Tag
TensorSSM proposes an Implicit Neural Representation for functional tensor decomposition that memorizes spatial positions in a dynamic State Space Model, encoding coarse-to-fine spatial tokens and cross-mode dependencies to improve reconstruction accuracy and storage efficiency across diverse tensor datasets.
提出LOTUS框架，通过低秩多线性分解从LLM隐藏状态中提取紧凑表示以检测幻觉，在多个LLM和问答基准上超越现有最先进方法。
提出 Tucker Bottleneck Attention（TuBA），利用低秩张量结构实现多维序列的高效全局 token 混合，在视频预测和全球天气预报上以显著更低的计算成本取得更优精度，最高减少 85.1% 计算量并实现 4.27 倍加速。
This paper introduces a novel semi-tensor product for tensors and develops a multi-term randomized tensor singular value decomposition (MSTP-SVD) to improve low-rank approximation accuracy and computational efficiency for visual data processing tasks.
This paper introduces MINT, a method that stacks recurrence (self-similarity) matrices of multiple time series into a tensor and applies tensor decomposition to mine co-clustered cross-sensor patterns. Experiments on transit, electricity, wind turbine, and traffic data show effective co-clustering of motifs in regular time series.
Proposes Hybrid Latent-Structural Fusion (HLSF), a weighted anomaly fusion framework combining CP-APR structural anomaly scores with latent-space density scores from normalizing flows, improving cyber anomaly detection on real-world compromised user credentials data.
Introduces (MPO)², a framework combining learned matrix product operator feature embeddings with compact polynomial weight tensors for efficient multivariate polynomial optimization, achieving improved performance over existing tensor decomposition based polynomial models.
This paper proposes FaStR, a method that factorizes the transition kernel in reinforcement learning using CP decomposition into separate state, action, and next-state encoders, improving sample efficiency especially in high-dimensional locomotion tasks.
This paper introduces low-cost High-Order Singular Value Decomposition (lcHOSVD), a tensor-based method for reconstructing high-dimensional environmental fields from sparse sensor measurements. Applied to urban flow and air-quality datasets, it achieves lower reconstruction errors and greater robustness to uneven sensor distributions compared to matrix-based approaches.
This paper introduces Separable Neural Architecture (SNA), a function class that combines neural approximation with tensor decomposition to efficiently solve parametric PDEs. The method achieves dramatic speedups (up to 150,000×) over traditional grid-based methods in engineering applications like laser powder bed fusion and material property prediction.
This paper introduces a distributional generalization of matrix completion where each entry is a probability distribution rather than a scalar, using kernel mean embeddings and Tucker rank to capture low-rank structure. The authors propose a novel estimator with non-asymptotic error bounds and demonstrate effectiveness on synthetic and real-world data.
This paper presents an online framework for modeling streaming time series as dynamic mixtures of time-delay systems, addressing regime shifts and memory constraints via a summary system tensor and tensor decomposition.