提出双流OSSE-LSTM与反事实归因的少标签时间序列分类
Get the story
2026年10月5日，arXiv（Artificial Intelligence，一手来源）报道提出 Dual-Stream OSSE-LSTM，用于大规模少标签时间序列分类。该框架属于 episodic metric-learning，将 Omni-Scale CNN 与 Squeeze-and-Excitation 重校准和双向 LSTM 融合，形成 prototype-oriented embedding。报道还提及反事实归因，但未给出其具体方法或实验细节。目前未见后续验证、代码或数据集信息。
Generated from reports · updated 16 hr ago
Timeline
Follow the coverage from different angles.
- arXiv · Artificial Intelligence双流 OSSE-LSTM 与反事实归因实现大规模少标签时间序列分类
论文提出 Dual-Stream OSSE-LSTM，一种 episodic metric-learning 框架，将 Omni-Scale CNN 与 Squeeze-and-Excitation 重校准和双向 LSTM 融合为 prototype-oriented embedding。
Heat trend
Current heat 7·Comparable peak 10（Oct 5）·Comparable change over 24 hours –
The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.