have shown that the proposed pipeline outperformed two SOTA segmentation models on our test dataset (MEASURE 1) with a mean Dice of 0.90, vs. a mean Dice of 0.73 for Mask R-CNN and 0.72 for U-Net. To further evaluate the generalization ability of VertXNet, the pre-trained pipeline was directly tested on two additional datasets (PREVENT and NHANES II) and consistent performance was observed with a mean Dice of 0.89 and 0.88, respectively. Overall, VertXNet demonstrated significantly improved performance for vertebra segmentation and labeling for spinal X-ray imaging, and evaluation on both in-house clinical trial data and publicly available data further proved its generalization.

Yuanhan Mo, Yao Chen, Aimee Readie et al.

Manually grading structural changes with the modified Stoke Ankylosing Spondylitis Spinal Score (mSASSS) on spinal X-ray imaging is costly and time-consuming due to bone shape complexity and image quality variations. In this study, we address this challenge by prototyping a 2-step auto-grading pipeline, called VertXGradeNet, to automatically predict mSASSS scores for the cervical and lumbar vertebral units (VUs) in X-ray spinal imaging. The VertXGradeNet utilizes VUs generated by our previously developed VU extraction pipeline (VertXNet) as input and predicts mSASSS based on those VUs. VertXGradeNet was evaluated on an in-house dataset of lateral cervical and lumbar X-ray images for axial spondylarthritis patients. Our results show that VertXGradeNet can predict the mSASSS score for each VU when the data is limited in quantity and imbalanced. Overall, it can achieve a balanced accuracy of 0.56 and 0.51 for 4 different mSASSS scores (i.e., a score of 0, 1, 2, 3) on two test datasets. The accuracy of the presented method shows the potential to streamline the spinal radiograph readings and therefore reduce the cost of future clinical trials.

Shuaiyi Nie, Siyu Ding, Wenyuan Zhang et al.

Large reasoning models trained with reinforcement learning and verifiable rewards (RLVR) achieve strong performance on complex reasoning tasks, yet often overthink, generating redundant reasoning without performance gains. Existing trajectory-level length penalties often fail to effectively shorten reasoning length and degrade accuracy, as they uniformly treat all reasoning steps and lack fine-grained signals to distinguish redundancy from necessity. Meanwhile, process-supervised methods are typically resource-intensive and suffer from inaccurate credit assignment. To address these issues, we propose ATTNPO, a low-overhead process-supervised RL framework that leverages the model's intrinsic attention signals for step-level credit assignment. We first identify a set of special attention heads that naturally focus on essential steps while suppressing redundant ones. By leveraging the attention scores of these heads, We then employ two sub-strategies to mitigate overthinking by discouraging redundant steps while preserving accuracy by reducing penalties on essential steps. Experimental results show that ATTNPO substantially reduces reasoning length while significantly improving performance across 9 benchmarks.

Qingyun Zou, Feng Yu, Hongshi Tan et al.

High-Level Synthesis (HLS) compiles algorithmic C/C++ descriptions into hardware, with Quality of Results (QoR) -- latency and resource utilization -- critically governed by pragma configurations and code structure. Existing LLM-based HLS approaches train for functional correctness but ignore QoR entirely. We observe that reinforcement learning (RL) for HLS does not require absolute synthesis results -- only relative comparisons between candidates. Based on this insight, we propose \textbf{HLS-Seek}, a QoR-aware NL-to-HLS framework that replaces expensive synthesis-in-the-loop RL with a comparative proxy reward model achieving 99.53\% Pareto-dominance accuracy. To prevent reward hacking, we introduce \textit{uncertainty-aware Monte Carlo (MC) dropout switching} that selectively invokes real Vitis HLS synthesis for low-confidence candidates and online updates the proxy, creating a self-improving reward system. HLS-Seek achieves 81.5\% syntax correctness pass@1 and 81.4\% Func@5 on HLS-eval with only 7B parameters, surpassing GPT-5.1 and other frontier models while achieving 8.5$\times$ faster training than real-reward RL. On QoR evaluation, HLS-Seek achieves the lowest latency on 16/30 kernels and Pareto-dominates HLS-specific baselines on 9 kernels.

Zining Zhang, Yao Chen, Bingsheng He et al.

