arXiv subjects

Junjie Tang

Publications and source records attributed to Junjie Tang.

9 recordsLinked to original sources

arXiv2026

Video-Text Temporal Localization via Multi-Scale Convolution and Dynamic Routing

Video-text temporal localization requires precise alignment between natural language queries and corresponding video segments, a fundamental challenge in multimodal understanding. We present a novel framework that addresses two critical limitations of existing methods: inadequate modeling of hierarchical temporal structure and inability to handle complex many-to-many correspondences between modalities. Our approach introduces a multi-scale temporal convolutional encoder that captures motion patterns across different temporal granularities - from instantaneous frame transitions to extended action sequences. We further propose a capsule-based dynamic routing mechanism that iteratively refines segment-query associations through structured agreement updates, enabling flexible modeling of non-monotonic alignments. These components are unified through a multi-task learning objective that jointly optimizes temporal boundary regression, cross-modal semantic alignment, and capsule diversity. Extensive experiments on ActivityNet Captions demonstrate significant improvements, achieving 42.9% Recall@0.5 and 41.1% mean IoU, surpassing strong transformer-based baselines while maintaining computational efficiency. Our results validate that combining hierarchical temporal modeling with structured semantic routing provides an effective solution for fine-grained video-language understanding.

cs.CV↗

arXiv2026

NKI-Agent: Domain-Specific Fine-Tuning and Agentic Tool Use for Neuron Kernel Generation

Recent agentic approaches to LLM-based kernel generation have achieved impressive results on CUDA. For emerging AI accelerators such as AWS Trainium and Inferentia, automated kernel generation and optimization remain largely unaddressed. Writing kernels for these chips via the Neuron Kernel Interface (NKI) is particularly challenging: developers must navigate a multi-engine architecture, tile-based programming, and explicit data movement across multi-level memory hierarchy. Moreover, no publicly-available training data, benchmarks, or tool-augmented agents exist for this domain. We introduce NKI-Agent, the first system combining domain-specific supervised fine-tuning (SFT) with a compile-verify-fix agent loop for NKI kernel generation. We adapt the existing CUDA-Agent framework to Neuron hardware, curate 6,000 NKI kernel generation tasks for training, and construct NKIBench, a 250-task benchmark across three difficulty levels. Evaluated on real Trn1 hardware, NKI-Agent with Claude Opus 4.8 and a rank-aware system prompt achieves a 77.3% pass rate on the 150-task NKIBench. We show that tool use is critical: Opus 4.8 scores 6% in single-shot mode without agent tools. On a 60-task subset, we show that an SFT-trained Qwen3-Coder-30B-A3B achieves 25.0% pass rate at 1/100th the cost, outperforming Claude Sonnet 4 (15.0%). We also report that Group Relative Policy Optimization (GRPO) with binary compilation reward fails to improve over SFT, providing guidance on reward design for RL-based kernel generation.

cs.LG↗

arXiv2026

Legendre Memory Unit with A Multi-Slice Compensation Model for Short-Term Wind Speed Forecasting Based on Wind Farm Cluster Data

With more wind farms clustered for integration, the short-term wind speed prediction of such wind farm clusters is critical for normal operation of power systems. This paper focuses on achieving accurate, fast, and robust wind speed prediction by full use of cluster data with spatial-temporal correlation. First, weighted mean filtering (WMF) is applied to denoise wind speed data at the single-farm level. The Legendre memory unit (LMU) is then innovatively applied for the wind speed prediction, in combination with the Compensating Parameter based on Kendall rank correlation coefficient (CPK) of wind farm cluster data, to construct the multi-slice LMU (MSLMU). Finally, an innovative ensemble model WMF-CPK-MSLMU is proposed herein, with three key blocks: data pre-processing, forecasting, and multi-slice compensation. Advantages include: 1) LMU jointly models linear and nonlinear dependencies among farms to capture spatial-temporal correlations through backpropagation; 2) MSLMU enhances forecasting by using CPK-derived weights instead of random initialization, allowing spatial correlations to fully activate hidden nodes across clustered wind farms.; 3) CPK adaptively weights the compensation model in MSLMU and complements missing data spatially, to facilitate the whole model highly accurate and robust. Test results on different wind farm clusters indicate the effectiveness and superiority of proposed ensemble model WMF-CPK-MSLMU in the short-term prediction of wind farm clusters compared to the existing models.

cs.LG↗

arXiv2023

FPTN: Fast Pure Transformer Network for Traffic Flow Forecasting

Traffic flow forecasting is challenging due to the intricate spatio-temporal correlations in traffic flow data. Existing Transformer-based methods usually treat traffic flow forecasting as multivariate time series (MTS) forecasting. However, too many sensors can cause a vector with a dimension greater than 800, which is difficult to process without information loss. In addition, these methods design complex mechanisms to capture spatial dependencies in MTS, resulting in slow forecasting speed. To solve the abovementioned problems, we propose a Fast Pure Transformer Network (FPTN) in this paper. First, the traffic flow data are divided into sequences along the sensor dimension instead of the time dimension. Then, to adequately represent complex spatio-temporal correlations, Three types of embeddings are proposed for projecting these vectors into a suitable vector space. After that, to capture the complex spatio-temporal correlations simultaneously in these vectors, we utilize Transformer encoder and stack it with several layers. Extensive experiments are conducted with 4 real-world datasets and 13 baselines, which demonstrate that FPTN outperforms the state-of-the-art on two metrics. Meanwhile, the computational time of FPTN spent is less than a quarter of other state-of-the-art Transformer-based models spent, and the requirements for computing resources are significantly reduced.

