Filters








9 Hits in 8.1 sec

The Incomplete Rosetta Stone Problem: Identifiability Results for Multi-View Nonlinear ICA [article]

Luigi Gresele, Paul K. Rubenstein, Arash Mehrjou, Francesco Locatello, Bernhard Schölkopf
2019 arXiv   pre-print
In contrast to known identifiability results for nonlinear ICA, we prove that independent latent sources with arbitrary mixing can be recovered as long as multiple, sufficiently different noisy views are  ...  When the views are considered separately, this reduces to nonlinear Independent Component Analysis (ICA) for which it is provably impossible to undo the mixing.  ...  Acknowledgements Thanks to Krikamol Muandet for providing his office for fruitful discussions, to Matthias Bauer and Manuel Wüthrich for proofreading and to Lucia Busso for interesting input about linguistics  ... 
arXiv:1905.06642v2 fatcat:mlis7vcwkndrplu5tzh2vxm7fa

Understanding Latent Correlation-Based Multiview Learning and Self-Supervision: An Identifiability Perspective [article]

Qi Lyu, Xiao Fu, Weiran Wang, Songtao Lu
2022 arXiv   pre-print
A finite sample analysis, which has been rare in nonlinear mixture identifiability study, is also presented. The theoretical results and newly designed regularization are tested on a series of tasks.  ...  An intuitive generative model of multiview data is adopted, where the views are different nonlinear mixtures of shared and private components.  ...  "identity-revealing" information to a certain extent, as in the MNIST case.  ... 
arXiv:2106.07115v3 fatcat:pldhjjctfza3jewxqzh5axeheq

Weakly-Supervised Disentanglement Without Compromises [article]

Francesco Locatello, Ben Poole, Gunnar Rätsch, Bernhard Schölkopf, Olivier Bachem, Michael Tschannen
2020 arXiv   pre-print
Overall, our results demonstrate that weak supervision enables learning of useful disentangled representations in realistic scenarios.  ...  We model such observations as pairs of non-i.i.d. images sharing at least one of the underlying factors of variation.  ...  Acknowledgments: The authors thank Stefan Bauer, Ilya Tolstikhin, Sarah Strauss and Josip Djolonga for helpful discussions and comments.  ... 
arXiv:2002.02886v4 fatcat:jbgrcjogyfhm7p3shu7jdz5dum

Disentangling Factors of Variation Using Few Labels [article]

Francesco Locatello, Michael Tschannen, Stefan Bauer, Gunnar Rätsch, Bernhard Schölkopf, Olivier Bachem
2020 arXiv   pre-print
We observe that a small number of labeled examples (0.01--0.5\% of the data set), with potentially imprecise and incomplete labels, is sufficient to perform model selection on state-of-the-art unsupervised  ...  Learning disentangled representations is considered a cornerstone problem in representation learning.  ...  Acknowledgments: Francesco Locatello is supported by the Max Planck ETH Center for Learning Systems, by an ETH core grant (to Gunnar Rätsch), and by a Google Ph.D. Fellowship.  ... 
arXiv:1905.01258v2 fatcat:3nvwprvr4vhc5jrl2lycls7sg4

Embrace the Gap: VAEs Perform Independent Mechanism Analysis [article]

Patrik Reizinger, Luigi Gresele, Jack Brady, Julius von Kügelgen, Dominik Zietlow, Bernhard Schölkopf, Georg Martius, Wieland Brendel, Michel Besserve
2022 arXiv   pre-print
The gap between ELBO and log-likelihood is therefore welcome, since it bears unanticipated benefits for nonlinear representation learning.  ...  We seek to elucidate this apparent paradox by studying nonlinear VAEs in the limit of near-deterministic decoders.  ...  covariance with a uniqueness result for diagonal Σ φ z|x .  ... 
arXiv:2206.02416v1 fatcat:xzszp3m4dja3zcrcjl6hngo7iy

Enforcing and Discovering Structure in Machine Learning [article]

Francesco Locatello
2021 arXiv   pre-print
The world is structured in countless ways.  ...  In this dissertation, we consider two different research areas that concern structuring a learning algorithm's solution: when the structure is known and when it has to be discovered.  ...  "The incomplete rosetta stone problem: Identifiability results for multi-view nonlinear ica". In: Conference on Uncertainty in Artificial Intelligence (UAI).  ... 
arXiv:2111.13693v1 fatcat:2urmfjeh6nhvjeiv3qoiy5hrum

Theoretical and experimental investigation of the insertion loss of a dissipative muffler

Christian Pommer, Viggo Tarnow
1986 Journal of the Acoustical Society of America  
However, for sources and receivers near the seafloor, Stonely waves are observed with velocities near the shear wave velocity.  ...  This results in a response that can be viewed as a summation of propagating and standing wave effects.  ...  From a user's point of view, these different types of instruments should provide identical measurement results--but this is not always the case in practice.  ... 
doi:10.1121/1.2024087 fatcat:tcee6inawvhubc6ftuk2ffhyce

Making galaxies passive: Insights from resolved studies of nearby galaxies

Miguel Querejeta
2016
We also find that major mergers can simultaneously account for the difference in angular momentum and concentration between spirals and S0s, as recently found by the CALIFA team.  ...  We start presenting our catalogue of stellar mass maps for more than 1500 nearby galaxies, which has been publicly released.  ...  The authors would like to thank the anonymous referee for a helpful report, as well as Daniela Calzetti, Nick Z. Scoville and Mari Polletta for making the HST/F190N mosaic available to us.  ... 
doi:10.11588/heidok.00020891 fatcat:s45og523eja7rgkma7svybxwdi

Forty-First Annual Meeting March 2-6, 1997 Ernest N. Morial Convention Center New Orleans, Louisiana. Wednesday Symposia and Posters, Part I

1997 Biophysical Journal  
Pollack)) Center for Bioeng. 357962, Univ. of Wash, Seattle WA 98195 W - W AM44 A MOLECULAR ROSETTA STONE: DOES THE CRYSTAL STRUCTURE OF THE KINESIN MOTOR DOMAIN REVEAL AN EVOLUTIONARY LINK BETWEEN MOLECULAR  ...  Our results are consistent with the view that myosin IIB perforns a direct role in facilitating filopodial extension of neuronal growth cones.  ... 
doi:10.1016/s0006-3495(97)78740-x fatcat:erhwjii3gzb6fo43eqzapv5b4u