The increasing size and complexity of Large Language Models (LLMs) pose challenges for their deployment on personal computers and mobile devices. Aggressive post-training model compression is necessary to reduce the models' size, but it often results in significant accuracy loss. To address this challenge, we propose a novel network pruning technology that utilizes over 0.7 sparsity and less than 8 bits of quantization. Our approach enables the compression of prevailing LLMs within a couple of hours while maintaining a relatively small accuracy loss. In experimental evaluations, our method demonstrates effectiveness and potential for practical deployment. By making LLMs available on domestic devices, our work can facilitate a new era of natural language processing applications with wide-ranging impacts.

Yao Chen, Jiabao Wang, Peichao Wang et al.

Low-resolution fine-grained image classification has recently made significant progress, largely thanks to the super-resolution techniques and knowledge distillation methods. However, these approaches lead to an exponential increase in the number of parameters and computational complexity of models. In order to solve this problem, in this letter, we propose a Vision Mamba Distillation (ViMD) approach to enhance the effectiveness and efficiency of low-resolution fine-grained image classification. Concretely, a lightweight super-resolution vision Mamba classification network (SRVM-Net) is proposed to improve its capability for extracting visual features by redesigning the classification sub-network with Mamba modeling. Moreover, we design a novel multi-level Mamba knowledge distillation loss boosting the performance, which can transfer prior knowledge obtained from a High-resolution Vision Mamba classification Network (HRVM-Net) as a teacher into the proposed SRVM-Net as a student. Extensive experiments on seven public fine-grained classification datasets related to benchmarks confirm our ViMD achieves a new state-of-the-art performance. While having higher accuracy, ViMD outperforms similar methods with fewer parameters and FLOPs, which is more suitable for embedded device applications. Code is available at https://github.com/boa2004plaust/ViMD.

Yibo Zhao, Liang Peng, Yang Yang et al.

Diffusion models have exhibited impressive prowess in the text-to-image task. Recent methods add image-level structure controls, e.g., edge and depth maps, to manipulate the generation process together with text prompts to obtain desired images. This controlling process is globally operated on the entire image, which limits the flexibility of control regions. In this paper, we explore a novel and practical task setting: local control. It focuses on controlling specific local region according to user-defined image conditions, while the remaining regions are only conditioned by the original text prompt. However, it is non-trivial to achieve local conditional controlling. The naive manner of directly adding local conditions may lead to the local control dominance problem, which forces the model to focus on the controlled region and neglect object generation in other regions. To mitigate this problem, we propose Regional Discriminate Loss to update the noised latents, aiming at enhanced object generation in non-control regions. Furthermore, the proposed Focused Token Response suppresses weaker attention scores which lack the strongest response to enhance object distinction and reduce duplication. Lastly, we adopt Feature Mask Constraint to reduce quality degradation in images caused by information differences across the local control region. All proposed strategies are operated at the inference stage. Extensive experiments demonstrate that our method can synthesize high-quality images aligned with the text prompt under local control conditions.

Borui Xu, Yao Chen, Zeyi Wen et al.

The increasing demand for efficient summarization tools in resource-constrained environments highlights the need for effective solutions. While large language models (LLMs) deliver superior summarization quality, their high computational resource requirements limit practical use applications. In contrast, small language models (SLMs) present a more accessible alternative, capable of real-time summarization on edge devices. However, their summarization capabilities and comparative performance against LLMs remain underexplored. This paper addresses this gap by presenting a comprehensive evaluation of 19 SLMs for news summarization across 2,000 news samples, focusing on relevance, coherence, factual consistency, and summary length. Our findings reveal significant variations in SLM performance, with top-performing models such as Phi3-Mini and Llama3.2-3B-Ins achieving results comparable to those of 70B LLMs while generating more concise summaries. Notably, SLMs are better suited for simple prompts, as overly complex prompts may lead to a decline in summary quality. Additionally, our analysis indicates that instruction tuning does not consistently enhance the news summarization capabilities of SLMs. This research not only contributes to the understanding of SLMs but also provides practical insights for researchers seeking efficient summarization solutions that balance performance and resource use.

Junyan Hu, Xue Xiao, Mengqi Zhang et al.

