Unverified50% confidenceFactExact time
Transformer architecture and contrastive learning have been used to create embedding representations of sperm whale coda sequences.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/6/2026
First Seen
Valid until: 1/4/2027
Sources
Related Entities
Related Claims
UnverifiedThe Hierarchical Behavior Encoder (HBE) uses a Temporal Transformer and a Set Transformer to model the temporal evolution of individuals and the collective dynamics of the fish population, respectively.68% similarUnverified模仿学习的代表性方法包括行为克隆、扩散策略以及ACT（Action Chunking with Transformers）65% similarUnverifiedTransformer通过Self-Attention机制计算输入序列中所有Token之间的相关性权重，放弃状态延续换取并行训练能力和水平扩展性65% similarUnverifiedSentence-Transformers (SBERT) was proposed by Nils Reimers et al. in 2019, adding a pooling layer on top of pre-trained BERT and fine-tuning through a siamese network structure.63% similarUnverifiedThe self-attention mechanism of the Transformer architecture can capture relationships between any two positions in a sequence, making it well-suited for capturing ultra-long-range dependencies in DNA sequences, and outperforming traditional convolutional neural networks in this regard.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/980300API
curl https://kongchang.com/api/v1/knowledge/claims/980300MCP
get_claim(id=980300)