GitHub

ESC: Emulating Self-attention with Convolution for Efficient Image Super-Resolution

In this paper, we tackle the high computational overhead of transformers for lightweight image super-resolution. (SR). Motivated by the observations of self-attention's inter-layer repetition, we ...
GitHub

$\infty$-Video: A Training-Free Approach to Long Video Understanding via Continuous-Time Memory Consolidation

Official implementation of the paper $\infty$-Video: A Training-Free Approach to Long Video Understanding via Continuous-Time Memory Consolidation. **Abstract**: *Current video-language models ...
AlphaGalileo

Adapting to the Stream: An Instance-Attention GNN Method for Irregular Multivariate Time Series Data

DynIMTS replaces static graphs with instance-attention that updates edge weights on the fly, delivering SOTA imputation and P12 classification ...
IEEE

Continuous-Time Data-Driven Unknown-Input Observers

Abstract: State estimation in the presence of unknown inputs remains a fundamental challenge in control and monitoring applications. Traditional unknown-input observer (UIO) relies on explicit system ...
IEEE

Generalized Metaplectic Convolution-Based Cohen's Class Time-Frequency Distribution: Theory and Application

Abstract: The convolution type of the Cohen's class time-frequency distribution (CCTFD) is a useful and effective time-frequency analysis tool for additive noises jamming signals. However, it can't ...