Matthijs Van keirsbilck  

 



  ![](/sites/default/files/person/P3231541-min-crop-tiny.jpg)

  

 Matthijs is a Senior Research Scientist at NVIDIA, working on foundations of neural networks and machine learning.
He has a broad field of interest, and is currently working on efficient neural network architecture design, leveraging structural sparsity and quantization.
He's also very interested in understanding the training dynamics of current models.
He has a background in Electrical Engineering, studying at KU Leuven (Belgium) and TU Munich (Germany).



   Research Area(s)

[Algorithms and Numerical Methods](/research-area/algorithms)

[Artificial Intelligence and Machine Learning ](/research-area/machine-learning-artificial-intelligence)

[Circuits and VLSI Design](/research-area/circuits)

[Computer Architecture](/research-area/computer-architecture)

[Speech Processing](/research-area/speech-processing)

 

 

  

 Main Field of Interest

[Artificial Intelligence and Machine Learning ](/research-area/machine-learning-artificial-intelligence)

 

  

 Google Scholar

[https://scholar.google.com/citations?user=qphCQwQAAAAJ&h](https://scholar.google.com/citations?user=qphCQwQAAAAJ&h)

 

  

 

 

 



 ### Publications

 

### 2025 

[Hymba: A Hybrid-head Architecture for Small Language Models](/publication/2025-04_hymba-hybrid-head-architecture-small-language-models)

Xin Dong, [Yonggan Fu\*](/person/yonggan-fu), Shizhe Diao, [Wonmin Byeon](/person/wonmin-byeon), Zijia Chen, Ameya Sunil Mahabaleshwarkar, Shih-Yang Liu, [Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Min-Hung Chen](/person/min-hung-chen), [Yoshi Nishi](/person/yoshi-nishi), Yingyan Celine Lin, [Jan Kautz](/person/jan-kautz), [Pavlo Molchanov](/person/pavlo-molchanov)



[Hymba - ICLR 2025](https://jankautz.com/publications/Hymba_ICLR25.pdf)



ICLR spotlight paper





### 2024 

[Conformer without Convolutions](/publication/2024-09_conformer-without-convolutions)

[Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller)



[Interspeech 2024](https://www.isca-archive.org/interspeech_2024/vankeirsbilck24_interspeech.pdf)









### 2022 

[Deep Learning-Based Synchronization for Uplink NB-IoT](/publication/2022-12_deep-learning-based-synchronization-uplink-nb-iot)

[Fayçal Aït Aoudia ](/person/faycal-ait-aoudia), [Jakob Hoydis](/person/jakob-hoydis), [Sebastian Cammerer](/person/sebastian-cammerer), [Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller)



[IEEE Global Communications Conference 2022](https://ieeexplore.ieee.org/abstract/document/10001353)









[Artificial Neural Networks generated by Low Discrepancy Sequences](/publication/2022-05_artificial-neural-networks-generated-low-discrepancy-sequences)

[Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller)



[arxiv link](https://arxiv.org/pdf/2103.03543.pdf)









### 2021 

[Compressing 1D Time-Channel Separable Convolutions using Sparse Random Ternary Matrices](/publication/2021-08_compressing-1d-time-channel-separable-convolutions-using-sparse-random-ternary)

Goncalo Mordido, [Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller)



[Interspeech 2021](https://www.isca-speech.org/archive/pdfs/interspeech_2021/mordido21_interspeech.pdf)









### 2020 

[Monte Carlo Gradient Quantization](/publication/2020-06_monte-carlo-gradient-quantization)

Goncalo Mordido, [Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller)



[CVPR2020 Workshop](https://openaccess.thecvf.com/content_CVPRW_2020/papers/w40/Mordido_Monte_Carlo_Gradient_Quantization_CVPRW_2020_paper.pdf)









### 2019 

[Instant Quantization of neural networks using Monte Carlo Methods](/publication/2019-12_instant-quantization-neural-networks-using-monte-carlo-methods)

Gonçalo Mordido, [Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller)



[NeurIPS 2019 EMC2 workshop (IEEExplore)](https://ieeexplore.ieee.org/document/9463561)









[Rethinking full connectivity in recurrent neural networks](/publication/2019-03_rethinking-full-connectivity-recurrent-neural-networks)

[Matthijs Van keirsbilck](/person/matthijs-van-keirsbilck), [Alex Keller](/person/alex-keller), Xiaodong Yang



[arxiv paper](https://arxiv.org/pdf/1905.12340.pdf)