PERSON RESPONSIBLE: MARK ABRAHAM

GROMACS is a versatile package to perform molecular dynamics, i.e. simulate the Newtonian equations of motion for systems with hundreds to millions of particles. It is primarily designed for biochemical molecules like proteins, lipids and nucleic acids that have a lot of complicated bonded interactions, but since GROMACS is extremely fast at calculating the nonbonded interactions (that usually dominate simulations) many groups are also using it for research on non-biological systems, e.g. polymers.

PERSON RESPONSIBLE: ROBERTO DI REMIGIO

VeloxChem solves the Schrödinger equation to study the electronic structure of molecular systems. The program can compute molecular energies and simulate the response of molecules subject to external electromagnetic fields. VeloxChem is built to exploit the aggregate resources of computing systems: from laptops to clusters. It can handle thousands of atoms and leverages a hybrid Python/C++ programming paradigm for fast development without sacrificing performance

PERSON RESPONSIBLE: JING GONG

Nek5000 is an open-source code for the simulation of incompressible flow. Nek5000 is widely used in a broad range of applications, including the study of thermal hydraulics in nuclear reactor cores, the modeling of ocean currents and the simulation of combustion in mechanical engines. The Nek5000 discretization scheme is based on the spectral-element method. In this approach, the incompressible Navier-Stokes equations are discretized in space by using high-order, weighted residual techniques employing tensor-product polynomial bases.

PERSON RESPONSIBLE: QIANG LI

ESSENSE is a research code for flow calculations by solving the compressible Navier-Stokes equations. Using a high order finite difference method in combination with summation-by-parts operators and weak boundary conditions makes it possible to efficiently and reliably handle large problems on structured grids for reasonably smooth geometries.

PERSON RESPONSIBLE:  Roberto Di Remigio

OpenMolcas is a molecular electronic structure package focused on multiconfigurational methods. The program can compute molecular ground- and excited-state energies with state-of-the-art accuracy for many complex electronic structure problems. OpenMolcas is open source (LGPL licensed) with an open development workflow. It is a large code, written in a mix of languages: primarily Fortran (77 and 90) with some C and external modules in C++. OpenMolcas has a large, international user base. It is among one of the most used software packages for multiconfigurational quantum chemistry, in Europe and around the world.

PERSON RESPONSIBLE: Jing Gong and Arten Zhmurov

BCPNNSim is an open-source code for scalable parallel simulation of Bayesian Confidence Propagation Neural Networks. A BCPNN module features Bayesian-Hebbian synaptic plasticity as well as structural plasticity for unsupervised and supervised learning. The code has been applied successfully to simulation of reduced brain models of e.g. associative memory and to Machine Learning benchmarks like MNIST, SVHN and CIFAR-10. Current focus is on parallel implementation on GPU and clusters of GPUs via MPI. Extensions are planned for spiking units, stacked layers, and improved support for multi-network architectures.

With RISE ( https://www.ri.se/en), ENCCS is helping build the next generation of Swedish language models from the BERT family. Currently we are training a DeBERTa-large model for Swedish with only a small amount of data by using transfer learning from the equivalent English models. This project is running as a pilot access in the early life of the BerzeLiUS AI supercomputer ( https://www.nsc.liu.se/systems/berzelius/).

Swedish Speech Synthesis
With Voxo AB ( https://www.voxo.ai/), ENCCS is using machine learning to develop Swedish-language speech-synthesis machine-learning models based on the Tacotron2 family of speech-synthesis model architectures. It will be a key component of Voxo’s conversational assistant capable of providing information in real time in response to spoken natural-language questions. It will be capable of learning to pronounce jargon relevant to particular domains, such as banking. It will generate audio streams quickly, so that users will be comfortable with natural conversation flow, without pauses for generating long replies. This project is using HPC time awarded via the PRACE SHAPE project (https://prace-ri.eu/hpc-access/shape-access/) on the German GPU-boosted supercomputer JUWELS ( https://www.fz-juelich.de/ias/jsc/EN/Home/home_node.html).

Traffic flow is a major contributor to the emission of greenhouse gases. Municipalities around the world have invested in what is known as Intelligent Transportation Systems (ITS) to optimize the traffic flow and reduce the emission. Preemptive traffic prediction can significantly help ITS to provide a better organization of the flow across cities. Machine learning methods have proved their usability in forecasting traffic flow. In a joint undertaking, Trafikverket, KTH, and ENCCS experts Mr. Christian Edfjall, Assoc. Prof. Xiaoliang Ma, and Dr. Hossein Ehteshami aim to predict traffic along the E4S highway using deep learning (DL) methods. In the first part of the project, the traffic flow will be modeled using appropriate DL models. In the second part, the already verified DL models will be coupled to a traffic-based pollution theory in order to study the dynamic of traffic pollutants at the E4S. The upcoming model will primarily be developed in Tensorflow (Keras API) in combination with in-house codes.

Sentiment analysis of texts and speech-to-text transformation are active areas of research and development in the field of Artificial Intelligence (AI). Two main ingredients of such endeavor are high-quality training data and a suitable deep neural network (NN) model, which uses the training data to tune its parameters. The reward is a system that not only can turn (almost) any speech to text but also “understand” the context and sentiment in it. Modern phones, laptops, and other gadgets are already using this technology to serve their owners. Nonetheless, most of the development in this field emerged around the English language model. Currently, there is a void for a Swedish counterpart. As a response to this void, the data lab (KBLab) at the National Library of Sweden (Kungliga Biblioteket) developed the KB-BERT model, the Swedish trained transformer model based on Google BERT architecture. KB-BERT, trained on the vast amount of high-quality data solely available at KB, proved to be a game-changer in this area.” and add a sentence that that ENCCS is assisting KB in this project that will run on Vega.

PERSON RESPONSIBLE: QIANG LI

ICON is a highly versatile next-generation global climate model. The model solves the equations of motion for the atmosphere and ocean and couple these together with unresolved processes such as small scale turbulence, cloud microphysics and radiation. The model code has been designed with parallelization in mind allowing scientists to achieve unprecedented kilometer-scale resolutions, enabling simulations of individual clouds and ocean eddies even on global grids.

PERSON RESPONSIBLE: QIANG LI

EC-Earth is a global climate model system based on the idea to use the world-leading weather forecast model of the ECMWF (European Centre of Medium Range Weather Forecast) in its seasonal prediction configuration as the base of climate model. The model system can be used in several configurations including the classical climate model (atmosphere, soil, ocean, sea ice) and Earth System configurations (adding atmospheric chemistry and aerosols, ocean bio-geo-chemistry, dynamic vegetation and a Greenland ice sheet). The model is developed by the European EC-Earth consortium with SMHI as core partner leading the development and other Swedish partners from the universities of Lund, Stockholm, Gothenburg and Uppsala. The model in its different configurations and resolutions is used for climate change projections, predictions and process studies.