As large language models (LLMs) grow in size, traditional full fine-tuning becomes increasingly impractical due to its high computational and storage costs. Although popular parameter-efficient fine-tuning methods, such as LoRA, have significantly reduced the number of tunable parameters, there is still room for further optimization. In this work, we propose ASLoRA, a cross-layer parameter-sharing strategy combining global sharing with partial adaptive sharing. Specifically, we share the low-rank matrix A across all layers and adaptively merge matrix B during training. This sharing mechanism not only mitigates overfitting effectively but also captures inter-layer dependencies, significantly enhancing the model's representational capability. We conduct extensive experiments on various NLP tasks, showing that ASLoRA outperforms LoRA while using less than 25% of the parameters, highlighting its flexibility and superior parameter efficiency. Furthermore, in-depth analyses of the adaptive sharing strategy confirm its significant advantages in enhancing both model flexibility and task adaptability.

Peichao Wang, Jiabao Wang, Yao Chen et al.

The data-driven method for infrared small target detection (IRSTD) has achieved promising results. However, due to the small scale of infrared small target datasets and the limited number of pixels occupied by the targets themselves, it is a challenging task for deep learning methods to directly learn from these samples. Utilizing human expert knowledge to assist deep learning methods in better learning is worthy of exploration. To effectively guide the model to focus on targets' spatial features, this paper proposes the Local Contrast Attention Enhanced infrared small target detection Network (LCAE-Net), combining prior knowledge with data-driven deep learning methods. LCAE-Net is a U-shaped neural network model which consists of two developed modules: a Local Contrast Enhancement (LCE) module and a Channel Attention Enhancement (CAE) module. The LCE module takes advantages of prior knowledge, leveraging handcrafted convolution operator to acquire Local Contrast Attention (LCA), which could realize background suppression while enhance the potential target region, thus guiding the neural network to pay more attention to potential infrared small targets' location information. To effectively utilize the response information throughout downsampling progresses, the CAE module is proposed to achieve the information fusion among feature maps' different channels. Experimental results indicate that our LCAE-Net outperforms existing state-of-the-art methods on the three public datasets NUDT-SIRST, NUAA-SIRST, and IRSTD-1K, and its detection speed could reach up to 70 fps. Meanwhile, our model has a parameter count and Floating-Point Operations (FLOPs) of 1.945M and 4.862G respectively, which is suitable for deployment on edge devices.

Xiangzhao Hao, Zefeng Zhang, Zhenyu Zhang et al.

Image degradation from blur, noise, compression, and poor illumination severely undermines multimodal understanding in real-world settings. Unified multimodal models that combine understanding and generation within a single architecture are a natural fit for this challenge, as their generative pathway can model the fine-grained visual structure that degradation destroys. Yet these models fail to leverage their own generative capacity on degraded inputs. We trace this disconnect to two compounding factors: existing training regimes never ask the model to invoke generation during reasoning, and the standard decode-reencode pathway does not support effective joint optimization. We present CLEAR, a framework that connects the two capabilities through three progressive steps: (1) supervised fine-tuning on a degradation-aware dataset to establish the generate-then-answer reasoning pattern; (2) a Latent Representation Bridge that replaces the decode-reencode detour with a direct, optimizable connection between generation and reasoning; (3) Interleaved GRPO, a reinforcement learning method that jointly optimizes text reasoning and visual generation under answer-correctness rewards. We construct MMD-Bench, covering three degradation severity levels across six standard multimodal benchmarks. Experiments show that CLEAR substantially improves robustness on degraded inputs while preserving clean-image performance. Our analysis further reveals that removing pixel-level reconstruction supervision leads to intermediate visual states with higher perceptual quality, suggesting that task-driven optimization and visual quality are naturally aligned.

Vicky Dong, Hao Yu, Yao Chen

This study introduces a novel approach to sentence-level relation extraction (RE) that integrates Graph Neural Networks (GNNs) with Large Language Models (LLMs) to generate contextually enriched support documents. By harnessing the power of LLMs to generate auxiliary information, our approach crafts an intricate graph representation of textual data. This graph is subsequently processed through a Graph Neural Network (GNN) to refine and enrich the embeddings associated with each entity ensuring a more nuanced and interconnected understanding of the data. This methodology addresses the limitations of traditional sentence-level RE models by incorporating broader contexts and leveraging inter-entity interactions, thereby improving the model's ability to capture complex relationships across sentences. Our experiments, conducted on the CrossRE dataset, demonstrate the effectiveness of our approach, with notable improvements in performance across various domains. The results underscore the potential of combining GNNs with LLM-generated context to advance the field of relation extraction.

Esteban Real, Yao Chen, Mirko Rossini et al.

