Zobrazeno 1 - 10
of 459
pro vyhledávání: '"Balazs Peter"'
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 the condition number of the weight matrix as a regularizing term into t
Externí odkaz:
http://arxiv.org/abs/2410.00169
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 raw audio are hard to train and
Externí odkaz:
http://arxiv.org/abs/2408.17358
Autor:
Köhldorfer, Lukas, Balazs, Peter
In this article we introduce several new examples of Wiener pairs $\mathcal{A} \subseteq \mathcal{B}$, where $\mathcal{B} = \mathcal{B}(\ell^2(X;\mathcal{H}))$ is the Banach algebra of bounded operators acting on the Hilbert space-valued Bochner sequ
Externí odkaz:
http://arxiv.org/abs/2407.16416
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, the ReLU function naturally interferes with this property, making the
Externí odkaz:
http://arxiv.org/abs/2406.15856
Kernel theorems, in general, provide a convenient representation of bounded linear operators. For the operator acting on a concrete function space, this means that its action on any element of the space can be expressed as a generalised integral oper
Externí odkaz:
http://arxiv.org/abs/2402.18367
In this paper we ask when it is possible to transform a given sequence into a frame or a lower semi frame by multiplying the elements by numbers. In other words, we ask when a given sequence is a weighted frame or a weighted lower semi frame and for
Externí odkaz:
http://arxiv.org/abs/2310.18957
Autor:
Balazs, Peter, Shamsabadi, Mitra
Frames and orthonormal bases are naturally linked to bounded operators. To tackle unbounded operators those sequences might not be well suited. This has already been noted by von Neumann in the 1920ies. But modern frame theory also investigates other
Externí odkaz:
http://arxiv.org/abs/2310.02088
Publikováno v:
IEEE Signal Processing Letters 31 (2024) 1084-1088
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 sys
Externí odkaz:
http://arxiv.org/abs/2309.05855
Autor:
Lostanlen, Vincent, Haider, Daniel, Han, Han, Lagrange, Mathieu, Balazs, Peter, Ehler, Martin
Publikováno v:
2023 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA 2023)
Waveform-based deep learning faces a dilemma between nonparametric and parametric approaches. On one hand, convolutional neural networks (convnets) may approximate any linear time-invariant system; yet, in practice, their frequency responses become m
Externí odkaz:
http://arxiv.org/abs/2307.13821
Autor:
Molnár, Dániel, Török, Tímea Nóra, Kövecs, Roland, Pósa, László, Balázs, Péter, Molnár, György, Olalla, Nadia Jimenez, Leuthold, Juerg, Volk, János, Csontos, Miklós, Halbritter, András
Analog tunable memristors are widely utilized as artificial synapses in various neural network applications. However, exploiting the dynamical aspects of their conductance change to implement active neurons is still in its infancy, awaiting the reali
Externí odkaz:
http://arxiv.org/abs/2307.13320