Filters








3 Hits in 4.2 sec

Compressing RNNs for IoT devices by 15-38x using Kronecker Products [article]

Urmish Thakker, Jesse Beu, Dibakar Gope, Chu Zhou, Igor Fedorov, Ganesh Dasika, Matthew Mattina
2020 arXiv   pre-print
This paper introduces a method to compress RNNs for resource constrained environments using Kronecker product (KP). KPs can compress RNN layers by 15-38x with minimal accuracy loss.  ...  Recurrent Neural Networks (RNN) can be difficult to deploy on resource constrained devices due to their size.As a result, there is a need for compression techniques that can significantly compress RNNs  ...  LG] 31 Jan 2020 Compressing RNNs for IoT devices by 15-38x using Kronecker Products 2 RELATED WORK KPs have been used in the deep learning community in the past (Jose et al., 2017; Zhou & Wu, 2015) .  ... 
arXiv:1906.02876v5 fatcat:dtxwn4wfarfwpmbbuykgk7clpy

Pushing the limits of RNN Compression [article]

Urmish Thakker, Igor Fedorov, Jesse Beu, Dibakar Gope, Chu Zhou, Ganesh Dasika, Matthew Mattina
2019 arXiv   pre-print
This paper introduces a method to compress RNNs for resource constrained environments using Kronecker product (KP). KPs can compress RNN layers by 16-38x with minimal accuracy loss.  ...  As a result, there is a need for compression techniques that can significantly compress RNNs without negatively impacting task accuracy.  ...  Similar observations can be made for the pruned HAR1 network. Conclusion We show how to compress RNN Cells by 15× to 38× using Kronecker products.  ... 
arXiv:1910.02558v2 fatcat:4diffzmavzejzfp2py7xtckyz4

Compressing Language Models using Doped Kronecker Products [article]

Urmish Thakker, Paul N. Whatmough, Zhi-Gang Liu, Matthew Mattina, Jesse Beu
2020 arXiv   pre-print
Kronecker Products (KP) have been used to compress IoT RNN Applications by 15-38x compression factors, achieving better results than traditional compression methods.  ...  We call this compression method doped kronecker product compression.  ...  Recently, Kronecker Products (KP) were used to compress IoT applications by 15-38x compression factors (Thakker et al., 2019d; and achieves better accuracy than pruning (Zhu & Gupta, 2017) , low-rank  ... 
arXiv:2001.08896v5 fatcat:pwyear3j3vepphqrh6zwp7eapm