Guangli Li

Shuoming Zhang, Qiuchu Yu, Yangyu Zhang et al.

LLM-based agents are increasingly used to generate GPU kernels, but they often know what optimizations to try without knowing when those optimizations are sound. We introduce KLineage, which learns this missing "when" knowledge from expert kernels: instead of relying on forward rollouts, KLineage walks expert implementations backward through validation-gated simplifications and reverses each accepted step into a reusable optimization skill. Each skill records not only the optimization intent, but also where it applies in code, what conditions made it valid, what effect it had, and what failures its assumptions avoid. A downstream LLM materializes these skills on new code surfaces under the same compile/correctness/profile gate. On five expert workloads across two NVIDIA architectures, these lineage-derived skills serve as an effective optimization curriculum, exceeding recent memory-based LLM-kernel baselines in both final kernel quality and optimization efficiency under the same fixed budget. We additionally use a separate 22-instance held-out check as a sanity test against source-case memorization.

Guangli Li, Canbiao Wu, Na Tian et al.

Electroencephalography (EEG)-based emotion recognition suffers from severe performance degradation when models are transferred across heterogeneous datasets due to physiological variability, experimental paradigm differences, and device inconsistencies. Existing domain adversarial methods primarily enforce global marginal alignment and often overlook class-conditional mismatch and decision boundary distortion, limiting cross-corpus generalization. In this work, we propose a unified Prototype-driven Adversarial Alignment (PAA) framework for cross-corpus EEG emotion recognition. The framework is progressively instantiated in three configurations: PAA-L, which performs prototype-guided local class-conditional alignment; PAA-C, which further incorporates contrastive semantic regularization to enhance intra-class compactness and inter-class separability; and PAA-M, the full boundary-aware configuration that integrates dual relation-aware classifiers within a three-stage adversarial optimization scheme to explicitly refine controversial samples near decision boundaries. By combining prototype-guided subdomain alignment, contrastive discriminative enhancement, and boundary-aware aggregation within a coherent adversarial architecture, the proposed framework reformulates emotion recognition as a relation-driven representation learning problem, reducing sensitivity to label noise and improving cross-domain stability. Extensive experiments on SEED, SEED-IV, and SEED-V demonstrate state-of-the-art performance under four cross-corpus evaluation protocols, with average improvements of 6.72\%, 5.59\%, 6.69\%, and 4.83\%, respectively. Furthermore, the proposed framework generalizes effectively to clinical depression identification scenarios, validating its robustness in real-world heterogeneous settings. The source code is available at \textit{https://github.com/WuCB-BCI/PAA}

Chenyang Song, Xu Han, Zhengyan Zhang et al.

Activation sparsity refers to the existence of considerable weakly-contributed elements among activation outputs. As a prevalent property of the models using the ReLU activation function, activation sparsity has been proven a promising paradigm to boost model inference efficiency. Nevertheless, most large language models (LLMs) adopt activation functions without intrinsic activation sparsity (e.g., GELU and Swish). Some recent efforts have explored introducing ReLU or its variants as the substitutive activation function to help LLMs achieve activation sparsity and inference acceleration, but few can simultaneously obtain high sparsity and comparable model performance. This paper introduces a simple and effective sparsification method named "ProSparse" to push LLMs for higher activation sparsity while maintaining comparable performance. Specifically, after substituting the activation function of LLMs with ReLU, ProSparse adopts progressive sparsity regularization with a factor smoothly increasing along the multi-stage sine curves. This can enhance activation sparsity and mitigate performance degradation by avoiding radical shifts in activation distributions. With ProSparse, we obtain high sparsity of 89.32% for LLaMA2-7B, 88.80% for LLaMA2-13B, and 87.89% for end-size MiniCPM-1B, respectively, achieving comparable performance to their original Swish-activated versions. These present the most sparsely activated models among open-source LLaMA versions and competitive end-size models, considerably surpassing ReluLLaMA-7B (66.98%) and ReluLLaMA-13B (71.56%). Our inference acceleration experiments further demonstrate the significant practical acceleration potential of LLMs with higher activation sparsity, obtaining up to 4.52$\times$ inference speedup.

Changxin Ke, Rui Zhang, Shuo Wang et al.

