Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox

Picture for Matthew Mattina

S2TA: Exploiting Structured Sparsity for Energy-Efficient Mobile CNN Acceleration


Jul 16, 2021
Zhi-Gang Liu, Paul N. Whatmough, Yuhao Zhu, Matthew Mattina


  Access Paper or Ask Questions

On the Effects of Quantisation on Model Uncertainty in Bayesian Neural Networks


Feb 22, 2021
Martin Ferianc, Partha Maji, Matthew Mattina, Miguel Rodrigues

* Code at: https://github.com/martinferianc/quantised-bayesian-nets 

  Access Paper or Ask Questions

Doping: A technique for efficient compression of LSTM models using sparse structured additive matrices


Feb 14, 2021
Urmish Thakker, Paul N. Whatmough, Zhigang Liu, Matthew Mattina, Jesse Beu

* Accepted to be published at MLSys 2021 

  Access Paper or Ask Questions

Information contraction in noisy binary neural networks and its implications


Feb 01, 2021
Chuteng Zhou, Quntao Zhuang, Matthew Mattina, Paul N. Whatmough

* 14 pages, 8 figures 

  Access Paper or Ask Questions

MicroNets: Neural Network Architectures for Deploying TinyML Applications on Commodity Microcontrollers


Oct 25, 2020
Colby Banbury, Chuteng Zhou, Igor Fedorov, Ramon Matas Navarro, Urmish Thakker, Dibakar Gope, Vijay Janapa Reddi, Matthew Mattina, Paul N. Whatmough

* 10 pages, 8 figures, 3 tables, appendix 

  Access Paper or Ask Questions

Rank and run-time aware compression of NLP Applications


Oct 06, 2020
Urmish Thakker, Jesse Beu, Dibakar Gope, Ganesh Dasika, Matthew Mattina

* Published at [email protected] 2020. arXiv admin note: text overlap with arXiv:1906.04886 

  Access Paper or Ask Questions

Sparse Systolic Tensor Array for Efficient CNN Hardware Acceleration


Sep 04, 2020
Zhi-Gang Liu, Paul N. Whatmough, Matthew Mattina


  Access Paper or Ask Questions

High Throughput Matrix-Matrix Multiplication between Asymmetric Bit-Width Operands


Aug 03, 2020
Dibakar Gope, Jesse Beu, Matthew Mattina


  Access Paper or Ask Questions

Efficient Residue Number System Based Winograd Convolution


Jul 23, 2020
Zhi-Gang Liu, Matthew Mattina

* Accepted by ECCV2020 Conference 

  Access Paper or Ask Questions

TinyLSTMs: Efficient Neural Speech Enhancement for Hearing Aids


May 20, 2020
Igor Fedorov, Marko Stamenovic, Carl Jensen, Li-Chia Yang, Ari Mandell, Yiming Gan, Matthew Mattina, Paul N. Whatmough

* First four authors contributed equally. For audio samples, see https://github.com/BoseCorp/efficient-neural-speech-enhancement 

  Access Paper or Ask Questions

Systolic Tensor Array: An Efficient Structured-Sparse GEMM Accelerator for Mobile CNN Inference


May 16, 2020
Zhi-Gang Liu, Paul N. Whatmough, Matthew Mattina

* Accepted by IEEE Computer Architecture Letters on 3/4/2020 

  Access Paper or Ask Questions

Searching for Winograd-aware Quantized Networks


Feb 25, 2020
Javier Fernandez-Marques, Paul N. Whatmough, Andrew Mundy, Matthew Mattina

* Published as a conference paper at MLSys 2020 

  Access Paper or Ask Questions

Compressing Language Models using Doped Kronecker Products


Jan 31, 2020
Urmish Thakker, Paul Whatamough, Matthew Mattina, Jesse Beu

* Presented at On-device Intelligence Workshop at Third Conference on Machine Learning and Systems (MLSys) 2020 
* Link to Workshop - https://mlsys.org/Conferences/2020/Schedule?showEvent=1297 

  Access Paper or Ask Questions

Noisy Machines: Understanding Noisy Neural Networks and Enhancing Robustness to Analog Hardware Errors Using Distillation


Jan 14, 2020
Chuteng Zhou, Prad Kadambi, Matthew Mattina, Paul N. Whatmough


  Access Paper or Ask Questions

ISP4ML: Understanding the Role of Image Signal Processing in Efficient Deep Learning Vision Systems


Nov 25, 2019
Patrick Hansen, Alexey Vilkin, Yury Khrustalev, James Imber, David Hanwell, Matthew Mattina, Paul N. Whatmough

* 13 pages, 11 figures 

  Access Paper or Ask Questions

Ternary MobileNets via Per-Layer Hybrid Filter Banks


Nov 04, 2019
Dibakar Gope, Jesse Beu, Urmish Thakker, Matthew Mattina


  Access Paper or Ask Questions

Pushing the limits of RNN Compression


Oct 09, 2019
Urmish Thakker, Igor Fedorov, Jesse Beu, Dibakar Gope, Chu Zhou, Ganesh Dasika, Matthew Mattina

* 5th edition of Workshop on Energy Efficient Machine Learning and Cognitive Computing at NeurIPS 2019 
* 6 pages. arXiv admin note: substantial text overlap with arXiv:1906.02876 

  Access Paper or Ask Questions

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


Jun 18, 2019
Urmish Thakker, Jesse Beu, Dibakar Gope, Chu Zhou, Igor Fedorov, Ganesh Dasika, Matthew Mattina


  Access Paper or Ask Questions

Run-Time Efficient RNN Compression for Inference on Edge Devices


Jun 18, 2019
Urmish Thakker, Jesse Beu, Dibakar Gope, Ganesh Dasika, Matthew Mattina

* Published at 4th edition of Workshop on Energy Efficient Machine Learning and Cognitive Computing for Embedded Applications at International Symposium of Computer Architecture 2019, Phoenix, Arizona (https://www.emc2-workshop.com/isca-19) colocated with ISCA 2019 

  Access Paper or Ask Questions

SpArSe: Sparse Architecture Search for CNNs on Resource-Constrained Microcontrollers


May 28, 2019
Igor Fedorov, Ryan P. Adams, Matthew Mattina, Paul N. Whatmough


  Access Paper or Ask Questions

Ternary Hybrid Neural-Tree Networks for Highly Constrained IoT Applications


Mar 04, 2019
Dibakar Gope, Ganesh Dasika, Matthew Mattina

* 2nd Conference on Systems and Machine Learning (SysML), 2019 

  Access Paper or Ask Questions

Efficient Winograd or Cook-Toom Convolution Kernel Implementation on Widely Used Mobile CPUs


Mar 04, 2019
Partha Maji, Andrew Mundy, Ganesh Dasika, Jesse Beu, Matthew Mattina, Robert Mullins

* HPCA.EMC2 Feb 17, 2019 

  Access Paper or Ask Questions

Learning low-precision neural networks without Straight-Through Estimator(STE)


Mar 04, 2019
Zhi-Gang Liu, Matthew Mattina


  Access Paper or Ask Questions

FixyNN: Efficient Hardware for Mobile Computer Vision via Transfer Learning


Feb 27, 2019
Paul N. Whatmough, Chuteng Zhou, Patrick Hansen, Shreyas Kolala Venkataramanaiah, Jae-sun Seo, Matthew Mattina

* 10 pages, 8 figures, paper accepted at SysML2019 conference 

  Access Paper or Ask Questions