Title | ||
---|---|---|
Self-Supervised Quantization Of Pre-Trained Neural Networks For Multiplierless Acceleration |
Abstract | ||
---|---|---|
To host intelligent algorithms such as Deep Neural Networks on embedded devices, it is beneficial to transform the data representation of neural networks into a fixed-point format with reduced bit-width. In this paper we present a novel quantization procedure for parameters and activations of pre-trained neural networks. For 8 bit linear quantization, our procedure achieves close to original network performance without retraining and consequently does not require labeled training data. Additionally, we evaluate our method for power-of-two quantization as well as for a two-hot quantization scheme, enabling shift-based inference. To underline the hardware benefits of a multiplierless accelerator, we propose the design of a shift-based processing element. |
Year | DOI | Venue |
---|---|---|
2019 | 10.23919/DATE.2019.8714901 | 2019 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION (DATE) |
Keywords | Field | DocType |
Quantization, Neural Networks, Hardware, Multiplierless Acceleration | Computer science,Parallel computing,Acceleration,Artificial intelligence,Artificial neural network,Quantization (signal processing) | Conference |
ISSN | Citations | PageRank |
1530-1591 | 1 | 0.34 |
References | Authors | |
0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Sebastian Vogel | 1 | 23 | 5.63 |
Jannik Springer | 2 | 1 | 0.68 |
Andre Guntoro | 3 | 20 | 11.05 |
Gerd Ascheid | 4 | 1205 | 144.76 |