The rise of GPU-based high-performance computing (HPC) has driven the widespread adoption of parallel programming models such as CUDA. Yet, the inherent complexity of parallel programming creates a demand for the automated sequential-to-parallel approaches. However, data scarcity poses a significant challenge for machine learning-based sequential-to-parallel code translation. Although recent back-translation methods show promise, they still fail to ensure functional equivalence in the translated code. In this paper, we propose \textbf{QiMeng-MuPa}, a novel \textbf{Mu}tual-Supervised Learning framework for Sequential-to-\textbf{Pa}rallel code translation, to address the functional equivalence issue. QiMeng-MuPa consists of two models, a Translator and a Tester. Through an iterative loop consisting of Co-verify and Co-evolve steps, the Translator and the Tester mutually generate data for each other and improve collectively. The Tester generates unit tests to verify and filter functionally equivalent translated code, thereby evolving the Translator, while the Translator generates translated code as augmented input to evolve the Tester. Experimental results demonstrate that QiMeng-MuPa significantly enhances the performance of the base models: when applied to Qwen2.5-Coder, it not only improves Pass@1 by up to 28.91% and boosts Tester performance by 68.90%, but also outperforms the previous state-of-the-art method CodeRosetta by 1.56 and 6.92 in BLEU and CodeBLEU scores, while achieving performance comparable to DeepSeek-R1 and GPT-4.1. Our code is available at https://github.com/kcxain/mupa.

Guangli Li, Canbiao Wu, Zhehao Zhou et al.

Emotion recognition based on electroencephalography (EEG) signals is increasingly becoming a key research hotspot in affective Brain-Computer Interfaces (aBCIs). However, the current transfer learning model greatly depends on the source domain and target domain data, which hinder the practical application of emotion recognition. Therefore, we propose a Multi-domain Aggregation Transfer Learning framework for EEG emotion recognition with Domain-Class prototype under unseen targets (MATL-DC). We design the feature decoupling module to decouple class-invariant domain features from domain-invariant class features from shallow features. In the model training stage, the multi-domain aggregation mechanism aggregates the domain feature space to form a superdomain, which enhances the characteristics of emotional EEG signals. In each superdomain, we further extract the class prototype representation by class features. In addition, we adopt the pairwise learning strategy to transform the sample classification problem into the similarity problem between sample pairs, which effectively alleviates the influence of label noise. It is worth noting that the target domain is completely unseen during the training process. In the inference stage, we use the trained domain-class prototypes for inference, and then realize emotion recognition. We rigorously validate it on the publicly available databases (SEED, SEED-IV and SEED-V). The results show that the accuracy of MATL-DC model is 84.70\%, 68.11\% and 61.08\%, respectively. MATL-DC achieves comparable or even better performance than methods that rely on both source and target domains. The source code is available at https://github.com/WuCB-BCI/MATL-DC.

Guangli Li, Canbiao Wu, Zhen Liang

Affective computing is a rapidly developing interdisciplinary research direction in the field of brain-computer interface. In recent years, the introduction of deep learning technology has greatly promoted the development of the field of emotion recognition. However, due to physiological differences between subjects, as well as the variations in experimental environments and equipment, cross-corpus emotion recognition faces serious challenges, especially for samples near the decision boundary. To solve the above problems, we propose an optimization method based on domain adversarial transfer learning to fine-grained alignment of affective features, named Maximum classifier discrepancy with Pairwise Learning (McdPL) framework. In McdPL, we design a dual adversarial classifier (Ada classifier and RMS classifier), and apply a three-stage adversarial training to maximize classification discrepancy and minimize feature distribution to align controversy samples near the decision boundary. In the process of domain adversarial training, the two classifiers also maintain an adversarial relationship, ultimately enabling precise cross-corpus feature alignment. In addition, the introduction of pairwise learning transforms the classification problem of samples into a similarity problem between samples, alleviating the influence of label noise. We conducted systematic experimental evaluation of the model using publicly available SEED, SEED-IV and SEED-V databases. The results show that the McdPL model is superior to other baseline models in the cross-corpus emotion recognition task, and the average accuracy improvements of 4.76\% and 3.97\%, respectively. Our work provides a promising solution for emotion recognition cross-corpus. The source code is available at https://github.com/WuCB-BCI/Mcd_PL.

Guangli Li, Canbiao Wu, Zhehao Zhou et al.

