ROSETTA: A Resource and Energy-Efficient Inference Processor for Recurrent Neural Networks Based on Programmable Data Formats and Fine Activation Pruning.
Jiho KimTae-Hwan KimPublished in: IEEE Trans. Emerg. Top. Comput. (2023)
Keyphrases
- human body
- recurrent neural networks
- efficient inference
- data formats
- databases
- processing units
- data types
- probabilistic inference
- database management systems
- neural network
- feed forward
- fully connected
- conditional random fields
- metadata
- markov random field
- hidden variables
- artificial neural networks
- human pose estimation
- hidden layer
- echo state networks
- parallel processing
- approximate inference
- exact inference
- graph structure
- database
- markov networks
- higher order
- structured prediction
- general purpose
- hidden markov models
- back propagation
- data management
- computing systems
- junction tree
- preprocessing
- database systems