paper-with-me

홈 › Papers

Instabilities in Convnets for Raw Audio

2023-09-11 · Daniel Haider, Vincent Lostanlen, Martin Ehler, Peter Balazs

What makes waveform-based deep learning so hard? Despite numerous attempts at training convolutional neural networks (convnets) for filterbank design, they often fail to outperform hand-crafted baselines. These baselines are linear time-invariant systems: as such, they can be approximated by convnets with wide receptive fields. Yet, in practice, gradient-based optimization leads to suboptimal approximations. In our article, we approach this phenomenon from the perspective of initialization. We present a theory of large deviations for the energy response of FIR filterbanks with random Gaussian weights. We find that deviations worsen for large filters and locally periodic input signals, which are both typical for audio signal processing applications. Numerical simulations align with our theory and suggest that the condition number of a convolutional layer follows a logarithmic scaling law between the number and length of the filters, which is reminiscent of discrete wavelet bases.

📄 PDF Abstract BibTeX arXiv:2309.05855

Code (1)

danedane-haider/random-filterbanks 공식 구현 pytorch

Tasks

Audio Signal Processing

Methods 이 논문이 사용한 방법론

ALIGN In the ALIGN method, visual and language representations are jointly trained from noisy image alt-text data. The image and text encoders are learned via contrastive loss…
fail 설명 없음

Similar Papers 제목 키워드 기반

UniRepLKNet: A Universal Perception Large-Kernel ConvNet for Audio Video Point Cloud Time-Series and Image Recognition

2024-01-01 · CVPR 2024 1 · Xiaohan Ding, Yiyuan Zhang, Yixiao Ge, Sijie Zhao 외

Large-kernel convolutional neural networks (ConvNets) have recently received extensive research attention but two unresolved and critical issues demand further investigation. 1) The architectures of existing large-ke…

Time SeriesTime Series Forecasting

UniRepLKNet: A Universal Perception Large-Kernel ConvNet for Audio, Video, Point Cloud, Time-Series and Image Recognition

2023-11-27 · Xiaohan Ding, Yiyuan Zhang, Yixiao Ge, Sijie Zhao 외

Large-kernel convolutional neural networks (ConvNets) have recently received extensive research attention, but two unresolved and critical issues demand further investigation. 1) The architectures of existing large-kerne…

Image ClassificationObject DetectionSemantic SegmentationTime Series+1

Spatio-Temporal Graph Convolutional Neural Networks for Physics-Aware Grid Learning Algorithms

2022-03-31 · Tong Wu, Ignacio Losada Carreno, Anna Scaglione, Daniel Arnold

This paper proposes a model-free Volt-VAR control (VVC) algorithm via the spatio-temporal graph ConvNet-based deep reinforcement learning (STGCN-DRL) framework, whose goal is to control smart inverters in an unbalanced d…

Deep Reinforcement Learningreinforcement-learningReinforcement LearningReinforcement Learning (RL)

Scaling Up Your Kernels: Large Kernel Design in ConvNets towards Universal Representations

2024-10-10 · Yiyuan Zhang, Xiaohan Ding, Xiangyu Yue

This paper proposes the paradigm of large convolutional kernels in designing modern Convolutional Neural Networks (ConvNets). We establish that employing a few large kernels, instead of stacking multiple smaller ones, ca…

Time Series ForecastingVideo Recognition

Hold Me Tight: Stable Encoder-Decoder Design for Speech Enhancement

2024-08-30 · Daniel Haider, Felix Perfler, Vincent Lostanlen, Martin Ehler 외

Convolutional layers with 1-D filters are often used as frontend to encode audio signals. Unlike fixed time-frequency representations, they can adapt to the local characteristics of input data. However, 1-D filters on ra…

DecoderSpeech Enhancement