A 6.54-to-26.03 TOPS/W Computing-In-Memory RNN Processor using Input Similarity Optimization and Attention-based Context-breaking with Output Speculation.
Ruiqi GuoHao LiRuhui LiuZhixiao ZhangLimei TangHao SunLeibo LiuMeng-Fan ChangShaojun WeiShouyi YinPublished in: VLSI Circuits (2021)
Keyphrases
- recurrent neural networks
- input data
- nearest neighbor
- multiple output
- optimization method
- context aware
- contextual information
- high speed
- global optimization
- memory management
- similarity measure
- euclidean distance
- input output
- similarity function
- hidden units
- desired output
- optimization algorithm
- input variables
- rbf network