Delft High Performance Computing Centre

Item 1 of 1

[Translate to English:]

Supercomputing power for racing cars and quantum states
[Translate to English:]

High Performace Computing has landed in Delft

Better material models thanks to artificial intelligence

About DHPC

Scientists increasingly need extensive computing power to solve complex problems in physics, mechanics and dynamics. The Delft High Performance Computing Centre (DHPC) deploys the infrastructure (hardware, software and staff) for TU Delft that is capable of complex analysis and modelling for researchers. At the same time we provide Bachelor, Master and PhD students with hands-on experience using the tools they will need in their careers.

Both high-performance simulations and high-performance data science are evolving rapidly and the combination of these techniques will lead to completely new insights into science and engineering, an increase in innovation, and the training of high-performance computing engineers for the future.

Due to the rapidly evolving hardware and tools for numerical simulations, HPC has significantly changed the way fundamental research is conducted at universities. Simulations not only replace experiments, but also add very valuable fundamental insights. We see the results in all kinds of disciplines, such as materials science, fluid dynamics, quantum mechanics, design optimization, big data mining and artificial intelligence.

Subscribe to our Newsletter

News

19 December 2023

Better material models thanks to artificial intelligence

24 November 2023

Expansion DelftBlue ready for use in January 2024

02 November 2023

Martin van Gijzen has been appointed as the new Scientific Director of the DHPC

Effective September 1, Martin van Gijzen has been appointed as the new Scientific Director of the Delft High Performance Computing Centre (DHPC). Van Gijzen brings a wealth of knowledge and expertise to the centre, and he is committed to further developing the centre into a knowledge hub and go-to portal for high performance computing expertise at the TU Delft.

Agenda

19 April 2024 12:30 till 13:15

[NA] Alena Kopaničáková : Enhancing Training of Deep Neural Networks Using Multilevel and Domain Decomposition Strategies

The training of deep neural networks (DNNs) is traditionally accomplished using stochastic gradient descent or its variants. While these methods have demonstrated certain robustness and accuracy, their convergence speed deteriorates for large-scale, highly ill-conditioned, and stiff problems, such as ones arising in scientific machine learning applications. Consequently, there is a growing interest in adopting more sophisticated training strategies that can not only accelerate convergence but may also enable parallelism, convergence control, and automatic selection of certain hyper-parameters.
In this talk, we propose to enhance the training of DNNs by leveraging nonlinear multilevel and domain decomposition strategies. We will discuss how to construct a multilevel hierarchy and how to decompose the parameters of the network by exploring the structure of the DNN architecture, properties of the loss function, and characteristics of the dataset. Furthermore, the dependency on a large number of hyper-parameters will be reduced by employing a trust-region globalization strategy. The effectiveness of the proposed training strategies will be demonstrated through a series of numerical experiments from the field of image classification and physics-informed neural networks.

References:
[1] A. Kopaničáková, H. Kothari, G. Karniadakis and R. Krause. Enhancing training of physics-informed neural networks using domain-decomposition based preconditioning strategies. Under review, 2023.
[2] S. Gratton, A. Kopaničáková, and Ph. Toint. Multilevel Objective-Function-Free Optimization with an Application to Neural Networks Training. SIAM, Journal on Optimization (Accepted), 2023.
[3] A. Kopaničáková. On the use of hybrid coarse-level models in multilevel minimization methods. Domain Decomposition Methods in Science and Engineering XXVII (Accepted), 2023.
[4] A. Kopaničáková, and R. Krause. Globally Convergent Multilevel Training of Deep Residual Networks. SIAM Journal on Scientific Computing, 2022.

23 April 2024 09:30 till 16:30

Command line and DelftBlue basics 23 April 2024

After taking this course, participants will be able to fully engage in remote HPC systems using the command line.

03 May 2024 12:30 till 13:15

[NA] Jakob Zech: Nonparametric Distribution Learning via Neural ODEs

In this talk, we explore approximation properties and statistical aspects of Neural Ordinary Differential Equations (Neural ODEs). Neural ODEs are a recently established technique in computational statistics and machine learning, that can be used to characterize complex distributions. Specifically, given a fixed set of independent and identically distributed samples from a target distribution, the goal is either to estimate the target density or to generate new samples. We first investigate the regularity properties of the velocity fields used to push forward a reference distribution to the target. This analysis allows us to deduce approximation rates achievable through neural network representations. We then derive a concentration inequality for the maximum likelihood estimator of general ODE-parametrized transport maps. By merging these findings, we are able to determine convergence rates in terms of both the network size and the number of required samples from the target distribution. Our discussion will particularly focus on target distributions within the class of positive \(C^k\) densities on the \(d\)-dimensional unit cube \([0,1]^d\).

31 May 2024 12:30 till 13:15

[NA] Hyea Hyun Kim: Partitioned neural network approximation to partial differential equations and its training performance enhancement utilizing domain decomposition algorithms

With the success of deep learning technologies in many scientific and engineering applications, neural network approximation methods have emerged as an active research area in numerical partial differential equations. However, the new approximation methods still need further validations on their accuracy, stability, and efficiency so as to be used as alternatives to classical approximation methods. In this talk, we first introduce partitioned neural network approximation to partial differential equations, where neural network functions localized in each small subdomains are employed as a solution surrogate in order to reduce the approximation and optimization errors in the standard single large neural network approximation. The parameters in each local neural network function are then optimized to minimize the corresponding cost function to the model problem. To enhance the parameter training efficiency further, iterative algorithms for the partitioned neural network function can be developed by utilizing classical domain decomposition algorithms and their convergence theory. We finally present promising features in this new approach as a way of enhancing the neural network solution accuracy, stability, and efficiency with some supporting numerical results.

12 June 2024 09:30 till 16:30

High-performance computing using OpenMP 12 June 2024