Chin-Yun Yu

Chin-Yun Yu

Researcher/Software Engineer/MIR Enthusiast/DjentleMan

Posts by Tags

differentiable IIR

Block-based Fast Differentiable IIR in PyTorch

16 minute read

Published:

I recently came across a presentation by Andres Ezequiel Viso from GPU Audio at ADC 2022, in which he talked about how they accelerate IIR filters on the GPU. The approach they use is to formulate the IIR filter as a state-space model (SSM) and augment the transition matrix so that each step processes multiple samples at once. The primary speedup stems from the fact that GPUs are very good at performing large matrix multiplications, and the SSM formulation enables us to leverage this capability.

Notes on differentiable TDF-II filter

9 minute read

Published:

This blog is a continuation of some of my early calculations for propagating gradients through general IIR filters, including direct-form and transposed-direct-form.

nmf

How to Train Deep NMF Model in PyTorch

3 minute read

Published:

Recently I updated the implementation of PyTorch-NMF to make it be able to scale on large and complex NMF models. In this blog post I will briefly explain how this was done thanks to the automatic differentiation of PyTorch.

pytorch

Block-based Fast Differentiable IIR in PyTorch

16 minute read

Published:

I recently came across a presentation by Andres Ezequiel Viso from GPU Audio at ADC 2022, in which he talked about how they accelerate IIR filters on the GPU. The approach they use is to formulate the IIR filter as a state-space model (SSM) and augment the transition matrix so that each step processes multiple samples at once. The primary speedup stems from the fact that GPUs are very good at performing large matrix multiplications, and the SSM formulation enables us to leverage this capability.

Notes on differentiable TDF-II filter

9 minute read

Published:

This blog is a continuation of some of my early calculations for propagating gradients through general IIR filters, including direct-form and transposed-direct-form.

How to Train Deep NMF Model in PyTorch

3 minute read

Published:

Recently I updated the implementation of PyTorch-NMF to make it be able to scale on large and complex NMF models. In this blog post I will briefly explain how this was done thanks to the automatic differentiation of PyTorch.

scientific computing

Block-based Fast Differentiable IIR in PyTorch

16 minute read

Published:

I recently came across a presentation by Andres Ezequiel Viso from GPU Audio at ADC 2022, in which he talked about how they accelerate IIR filters on the GPU. The approach they use is to formulate the IIR filter as a state-space model (SSM) and augment the transition matrix so that each step processes multiple samples at once. The primary speedup stems from the fact that GPUs are very good at performing large matrix multiplications, and the SSM formulation enables us to leverage this capability.

Notes on differentiable TDF-II filter

9 minute read

Published:

This blog is a continuation of some of my early calculations for propagating gradients through general IIR filters, including direct-form and transposed-direct-form.

How to Train Deep NMF Model in PyTorch

3 minute read

Published:

Recently I updated the implementation of PyTorch-NMF to make it be able to scale on large and complex NMF models. In this blog post I will briefly explain how this was done thanks to the automatic differentiation of PyTorch.

side project

How to Train Deep NMF Model in PyTorch

3 minute read

Published:

Recently I updated the implementation of PyTorch-NMF to make it be able to scale on large and complex NMF models. In this blog post I will briefly explain how this was done thanks to the automatic differentiation of PyTorch.

state-space model

Block-based Fast Differentiable IIR in PyTorch

16 minute read

Published:

I recently came across a presentation by Andres Ezequiel Viso from GPU Audio at ADC 2022, in which he talked about how they accelerate IIR filters on the GPU. The approach they use is to formulate the IIR filter as a state-space model (SSM) and augment the transition matrix so that each step processes multiple samples at once. The primary speedup stems from the fact that GPUs are very good at performing large matrix multiplications, and the SSM formulation enables us to leverage this capability.