REFINE: Trajectory Representation Learning via Closed-Loop Transcription
针对轨迹表示学习中现有方法的局限性,提出REFINE框架,通过闭环转录精炼结合反馈控制理论,提升模型对局部和全局时空依赖性的捕捉能力。
针对轨迹表示学习中现有方法的局限性,提出REFINE框架,通过闭环转录精炼结合反馈控制理论,提升模型对局部和全局时空依赖性的捕捉能力。
Catastrophic forgetting severely hinders long-term adaptability in continual learning. This paper proposes the first Mamba-based, forgetfulness-free fine-tuning framework for class-incremental continual learning. Our method maps historical task features into a subspace and applies orthogonal parameter updates within its nullspace—thereby preserving output consistency of the State Space Model (SSM) core across tasks. We theoretically derive and enforce consistency constraints on four time-invariant SSM parameters, simplifying both the recurrent structure and discretization procedure. Crucially, this work introduces nullspace projection to the Mamba architecture for the first time, enabling efficient, replay-free, and regularization-free continual learning. Evaluated on four standard class-incremental benchmarks, our approach consistently outperforms state-of-the-art methods. The implementation is publicly available.
This work addresses beamforming design for joint target sensing and active physical-layer eavesdropping (JTSAPE) systems, where a shared waveform at the base station simultaneously enables radar-like target parameter estimation, conveys information to the legitimate receiver, and acts as artificial noise to jam the illegitimate receiver—thereby enhancing eavesdropping performance. We propose the first normalized weighted framework jointly optimizing sensing accuracy (by minimizing the Cramér–Rao bound) and eavesdropping efficacy (by maximizing the eavesdropping signal-to-interference-plus-noise ratio). To tackle the resulting non-convex optimization under strong eavesdropper channels, we develop a stepwise iterative algorithm based on sequential rank-one constraint relaxation (SROCR). Simulation results demonstrate that the proposed method significantly improves both SINR and estimation accuracy in multi-target and time-varying channel scenarios, yielding high-quality suboptimal beam covariance solutions with strong robustness and practical applicability.
本文提出一种自适应皮层约束的EEG-视觉对齐方法,通过重建EEG响应并采用基于证据的自适应视觉监督策略,解决零样本脑-图像检索中信号噪声和响应变异性问题。
本文提出NeuroGlyph方法,通过学习视觉层次中不同深度的信息来优化脑-图像检索目标,超越了仅使用最终层的方法。
本文提出一种自适应皮层约束的EEG-视觉对齐方法,通过重建EEG响应并采用基于证据的自适应视觉监督策略,解决零样本脑-图像检索中信号噪声和响应变异性问题。
本文提出NeuroGlyph方法,通过学习视觉层次中不同深度的信息来优化脑-图像检索目标,超越了仅使用最终层的方法。
本文针对同行评审中证据冲突和可靠性差异问题,提出一种基于证据可靠性聚合的元评审生成方法,有效识别并解决冲突,提升评审质量。
本文提出了一种基于捏合天线的集成感知与通信架构,通过动态重构辐射点来应对动态用户和目标,实现灵活且低成本的解决方案。
针对多视图城市区域表示学习中因共享潜在因素导致的误导性关联问题,提出CURE框架,通过估计并减少共享潜成分的影响来增强预测稳定性。