Unverified50% confidenceFactExact time
The mathematical foundation of linear attention originates from the work of Katharopoulos et al. in 2020, which replaces the exp(QKᵀ) in Softmax attention with a kernel function.
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
Unverified线性注意力的数学基础源于2020年 Katharopoulos 等人发表于ICML的《Transformers are RNNs》，将时间复杂度从 O(N²) 降至 O(N)68% similarUnverified傅里叶神经算子（FNO）由Zongyi Li等人于2020年提出，学习的是无穷维函数空间之间的映射，能以比传统数值方法快数个数量级的速度求解偏微分方程61% similarUnverified拉格朗日对偶通过引入拉格朗日乘子将约束吸收进目标函数，再利用 KKT 条件转化为对偶问题，引出支持向量和核技巧的理论基础59% similarUnverifiedColBERT由Stanford DAWN实验室的Omar Khattab等人于2020年提出，采用延迟交互机制和MaxSim（Chamfer相似度）打分58% similarUnverifiedThe finite state automaton (FSA) was formally proposed by mathematician Stephen Kleene in 1956 and is completely equivalent to regular expressions in expressive power.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/590287API
curl https://kongchang.com/api/v1/knowledge/claims/590287MCP
get_claim(id=590287)