ETA: An Efficient Training Accelerator for DNNs Based on Hardware-Algorithm Co-Optimization.
Jinming LuChao NiZhongfeng WangPublished in: IEEE Trans. Neural Networks Learn. Syst. (2023)
Keyphrases
- optimization algorithm
- dynamic programming
- detection algorithm
- computational complexity
- optimal solution
- high accuracy
- experimental evaluation
- improved algorithm
- training process
- worst case
- computational cost
- significant improvement
- objective function
- constrained optimization
- optimization process
- parallel implementation
- training phase
- evolution strategy
- classification algorithm
- computationally efficient
- preprocessing
- matching algorithm
- optimization method
- times faster
- recognition algorithm
- hardware implementation
- expectation maximization
- low cost
- k means
- training algorithm
- search space
- real time
- hardware architecture
- stochastic gradient descent