Reproduce “Event-LSTM: An Unsupervised and Asynchronous Learning-based Representation for Event-based Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lakshmi Annamalai, Vignesh Ramanathan, Chetan Singh Thakur
Event-LSTM is an unsupervised learning framework that generates meaningful 2D grid representations directly from raw event streams without requiring labeled data. It produces speed-invariant features through asynchronous windowing, improves noise robustness with a memory-based filter, and achieves superior classification performance compared to hand-crafted features.
Key insights
- Unsupervised learning with LSTMs can extract meaningful temporal patterns from event data without requiring labeled ground truth.
- Processing event data asynchronously, based on event count rather than fixed time windows, leads to speed-invariant features and improved energy efficiency.
- Incorporating historical event data (memory) into denoising filters significantly enhances robustness against spurious noise events.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.