Jan 29, 2021 · Neural network accelerator is a key enabler for the on-device AI inference, for which energy efficiency is an important metric.

Feb 13, 2020 · Based on the insight, we propose a post-training optimization algorithm and a hamming-distance-aware training algorithm to co-design and co- ...

A series of post-training and training-aware techniques are proposed to co-design and co-optimize the accelerator and the network to reduce the hamming distance ...

Feb 13, 2020 · We discover the correlation between the datapath en- ergy and the hamming distance when streaming the input operands and further propose the ...

Jan 21, 2021 · 6.3 Combined Hamming Distance Optimization. We now combine the post-training optimization techniques with the training-aware optimization ...

Based on the insight, we propose a post-training optimization algorithm and a hamming-distance-aware training algorithm to co-design and co-optimize the ...

In dataflow processing, operands are streamed into the compute array. Datapath energy is determined by the total bit flips induced by operand streaming.

In this paper, a new algorithm that generates low complexity DCT approximation matrices with minimum number of low-frequency coefficients for all transform ...

Improving Efficiency in Neural Network Accelerator using Operands Hamming Distance Optimization. M. Li, Y. Li, and V. Chandra. ASP-DAC, page 599-604.

Improving efficiency in neural network accelerator using operands hamming distance optimization. Meng Li, Yilei Li, Vikas Chandra. Co-exploration of neural ...