Computers calculate transcendental functions by approximating them through the composition of a few limited-precision instructions. For example, an exponential can be calculated with a Taylor series. These approximation methods were developed over the centuries by mathematicians, who emphasized the attainability of arbitrary precision. Computers, however, operate on few limited precision types, such as the popular float32. In this study, we show that when aiming for limited precision, existing approximation methods can be outperformed by programs automatically discovered from scratch by a simple evolutionary algorithm. In particular, over real numbers, our method can approximate the exponential function reaching orders of magnitude more precision for a given number of operations when compared to previous approaches. More practically, over float32 numbers and constrained to less than 1 ULP of error, the same method attains a speedup over baselines by generating code that triggers better XLA/LLVM compilation paths. In other words, in both cases, evolution searched a vast space of possible programs, without knowledge of mathematics, to discover previously unknown optimized approximations to high precision, for the first time. We also give evidence that these results extend beyond the exponential. The ubiquity of transcendental functions suggests that our method has the potential to reduce the cost of scientific computing applications.

Yao Chen, David Ohlssen, Aimee Readie et al.

Artificial intelligence (AI) holds great promise for supporting clinical trials, from patient recruitment and endpoint assessment to treatment response prediction. However, deploying AI without safeguards poses significant risks, particularly when evaluating patient endpoints that directly impact trial conclusions. We compared two AI frameworks against human-only assessment for medical image-based disease evaluation, measuring cost, accuracy, robustness, and generalization ability. To stress-test these frameworks, we injected bad models, ranging from random guesses to naive predictions, to ensure that observed treatment effects remain valid even under severe model degradation. We evaluated the frameworks using two randomized controlled trials with endpoints derived from spinal X-ray images. Our findings indicate that using AI as a supporting reader (AI-SR) is the most suitable approach for clinical trials, as it meets all criteria across various model types, even with bad models. This method consistently provides reliable disease estimation, preserves clinical trial treatment effect estimates and conclusions, and retains these advantages when applied to different populations.

Yu Zhang, Yilong Luo, Mingyuan Ma et al.

Discovering quasi-cliques -- subgraphs whose edge density exceeds a given threshold -- is a fundamental task in graph mining with applications to web spam detection, fraud screening, and e-commerce recommendation. However, existing methods for quasi-clique discovery on large-scale web graphs are often sensitive to random seeds or lack of explicit edge-density guarantees, making the task challenging in practice. This paper presents EDQC, an energy diffusion-based method for quasi-clique discovery. EDQC first employs an adaptive energy diffusion process to generate an energy ranking that highlights structurally cohesive regions. Guided by this energy ranking, the algorithm identifies a high-quality subgraph by minimizing conductance, a standard measure from community detection. This subgraph is then refined to meet the specified density threshold. Extensive experiments on 75 real-world graphs show that EDQC finds larger quasi-cliques on most datasets, with consistently lower variance across runs and competitive runtime. To the best of our knowledge, EDQC is the first method to incorporate energy diffusion into quasi-clique discovery.

Xinheng Liu, Yao Chen, Prakhar Ganesh et al.

Quantization for Convolutional Neural Network (CNN) has shown significant progress with the intention of reducing the cost of computation and storage with low-bitwidth data inputs. There are, however, no systematic studies on how an existing full-bitwidth processing unit, such as CPUs and DSPs, can be better utilized to carry out significantly higher computation throughput for convolution under various quantized bitwidths. In this study, we propose HiKonv, a unified solution that maximizes the compute throughput of a given underlying processing unit to process low-bitwidth quantized data inputs through novel bit-wise parallel computation. We establish theoretical performance bounds using a full-bitwidth multiplier for highly parallelized low-bitwidth convolution, and demonstrate new breakthroughs for high-performance computing in this critical domain. For example, a single 32-bit processing unit can deliver 128 binarized convolution operations (multiplications and additions) under one CPU instruction, and a single 27x18 DSP core can deliver eight convolution operations with 4-bit inputs in one cycle. We demonstrate the effectiveness of HiKonv on CPU and FPGA for both convolutional layers or a complete DNN model. For a convolutional layer quantized to 4-bit, HiKonv achieves a 3.17x latency improvement over the baseline implementation using C++ on CPU. Compared to the DAC-SDC 2020 champion model for FPGA, HiKonv achieves a 2.37x throughput improvement and 2.61x DSP efficiency improvement, respectively.

