Publications (10)
Hold Me Tight: Stable Encoder-Decoder Design for Speech Enhancement
Daniel Haider, Felix Perfler, Vincent Lostanlen +2
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 o…
Instabilities in Convnets for Raw Audio
Daniel Haider, Vincent Lostanlen, Martin Ehler +1
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…
Injectivity of ReLU-layers: Tools from Frame Theory
Daniel Haider, Martin Ehler, Peter Balazs
Injectivity is the defining property of a mapping that ensures no information is lost and any input can be perfectly reconstructed from its output. By performing hard thresholding,…
Phase-Based Signal Representations for Scattering
Daniel Haider, Peter Balazs, Nicki Holighaus
The scattering transform is a non-linear signal representation method based on cascaded wavelet transform magnitudes. In this paper we introduce phase scattering, a novel approach…
(Almost) Smooth Sailing: Towards Numerical Stability of Neural Networks Through Differentiable Regularization of the Condition Number
Rossen Nenov, Daniel Haider, Peter Balazs
Maintaining numerical stability in machine learning models is crucial for their reliability and performance. One approach to maintain stability of a network layer is to integrate t…
Fitting Auditory Filterbanks with Multiresolution Neural Networks
Vincent Lostanlen, Daniel Haider, Han Han +3
Waveform-based deep learning faces a dilemma between nonparametric and parametric approaches. On one hand, convolutional neural networks (convnets) may approximate any linear time-…
ISAC: An Invertible and Stable Auditory Filter Bank with Customizable Kernels for ML Integration
Daniel Haider, Felix Perfler, Peter Balazs +2
This paper introduces ISAC, an invertible and stable, perceptually-motivated filter bank that is specifically designed to be integrated into machine learning paradigms. More precis…
Optimal lower Lipschitz bounds for ReLU layers, saturation, and phase retrieval
Daniel Freeman, Daniel Haider
The injectivity of ReLU layers in neural networks, the recovery of vectors from clipped or saturated measurements, and (real) phase retrieval in allow for a similar…
Aliasing in Convnets: A Frame-Theoretic Perspective
Daniel Haider, Vincent Lostanlen, Martin Ehler +2
Using a stride in a convolutional layer inherently introduces aliasing, which has implications for numerical stability and statistical generalization. While techniques such as the…
Convex Geometry of ReLU-layers, Injectivity on the Ball and Local Reconstruction
Daniel Haider, Martin Ehler, Peter Balazs
The paper uses a frame-theoretic setting to study the injectivity of a ReLU-layer on the closed ball of and its non-negative part. In particular, the interplay betwe…