cs.LG↗

arXiv2022

Network Analysis of Count Data from Mixed Populations

In applications such as gene regulatory network analysis based on single-cell RNA sequencing data, samples often come from a mixture of different populations and each population has its own unique network. Available graphical models often assume that all samples are from the same population and share the same network. One has to first cluster the samples and use available methods to infer the network for every cluster separately. However, this two-step procedure ignores uncertainty in the clustering step and thus could lead to inaccurate network estimation. Motivated by these applications, we consider the mixture Poisson log-normal model for network inference of count data from mixed populations. The latent precision matrices of the mixture model correspond to the networks of different populations and can be jointly estimated by maximizing the lasso-penalized log-likelihood. Under rather mild conditions, we show that the mixture Poisson log-normal model is identifiable and has the positive definite Fisher information matrix. Consistency of the maximum lasso-penalized log-likelihood estimator is also established. To avoid the intractable optimization of the log-likelihood, we develop an algorithm called VMPLN based on the variational inference method. Comprehensive simulation and real single-cell RNA sequencing data analyses demonstrate the superior performance of VMPLN.

stat.ME↗

arXiv2022

Single-cell gene regulatory network analysis for mixed cell populations with applications to COVID-19 single cell data

Gene regulatory network (GRN) refers to the complex network formed by regulatory interactions between genes in living cells. In this paper, we consider inferring GRNs in single cells based on single cell RNA sequencing (scRNA-seq) data. In scRNA-seq, single cells are often profiled from mixed populations and their cell identities are unknown. A common practice for single cell GRN analysis is to first cluster the cells and infer GRNs for every cluster separately. However, this two-step procedure ignores uncertainty in the clustering step and thus could lead to inaccurate estimation of the networks. To address this problem, we propose to model scRNA-seq by the mixture multivariate Poisson log-normal (MPLN) distribution. The precision matrices of the MPLN are the GRNs of different cell types and can be jointly estimated by maximizing MPLN's lasso-penalized log-likelihood. We show that the MPLN model is identifiable and the resulting penalized log-likelihood estimator is consistent. To avoid the intractable optimization of the MPLN's log-likelihood, we develop an algorithm called VMPLN based on the variational inference method. Comprehensive simulation and real scRNA-seq data analyses reveal that VMPLN performs better than the state-of-the-art single cell GRN methods.

q-bio.MN↗

arXiv2021

Gene regulatory network in single cells based on the Poisson log-normal model

Gene regulatory network inference is crucial for understanding the complex molecular interactions in various genetic and environmental conditions. The rapid development of single-cell RNA sequencing (scRNA-seq) technologies unprecedentedly enables gene regulatory networks inference at the single cell resolution. However, traditional graphical models for continuous data, such as Gaussian graphical models, are inappropriate for network inference of scRNA-seq's count data. Here, we model the scRNA-seq data using the multivariate Poisson log-normal (PLN) distribution and represent the precision matrix of the latent normal distribution as the regulatory network. We propose to first estimate the latent covariance matrix using a moment estimator and then estimate the precision matrix by minimizing the lasso-penalized D-trace loss function. We establish the convergence rate of the covariance matrix estimator and further establish the convergence rates and the sign consistency of the proposed PLNet estimator of the precision matrix in the high dimensional setting. The performance of PLNet is evaluated and compared with available methods using simulation and gene regulatory network analysis of scRNA-seq data.

stat.ME↗

arXiv2021

Denoising User-aware Memory Network for Recommendation

For better user satisfaction and business effectiveness, more and more attention has been paid to the sequence-based recommendation system, which is used to infer the evolution of users' dynamic preferences, and recent studies have noticed that the evolution of users' preferences can be better understood from the implicit and explicit feedback sequences. However, most of the existing recommendation techniques do not consider the noise contained in implicit feedback, which will lead to the biased representation of user interest and a suboptimal recommendation performance. Meanwhile, the existing methods utilize item sequence for capturing the evolution of user interest. The performance of these methods is limited by the length of the sequence, and can not effectively model the long-term interest in a long period of time. Based on this observation, we propose a novel CTR model named denoising user-aware memory network (DUMN). Specifically, the framework: (i) proposes a feature purification module based on orthogonal mapping, which use the representation of explicit feedback to purify the representation of implicit feedback, and effectively denoise the implicit feedback; (ii) designs a user memory network to model the long-term interests in a fine-grained way by improving the memory network, which is ignored by the existing methods; and (iii) develops a preference-aware interactive representation component to fuse the long-term and short-term interests of users based on gating to understand the evolution of unbiased preferences of users. Extensive experiments on two real e-commerce user behavior datasets show that DUMN has a significant improvement over the state-of-the-art baselines. The code of DUMN model has been uploaded as an additional material.

cs.IR↗

arXiv2020

LRC-BERT: Latent-representation Contrastive Knowledge Distillation for Natural Language Understanding

The pre-training models such as BERT have achieved great results in various natural language processing problems. However, a large number of parameters need significant amounts of memory and the consumption of inference time, which makes it difficult to deploy them on edge devices. In this work, we propose a knowledge distillation method LRC-BERT based on contrastive learning to fit the output of the intermediate layer from the angular distance aspect, which is not considered by the existing distillation methods. Furthermore, we introduce a gradient perturbation-based training architecture in the training phase to increase the robustness of LRC-BERT, which is the first attempt in knowledge distillation. Additionally, in order to better capture the distribution characteristics of the intermediate layer, we design a two-stage training method for the total distillation loss. Finally, by verifying 8 datasets on the General Language Understanding Evaluation (GLUE) benchmark, the performance of the proposed LRC-BERT exceeds the existing state-of-the-art methods, which proves the effectiveness of our method.

cs.CL↗