Prakhar Ganesh, Yao Chen, Yin Yang et al.

Performance of object detection models has been growing rapidly on two major fronts, model accuracy and efficiency. However, in order to map deep neural network (DNN) based object detection models to edge devices, one typically needs to compress such models significantly, thus compromising the model accuracy. In this paper, we propose a novel edge GPU friendly module for multi-scale feature interaction by exploiting missing combinatorial connections between various feature scales in existing state-of-the-art methods. Additionally, we propose a novel transfer learning backbone adoption inspired by the changing translational information flow across various tasks, designed to complement our feature interaction module and together improve both accuracy as well as execution speed on various edge GPU devices available in the market. For instance, YOLO-ReT with MobileNetV2x0.75 backbone runs real-time on Jetson Nano, and achieves 68.75 mAP on Pascal VOC and 34.91 mAP on COCO, beating its peers by 3.05 mAP and 0.91 mAP respectively, while executing faster by 3.05 FPS. Furthermore, introducing our multi-scale feature interaction module in YOLOv4-tiny and YOLOv4-tiny (3l) improves their performance to 41.5 and 48.1 mAP respectively on COCO, outperforming the original versions by 1.3 and 0.9 mAP.

Yao Chen, Qingyi Gao, Xiao Wang

Generative Adversarial Networks (GANs) have been impactful on many problems and applications but suffer from unstable training. The Wasserstein GAN (WGAN) leverages the Wasserstein distance to avoid the caveats in the minmax two-player training of GANs but has other defects such as mode collapse and lack of metric to detect the convergence. We introduce a novel inferential Wasserstein GAN (iWGAN) model, which is a principled framework to fuse auto-encoders and WGANs. The iWGAN model jointly learns an encoder network and a generator network motivated by the iterative primal dual optimization process. The encoder network maps the observed samples to the latent space and the generator network maps the samples from the latent space to the data space. We establish the generalization error bound of the iWGAN to theoretically justify its performance. We further provide a rigorous probabilistic interpretation of our model under the framework of maximum likelihood estimation. The iWGAN, with a clear stopping criteria, has many advantages over other autoencoder GANs. The empirical experiments show that the iWGAN greatly mitigates the symptom of mode collapse, speeds up the convergence, and is able to provide a measurement of quality check for each individual sample. We illustrate the ability of the iWGAN by obtaining competitive and stable performances for benchmark datasets.

Lingdong Kong, Prakhar Ganesh, Tan Wang et al.

We unveil a long-standing problem in the prevailing co-saliency detection systems: there is indeed inconsistency between training and testing. Constructing a high-quality co-saliency detection dataset involves time-consuming and labor-intensive pixel-level labeling, which has forced most recent works to rely instead on semantic segmentation or saliency detection datasets for training. However, the lack of proper co-saliency and the absence of multiple foreground objects in these datasets can lead to spurious variations and inherent biases learned by models. To tackle this, we introduce the idea of counterfactual training through context adjustment and propose a "cost-free" group-cut-paste (GCP) procedure to leverage off-the-shelf images and synthesize new samples. Following GCP, we collect a novel dataset called Context Adjustment Training (CAT). CAT consists of 33,500 images, which is four times larger than the current co-saliency detection datasets. All samples are automatically annotated with high-quality mask annotations, object categories, and edge maps. Extensive experiments on recent benchmarks are conducted, show that CAT can improve various state-of-the-art models by a large margin (5% ~ 25%). We hope that the scale, diversity, and quality of our dataset can benefit researchers in this area and beyond. Our dataset will be publicly accessible through our project page.

Xinheng Liu, Yao Chen, Cong Hao et al.

The combination of Winograd's algorithm and systolic array architecture has demonstrated the capability of improving DSP efficiency in accelerating convolutional neural networks (CNNs) on FPGA platforms. However, handling arbitrary convolution kernel sizes in FPGA-based Winograd processing elements and supporting efficient data access remain underexplored. In this work, we are the first to propose an optimized Winograd processing element (WinoPE), which can naturally support multiple convolution kernel sizes with the same amount of computing resources and maintains high runtime DSP efficiency. Using the proposed WinoPE, we construct a highly efficient systolic array accelerator, termed WinoCNN. We also propose a dedicated memory subsystem to optimize the data access. Based on the accelerator architecture, we build accurate resource and performance modeling to explore optimal accelerator configurations under different resource constraints. We implement our proposed accelerator on multiple FPGAs, which outperforms the state-of-the-art designs in terms of both throughput and DSP efficiency. Our implementation achieves DSP efficiency up to 1.33 GOPS/DSP and throughput up to 3.1 TOPS with the Xilinx ZCU102 FPGA. These are 29.1\% and 20.0\% better than the best solutions reported previously, respectively.

