no code implementations • 8 Jan 2025 • Tâm Le Minh, Julyan Arbel, Thomas Möllenhoff, Mohammad Emtiyaz Khan, Florence Forbes
We introduce a new multimodal optimization approach called Natural Variational Annealing (NVA) that combines the strengths of three foundational concepts to simultaneously search for multiple global and local modes of black-box nonconvex objectives.
no code implementations • 11 Dec 2024 • Hugo Monzón Maldonado, Thomas Möllenhoff, Nico Daheim, Iryna Gurevych, Mohammad Emtiyaz Khan
When finetuning multiple tasks altogether, it is important to carefully weigh them to get a good performance, but searching for good weights can be difficult and costly.
1 code implementation • 7 Nov 2024 • Bai Cong, Nico Daheim, Yuesong Shen, Daniel Cremers, Rio Yokota, Mohammad Emtiyaz Khan, Thomas Möllenhoff
We replace AdamW by the Improved Variational Online Newton (IVON) algorithm to finetune large language models.
1 code implementation • 12 Apr 2024 • Etash Guha, Shlok Natarajan, Thomas Möllenhoff, Mohammad Emtiyaz Khan, Eugene Ndiaye
Conformal prediction (CP) for regression can be challenging, especially when the output distribution is heteroscedastic, multimodal, or skewed.
1 code implementation • 27 Feb 2024 • Yuesong Shen, Nico Daheim, Bai Cong, Peter Nickl, Gian Maria Marconi, Clement Bazan, Rio Yokota, Iryna Gurevych, Daniel Cremers, Mohammad Emtiyaz Khan, Thomas Möllenhoff
We give extensive empirical evidence against the common belief that variational learning is ineffective for large neural networks.
1 code implementation • 30 Oct 2023 • Peter Nickl, Lu Xu, Dharmesh Tailor, Thomas Möllenhoff, Mohammad Emtiyaz Khan
Understanding model's sensitivity to its training data is crucial but can also be challenging and costly, especially during training.
1 code implementation • 19 Oct 2023 • Nico Daheim, Thomas Möllenhoff, Edoardo Maria Ponti, Iryna Gurevych, Mohammad Emtiyaz Khan
Models trained on different datasets can be merged by a weighted-averaging of their parameters, but why does it work and when can it fail?
no code implementations • 8 Mar 2023 • Eren Mehmet Kıral, Thomas Möllenhoff, Mohammad Emtiyaz Khan
This simplifies all three difficulties for many cases, providing flexible parametrizations through group's action, simple gradient computation through reparameterization, and updates that always stay on the manifold.
1 code implementation • 4 Oct 2022 • Thomas Möllenhoff, Mohammad Emtiyaz Khan
Sharpness-aware minimization (SAM) and related adversarial deep-learning methods can drastically improve generalization, but their underlying mechanisms are not yet fully understood.
no code implementations • 13 Jul 2021 • Hartmut Bauermeister, Emanuel Laude, Thomas Möllenhoff, Michael Moeller, Daniel Cremers
In contrast to existing discretizations which suffer from a grid bias, we show that a piecewise polynomial discretization better preserves the continuous nature of our problem.
1 code implementation • 27 Feb 2020 • Zhenzhang Ye, Thomas Möllenhoff, Tao Wu, Daniel Cremers
Structured convex optimization on weighted graphs finds numerous applications in machine learning and computer vision.
no code implementations • 4 Dec 2019 • Pierre Bréchet, Tao Wu, Thomas Möllenhoff, Daniel Cremers
We tackle the challenge of disentangled representation learning in generative adversarial networks (GANs) from the perspective of regularized optimal transport (OT).
1 code implementation • 12 May 2019 • Thomas Möllenhoff, Daniel Cremers
We take the novel perspective to view data not as a probability distribution but rather as a current.
no code implementations • 2 May 2019 • Thomas Möllenhoff, Daniel Cremers
Numerous tasks in imaging and vision can be formulated as variational problems over vector-valued maps.
no code implementations • ICCV 2019 • Michael Moeller, Thomas Möllenhoff, Daniel Cremers
The last decade has shown a tremendous success in solving various computer vision problems with the help of deep learning techniques.
no code implementations • 16 Jan 2018 • Thomas Möllenhoff, Zhenzhang Ye, Tao Wu, Daniel Cremers
We present a novel preconditioning technique for proximal optimization methods that relies on graph algorithms to construct effective preconditioners.
1 code implementation • ICLR 2018 • Thomas Frerix, Thomas Möllenhoff, Michael Moeller, Daniel Cremers
Specifically, we show that backpropagation of a prediction error is equivalent to sequential gradient descent steps on a quadratic penalty energy, which comprises the network activations as variables of the optimization.
no code implementations • ICCV 2017 • Thomas Möllenhoff, Daniel Cremers
In this work we show how sublabel-accurate multilabeling approaches can be derived by approximating a classical label-continuous convex relaxation of nonconvex free-discontinuity problems.
1 code implementation • 7 Apr 2016 • Emanuel Laude, Thomas Möllenhoff, Michael Moeller, Jan Lellmann, Daniel Cremers
Convex relaxations of nonconvex multilabel problems have been demonstrated to produce superior (provably optimal or near-optimal) solutions to a variety of classical computer vision problems.
2 code implementations • CVPR 2016 • Thomas Möllenhoff, Emanuel Laude, Michael Moeller, Jan Lellmann, Daniel Cremers
We propose a novel spatially continuous framework for convex relaxations based on functional lifting.
no code implementations • 7 Jul 2014 • Thomas Möllenhoff, Evgeny Strekalovskiy, Michael Moeller, Daniel Cremers
This paper deals with the analysis of a recent reformulation of the primal-dual hybrid gradient method [Zhu and Chan 2008, Pock, Cremers, Bischof and Chambolle 2009, Esser, Zhang and Chan 2010, Chambolle and Pock 2011], which allows to apply it to nonconvex regularizers as first proposed for truncated quadratic penalization in [Strekalovskiy and Cremers 2014].