Zobrazeno 1 - 5
of 5
pro vyhledávání: '"Mikhaeil, Jonas"'
Label bias occurs when the outcome of interest is not directly observable and instead modeling is performed with proxy labels. When the difference between the true outcome and the proxy label is correlated with predictors, this can yield systematic d
Externí odkaz:
http://arxiv.org/abs/2403.00639
Autor:
Brenner, Manuel, Hess, Florian, Mikhaeil, Jonas M., Bereska, Leonard, Monfared, Zahra, Kuo, Po-Chen, Durstewitz, Daniel
In many scientific disciplines, we are interested in inferring the nonlinear dynamical system underlying a set of observed time series, a challenging task in the face of chaotic behavior and noise. Previous deep learning approaches toward this goal o
Externí odkaz:
http://arxiv.org/abs/2207.02542
Recurrent neural networks (RNNs) are wide-spread machine learning tools for modeling sequential and time series data. They are notoriously hard to train because their loss gradients backpropagated in time tend to saturate or diverge during training.
Externí odkaz:
http://arxiv.org/abs/2110.07238
Bernstein Conference 2021 abstract. http://bernstein-conference.de
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::ac57da100e7066af393dceeddb9efcf7
Autor:
Bereska, Leonard, Brenner, Manuel, Kuo, Po C., Monfared, Zahra, Mikhaeil, Jonas, Durstewitz, Daniel
Bernstein Conference 2021 abstract. http://bernstein-conference.de
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::dfcf20a00dc7472025c1515d3e293e69