Yao Chen, Cole Hawkins, Kaiqi Zhang et al.

The deep neural network (DNN) based AI applications on the edge require both low-cost computing platforms and high-quality services. However, the limited memory, computing resources, and power budget of the edge devices constrain the effectiveness of the DNN algorithms. Developing edge-oriented AI algorithms and implementations (e.g., accelerators) is challenging. In this paper, we summarize our recent efforts for efficient on-device AI development from three aspects, including both training and inference. First, we present on-device training with ultra-low memory usage. We propose a novel rank-adaptive tensor-based tensorized neural network model, which offers orders-of-magnitude memory reduction during training. Second, we introduce an ultra-low bitwidth quantization method for DNN model compression, achieving the state-of-the-art accuracy under the same compression ratio. Third, we introduce an ultra-low latency DNN accelerator design, practicing the software/hardware co-design methodology. This paper emphasizes the importance and efficacy of training, quantization and accelerator design, and calls for more research breakthroughs in the area for AI on the edge.

Yao Chen, Jiangang Liu, Zhe Zhang et al.

In this work, we propose a novel Knowledge Graph Embedding (KGE) strategy, called MöbiusE, in which the entities and relations are embedded to the surface of a Möbius ring. The proposition of such a strategy is inspired by the classic TorusE, in which the addition of two arbitrary elements is subject to a modulus operation. In this sense, TorusE naturally guarantees the critical boundedness of embedding vectors in KGE. However, the nonlinear property of addition operation on Torus ring is uniquely derived by the modulus operation, which in some extent restricts the expressiveness of TorusE. As a further generalization of TorusE, MöbiusE also uses modulus operation to preserve the closeness of addition operation on it, but the coordinates on Möbius ring interacts with each other in the following way: {\em \color{red} any vector on the surface of a Möbius ring moves along its parametric trace will goes to the right opposite direction after a cycle}. Hence, MöbiusE assumes much more nonlinear representativeness than that of TorusE, and in turn it generates much more precise embedding results. In our experiments, MöbiusE outperforms TorusE and other classic embedding strategies in several key indicators.

Cong Hao, Yao Chen, Xiaofan Zhang et al.

High quality AI solutions require joint optimization of AI algorithms, such as deep neural networks (DNNs), and their hardware accelerators. To improve the overall solution quality as well as to boost the design productivity, efficient algorithm and accelerator co-design methodologies are indispensable. In this paper, we first discuss the motivations and challenges for the Algorithm/Accelerator co-design problem and then provide several effective solutions. Especially, we highlight three leading works of effective co-design methodologies: 1) the first simultaneous DNN/FPGA co-design method; 2) a bi-directional lightweight DNN and accelerator co-design method; 3) a differentiable and efficient DNN and accelerator co-search method. We demonstrate the effectiveness of the proposed co-design approaches using extensive experiments on both FPGAs and GPUs, with comparisons to existing works. This paper emphasizes the importance and efficacy of algorithm-accelerator co-design and calls for more research breakthroughs in this interesting and demanding area.

Cheng Gong, Yao Chen, Ye Lu et al.

Quantization has been proven to be an effective method for reducing the computing and/or storage cost of DNNs. However, the trade-off between the quantization bitwidth and final accuracy is complex and non-convex, which makes it difficult to be optimized directly. Minimizing direct quantization loss (DQL) of the coefficient data is an effective local optimization method, but previous works often neglect the accurate control of the DQL, resulting in a higher loss of the final DNN model accuracy. In this paper, we propose a novel metric called Vector Loss. Based on this new metric, we develop a new quantization solution called VecQ, which can guarantee minimal direct quantization loss and better model accuracy. In addition, in order to speed up the proposed quantization process during model training, we accelerate the quantization process with a parameterized probability estimation method and template-based derivation calculation. We evaluate our proposed algorithm on MNIST, CIFAR, ImageNet, IMDB movie review and THUCNews text data sets with numerical DNN models. The results demonstrate that our proposed quantization solution is more accurate and effective than the state-of-the-art approaches yet with more flexible bitwidth support. Moreover, the evaluation of our quantized models on Saliency Object Detection (SOD) tasks maintains comparable feature extraction quality with up to 16$\times$ weight size reduction.