Electroencephalogram (EEG) signals serve as a powerful tool in affective Brain-Computer Interfaces (aBCIs) and play a crucial role in affective computing. In recent years, the introduction of deep learning techniques has significantly advanced the development of aBCIs. However, the current emotion recognition methods based on deep transfer learning face the challenge of the dual dependence of the model on source domain and target domain, As well as being affected by label noise, which seriously affects the performance and generalization ability of the model. To overcome this limitation, we proposes a Pairwise Learning framework with Domain and Category Prototypes for EEG emotion recognition under unseen target conditions (PL-DCP), and integrating concepts of feature disentanglement and prototype inference. Here, the feature disentanglement module extracts and decouples the emotional EEG features to form domain features and class features, and further calculates the dual prototype representation. The Domain-pprototype captures the individual variations across subjects, while the class-prototype captures the cross-individual commonality of emotion categories. In addition, the pairwise learning strategy effectively reduces the noise effect caused by wrong labels. The PL-DCP framework conducts a systematic experimental evaluation on the published datasets SEED, SEED-IV and SEED-V, and the accuracy are 82.88\%, 65.15\% and 61.29\%, respectively. The results show that compared with other State-of-the-Art(SOTA) Methods, the PL-DCP model still achieves slightly better performance than the deep transfer learning method that requires both source and target data, although the target domain is completely unseen during the training. This work provides an effective and robust potential solution for emotion recognition. The source code is available at https://github.com/WuCB-BCI/PL_DCP.

Jiansong Li, Xiao Dong, Guangli Li et al.

The training of deep neural networks (DNNs) is usually memory-hungry due to the limited device memory capacity of DNN accelerators. Characterizing the memory behaviors of DNN training is critical to optimize the device memory pressures. In this work, we pinpoint the memory behaviors of each device memory block of GPU during training by instrumenting the memory allocators of the runtime system. Our results show that the memory access patterns of device memory blocks are stable and follow an iterative fashion. These observations are useful for the future optimization of memory-efficient training from the perspective of raw memory access patterns.

Guangli Li, Xiu Ma, Xueying Wang et al.

The increasing computational cost of deep neural network models limits the applicability of intelligent applications on resource-constrained edge devices. While a number of neural network pruning methods have been proposed to compress the models, prevailing approaches focus only on parametric operators (e.g., convolution), which may miss optimization opportunities. In this paper, we present a novel fusion-catalyzed pruning approach, called FuPruner, which simultaneously optimizes the parametric and non-parametric operators for accelerating neural networks. We introduce an aggressive fusion method to equivalently transform a model, which extends the optimization space of pruning and enables non-parametric operators to be pruned in a similar manner as parametric operators, and a dynamic filter pruning method is applied to decrease the computational cost of models while retaining the accuracy requirement. Moreover, FuPruner provides configurable optimization options for controlling fusion and pruning, allowing much more flexible performance-accuracy trade-offs to be made. Evaluation with state-of-the-art residual neural networks on five representative intelligent edge platforms, Jetson TX2, Jetson Nano, Edge TPU, NCS, and NCS2, demonstrates the effectiveness of our approach, which can accelerate the inference of models on CIFAR-10 and ImageNet datasets.

Guangli Li, Lei Liu, Xueying Wang et al.

Accelerating deep convolutional neural networks has become an active topic and sparked an interest in academia and industry. In this paper, we propose an efficient low-precision quantized Winograd convolution algorithm, called LANCE, which combines the advantages of fast convolution and quantization techniques. By embedding linear quantization operations into the Winograd-domain, the fast convolution can be performed efficiently under low-precision computation on graphics processing units. We test neural network models with LANCE on representative image classification datasets, including SVHN, CIFAR, and ImageNet. The experimental results show that our 8-bit quantized Winograd convolution improves the performance by up to 2.40x over the full-precision convolution with trivial accuracy loss.

Xueying Wang, Lei Liu, Guangli Li et al.

Background subtraction is a significant component of computer vision systems. It is widely used in video surveillance, object tracking, anomaly detection, etc. A new data source for background subtraction appeared as the emergence of low-cost depth sensors like Microsof t Kinect, Asus Xtion PRO, etc. In this paper, we propose a background subtraction approach on depth videos, which is based on convolutional neural networks (CNNs), called BGSNet-D (BackGround Subtraction neural Networks for Depth videos). The method can be used in color unavailable scenarios like poor lighting situations, and can also be applied to combine with existing RGB background subtraction methods. A preprocessing strategy is designed to reduce the influences incurred by noise from depth sensors. The experimental results on the SBM-RGBD dataset show that the proposed method outperforms existing methods on depth data.

Guangli Li, Lei Liu, Xueying Wang et al.

Recently, deep neural networks (DNNs) have been widely applied in mobile intelligent applications. The inference for the DNNs is usually performed in the cloud. However, it leads to a large overhead of transmitting data via wireless network. In this paper, we demonstrate the advantages of the cloud-edge collaborative inference with quantization. By analyzing the characteristics of layers in DNNs, an auto-tuning neural network quantization framework for collaborative inference is proposed. We study the effectiveness of mixed-precision collaborative inference of state-of-the-art DNNs by using ImageNet dataset. The experimental results show that our framework can generate reasonable network partitions and reduce the storage on mobile devices with trivial loss of accuracy.