Zobrazeno 1 - 10
of 3 972
pro vyhledávání: '"Gresele, A. At"'
We analyze identifiability as a possible explanation for the ubiquity of linear properties across language models, such as the vector difference between the representations of "easy" and "easiest" being parallel to that between "lucky" and "luckiest"
Externí odkaz:
http://arxiv.org/abs/2410.23501
Autor:
Ghosh, Shubhangi, Gresele, Luigi, von Kügelgen, Julius, Besserve, Michel, Schölkopf, Bernhard
Independent Mechanism Analysis (IMA) seeks to address non-identifiability in nonlinear Independent Component Analysis (ICA) by assuming that the Jacobian of the mixing function has orthogonal columns. As typical in ICA, previous work focused on the c
Externí odkaz:
http://arxiv.org/abs/2312.13438
Autor:
Jin, Zhijing, Chen, Yuen, Leeb, Felix, Gresele, Luigi, Kamal, Ojasv, Lyu, Zhiheng, Blin, Kevin, Adauto, Fernando Gonzalez, Kleiman-Weiner, Max, Sachan, Mrinmaya, Schölkopf, Bernhard
The ability to perform causal reasoning is widely considered a core feature of intelligence. In this work, we investigate whether large language models (LLMs) can coherently reason about causality. Much of the existing work in natural language proces
Externí odkaz:
http://arxiv.org/abs/2312.04350
Autor:
von Kügelgen, Julius, Besserve, Michel, Wendong, Liang, Gresele, Luigi, Kekić, Armin, Bareinboim, Elias, Blei, David M., Schölkopf, Bernhard
We study causal representation learning, the task of inferring latent causal variables and their causal relations from high-dimensional mixtures of the variables. Prior work relies on weak supervision, in the form of counterfactual pre- and post-inte
Externí odkaz:
http://arxiv.org/abs/2306.00542
Autor:
Wendong, Liang, Kekić, Armin, von Kügelgen, Julius, Buchholz, Simon, Besserve, Michel, Gresele, Luigi, Schölkopf, Bernhard
Independent Component Analysis (ICA) aims to recover independent latent variables from observed mixtures thereof. Causal Representation Learning (CRL) aims instead to infer causally related (thus often statistically dependent) latent variables, toget
Externí odkaz:
http://arxiv.org/abs/2305.17225
Autor:
Paolo Gresele, Giuseppe Guglielmini, Maurizio Del Pinto, Paolo Calabrò, Pasquale Pignatelli, Giuseppe Patti, Vittorio Pengo, Emilia Antonucci, Plinio Cirillo, Tiziana Fierro, Gualtiero Palareti, Rossella Marcucci, START Antiplatelet Registry Group
Publikováno v:
Scientific Reports, Vol 14, Iss 1, Pp 1-10 (2024)
Abstract Some previous observations suggest that a low platelet count is associated with an increased risk of adverse outcomes in patients with acute coronary syndromes (ACS). However, most of the data come from post-hoc analyses of randomized contro
Externí odkaz:
https://doaj.org/article/bb1cd95c94b44906bfe953431c75206f
Autor:
Kekić, Armin, Dehning, Jonas, Gresele, Luigi, von Kügelgen, Julius, Priesemann, Viola, Schölkopf, Bernhard
Early on during a pandemic, vaccine availability is limited, requiring prioritisation of different population groups. Evaluating vaccine allocation is therefore a crucial element of pandemics response. In the present work, we develop a model to retro
Externí odkaz:
http://arxiv.org/abs/2212.08498
Autor:
Paolo Gresele
Publikováno v:
Bleeding, Thrombosis and Vascular Biology, Vol 3, Iss 3 (2024)
Externí odkaz:
https://doaj.org/article/f7b022766a914ae4b3fbf4771872a9b0
One aim of representation learning is to recover the original latent code that generated the data, a task which requires additional information or inductive biases. A recently proposed approach termed Independent Mechanism Analysis (IMA) postulates t
Externí odkaz:
http://arxiv.org/abs/2207.06137
Autor:
Reizinger, Patrik, Gresele, Luigi, Brady, Jack, von Kügelgen, Julius, Zietlow, Dominik, Schölkopf, Bernhard, Martius, Georg, Brendel, Wieland, Besserve, Michel
Variational autoencoders (VAEs) are a popular framework for modeling complex data distributions; they can be efficiently trained via variational inference by maximizing the evidence lower bound (ELBO), at the expense of a gap to the exact (log-)margi
Externí odkaz:
http://arxiv.org/abs/2206.02416