Ruichu Cai, Zhihao Liang, Boyan Xu et al.

Existing leading code comment generation approaches with the structure-to-sequence framework ignores the type information of the interpretation of the code, e.g., operator, string, etc. However, introducing the type information into the existing framework is non-trivial due to the hierarchical dependence among the type information. In order to address the issues above, we propose a Type Auxiliary Guiding encoder-decoder framework for the code comment generation task which considers the source code as an N-ary tree with type information associated with each node. Specifically, our framework is featured with a Type-associated Encoder and a Type-restricted Decoder which enables adaptive summarization of the source code. We further propose a hierarchical reinforcement learning method to resolve the training difficulties of our proposed framework. Extensive evaluations demonstrate the state-of-the-art performance of our framework with both the auto-evaluated metrics and case studies.

Yuhong Li, Cong Hao, Xiaofan Zhang et al.

High quality AI solutions require joint optimization of AI algorithms and their hardware implementations. In this work, we are the first to propose a fully simultaneous, efficient differentiable DNN architecture and implementation co-search (EDD) methodology. We formulate the co-search problem by fusing DNN search variables and hardware implementation variables into one solution space, and maximize both algorithm accuracy and hardware implementation quality. The formulation is differentiable with respect to the fused variables, so that gradient descent algorithm can be applied to greatly reduce the search time. The formulation is also applicable for various devices with different objectives. In the experiments, we demonstrate the effectiveness of our EDD methodology by searching for three representative DNNs, targeting low-latency GPU implementation and FPGA implementations with both recursive and pipelined architectures. Each model produced by EDD achieves similar accuracy as the best existing DNN models searched by neural architecture search (NAS) methods on ImageNet, but with superior performance obtained within 12 GPU-hour searches. Our DNN targeting GPU is 1.40x faster than the state-of-the-art solution reported in Proxyless, and our DNN targeting FPGA delivers 1.45x higher throughput than the state-of-the-art solution reported in DNNBuilder.

Sicong Du, Hengkai Guo, Yao Chen et al.

Initialization is essential to monocular Simultaneous Localization and Mapping (SLAM) problems. This paper focuses on a novel initialization method for monocular SLAM based on planar features. The algorithm starts by homography estimation in a sliding window. It then proceeds to a global plane optimization (GPO) to obtain camera poses and the plane normal. 3D points can be recovered using planar constraints without triangulation. The proposed method fully exploits the plane information from multiple frames and avoids the ambiguities in homography decomposition. We validate our algorithm on the collected chessboard dataset against baseline implementations and present extensive analysis. Experimental results show that our method outperforms the fine-tuned baselines in both accuracy and real-time.

Prakhar Ganesh, Yao Chen, Xin Lou et al.

Pre-trained Transformer-based models have achieved state-of-the-art performance for various Natural Language Processing (NLP) tasks. However, these models often have billions of parameters, and, thus, are too resource-hungry and computation-intensive to suit low-capability devices or applications with strict latency requirements. One potential remedy for this is model compression, which has attracted a lot of research attention. Here, we summarize the research in compressing Transformers, focusing on the especially popular BERT model. In particular, we survey the state of the art in compression for BERT, we clarify the current best practices for compressing large-scale Transformer models, and we provide insights into the workings of various methods. Our categorization and analysis also shed light on promising future research directions for achieving lightweight, accurate, and generic NLP models.

Cong Hao, Yao Chen, Xinheng Liu et al.

The rapidly growing demands for powerful AI algorithms in many application domains have motivated massive investment in both high-quality deep neural network (DNN) models and high-efficiency implementations. In this position paper, we argue that a simultaneous DNN/implementation co-design methodology, named Neural Architecture and Implementation Search (NAIS), deserves more research attention to boost the development productivity and efficiency of both DNN models and implementation optimization. We propose a stylized design methodology that can drastically cut down the search cost while preserving the quality of the end solution.As an illustration, we discuss this DNN/implementation methodology in the context of both FPGAs and GPUs. We take autonomous driving as a key use case as it is one of the most demanding areas for high quality AI algorithms and accelerators. We discuss how such a co-design methodology can impact the autonomous driving industry significantly. We identify several research opportunities in this exciting domain.