no code implementations • 28 Mar 2024 • Tim Johnston, Nikolaos Makras, Sotirios Sabanis
Recent advances in stochastic optimization have yielded the interactive particle Langevin algorithm (IPLA), which leverages the notion of interacting particle systems (IPS) to efficiently sample from approximate posterior densities.
no code implementations • 22 Nov 2023 • Stefano Bruno, Ying Zhang, Dong-Young Lim, Ömer Deniz Akyildiz, Sotirios Sabanis
As a result, we obtain the best known upper bound estimates in terms of key quantities of interest, such as the dimension and rates of convergence, for the Wasserstein-2 distance between the data distribution (Gaussian with unknown mean) and our sampling algorithm.
no code implementations • 15 Nov 2023 • Iosif Lytras, Sotirios Sabanis
In this article we propose a novel taming Langevin-based scheme called $\mathbf{sTULA}$ to sample from distributions with superlinearly growing log-gradient which also satisfy a Log-Sobolev inequality.
no code implementations • 23 Mar 2023 • Ö. Deniz Akyildiz, Francesca Romana Crucinio, Mark Girolami, Tim Johnston, Sotirios Sabanis
We achieve this by formulating a continuous-time interacting particle system which can be seen as a Langevin diffusion over an extended state space of parameters and latent variables.
no code implementations • 19 Jan 2023 • Tim Johnston, Iosif Lytras, Sotirios Sabanis
In this article we consider sampling from log concave distributions in Hamiltonian setting, without assuming that the objective gradient is globally Lipschitz.
1 code implementation • 24 Oct 2022 • Dong-Young Lim, Ariel Neufeld, Sotirios Sabanis, Ying Zhang
We introduce a new Langevin dynamics based algorithm, called e-TH$\varepsilon$O POULA, to solve optimization problems with discontinuous stochastic gradients which naturally appear in real-world applications such as quantile estimation, vector quantization, CVaR minimization, and regularized optimization problems involving ReLU neural networks.
1 code implementation • 21 Oct 2021 • Ömer Deniz Akyildiz, Connor Duffin, Sotirios Sabanis, Mark Girolami
Through embedding uncertainty inside of the governing equations, finite element solutions are updated to give a posterior distribution which quantifies all sources of uncertainty associated with the model.
1 code implementation • 19 Jul 2021 • Dong-Young Lim, Ariel Neufeld, Sotirios Sabanis, Ying Zhang
To illustrate the applicability of the main results, we consider an example from transfer learning with ReLU neural networks, which represents a key paradigm in machine learning.
1 code implementation • 28 May 2021 • Dong-Young Lim, Sotirios Sabanis
We present a new class of Langevin based algorithms, which overcomes many of the known shortcomings of popular adaptive optimizers that are currently used for the fine tuning of deep learning models.
no code implementations • 2 Jul 2020 • Sotirios Sabanis, Ying Zhang
We are thus able to provide theoretical guarantees for the algorithm's convergence in (standard) Wasserstein distances for both convex and non-convex objective functions.
no code implementations • 25 Jun 2020 • Attila Lovas, Iosif Lytras, Miklós Rásonyi, Sotirios Sabanis
We offer a new learning algorithm based on an appropriately constructed variant of the popular stochastic gradient Langevin dynamics (SGLD), which is called tamed unadjusted stochastic Langevin algorithm (TUSLA).
no code implementations • 13 Feb 2020 • Ömer Deniz Akyildiz, Sotirios Sabanis
We provide a nonasymptotic analysis of the convergence of the stochastic gradient Hamiltonian Monte Carlo (SGHMC) to a target measure in Wasserstein-2 distance without assuming log-concavity.
no code implementations • 4 Oct 2019 • Ying Zhang, Ömer Deniz Akyildiz, Theodoros Damoulas, Sotirios Sabanis
In this paper, we are concerned with a non-asymptotic analysis of sampling algorithms used in nonconvex optimization.
no code implementations • 3 Jun 2019 • Mathias Barkhagen, Brian Fleming, Sergio Garcia Quiles, Jacek Gondzio, Joerg Kalcsics, Jens Kroeske, Sotirios Sabanis, Arne Staal
It is based on a novel concept called portfolio dimensionality that connects diversification to the non-Gaussianity of portfolio returns and can typically be defined in terms of the ratio of risk measures which are homogenous functions of equal degree.
no code implementations • 30 May 2019 • Ngoc Huy Chau, Éric Moulines, Miklos Rásonyi, Sotirios Sabanis, Ying Zhang
We consider the problem of sampling from a target distribution, which is \emph {not necessarily logconcave}, in the context of empirical risk minimization and stochastic optimization as presented in Raginsky et al. (2017).
no code implementations • 24 Jul 2016 • Raj Kumari Bahl, Sotirios Sabanis
In this paper, we are concerned with the valuation of Catastrophic Mortality Bonds and, in particular, we examine the case of the Swiss Re Mortality Bond 2003 as a primary example of this class of assets.