A 28nm 1.644TFLOPS/W Floating-Point Computation SRAM Macro with Variable Precision for Deep Neural Network Inference and Training.
Sangsu JeongJeongwoo ParkDongsuk JeonPublished in: ESSCIRC (2022)
Keyphrases
- floating point
- variable precision
- neural network
- floating point arithmetic
- decision table
- fixed point
- rough sets
- knowledge structures
- decision trees
- multi layer perceptron
- reduction method
- instruction set
- artificial neural networks
- training set
- pattern recognition
- fuzzy logic
- power consumption
- rough set theory
- natural language
- dominance relation
- neural network model
- bayesian networks