Speaker: Morten Brun

Place: Room 3137, 3rd floor Høyteknologisenteret

Time: Friday November 13, 2015, from 10:15am to 11:00am

Abstract:
I will explain how we have used persistent homology to examine meteorological data. In particular we have studied patterns of weather fronts that have reached Bergen. This is an ongoing joint project between the department of geophysics and the department of mathematics here in Bergen. If time permits, at the end of the talk I will sketch how we envision to use topological data analysis in the study of the stress imposed on cod liver by pollution.

Additional material: Flyer

Speaker: Junyong You

Place: Room 3137, 3rd floor Høyteknologisenteret

Time: Friday October 16, 2015, from 10:15am to 11:00am

Abstract:
Visual signals are often distorted due to unavoidable factors, e.g., lossy compression, transmission over error prone networks. Accurate assessment of visual quality plays an important role in multimedia services in order to provide the best viewing experience to end-users. Traditionally used quality metrics, e.g., PSNR (peak signal-to-noise ratio) often cannot accurately represent the actual distortions perceived by human users. For example, the distorted images below all have same PSNR values with respect to the original reference image, while they present different levels of perceived distortions. This talk presents methodologies of perceived visual quality assessment, including visual mechanism modelling and applications in objective visual quality metrics. A video quality metric driven visual attention and foveation mechanism that was developed by the presenter will be introduced as an example. Applications in other areas will also be briefly mentioned.

Additional material: Flyer

Speaker: Jan Byska

Place: Room 3137, 3rd floor Høyteknologisenteret

Time: Friday September 18, 2015, from 10:15am to 11:00am

Abstract:
Proteins are highly complex systems contributing to all functions in living organisms. Understanding their function helps to reveal the fundamentals of biochemical processes that are taking place in living cells. This search process, however, consumes an enormous amount of time and resources, due to the internal protein dynamics, which constantly change structural and biochemical characteristics of the protein over time. Thus, several in-silico predictive methods already appeared, which focus mostly on the analysis of the protein structure. In this talk we will mainly focus on methods for interactive exploration of protein inner void space including its geometry, physico chemical properties and dynamics to evaluate or modify the reactivity of the protein with other small molecules.

Additional material: Flyer

Speaker: Ivan Viola

Place: Room 3137, 3rd floor Høyteknologisenteret

Time: Friday August 28, 2015, from 10:15am to 11:00am

Abstract:
The study of biological processes carried out in living organisms is among the central foci of modern science. The field is nowadays by large extent computational, there are many kinds of digital models that characterize particular aspects of life. To provide a comprehensive view on biological phenomena, visualization offers itself for integrating multiple models into one visual environment. One of the interesting challenges, associated with such a visual integration, is to communicate phenomena that are simultaneously described on several spatial and temporal scales. In my talk I will discuss visualization techniques that bridge five orders of magnitude of spatial scale by interactively displaying structural information from a single atom level up to entire bacteria cell size with complete molecular machinery When dealing with simulation of molecular interactions in living organisms, agent-based models are used to model long sequences of Brownian motion that randomly carry out several physiologically relevant events. One such event occurs on average every thousandth simulation iteration in the agent based simulation. In order to perceive random motion but also to see relevant events simultaneously, we bridge two temporal scales that are three orders of magnitude apart from each other. The approach is based on a special timelapse approach inspired from scientific animation techniques. In the last part of my talk I will present a technique for interactive 3D visualization of molecular reaction pathways that is controlled in a top-down manner by quantitative simulation and is simultaneously covisualized. In such visual environment the viewer can interactively control the visualization as well as simulation parameters.

Additional material: Flyer

Speaker: Prof. Dr.-Ing. Lars Linsen

Place: Big Auditorium, 2rd floor, Høyteknologisenteret (data blokk)

Time: Thursday May 20, 2015, from 14:00am to 15:30am

Abstract:
Mathematical models are used for the description and the understanding of phenomena in all sciences. Numerical simulations support the validation of the models and data assimilation purposes. Such simulations often depend on a number of simulation parameters and initial configurations. The selection of these parameters and configurations is often not exactly known or their impact is part of the underlying research tasks. Therefore, multiple simulation runs with varying parameter settings or ensembles of simulations with varying configurations are executed. The analysis of such simulation ensembles is complex, especially when each simulation run represents a four-dimensional spatio-temporal phenomenon. The amount of data of a simulation ensemble often adds up to hundreds of Gigabytes or even Terabytes. The analysis of such complex data is no longer possible without the use of computers. On the other hand, such an analysis typically requires the expertise of a human. As visual representations are intuitive and can be processed efficiently by humans, it is a suitable approach to combine visual representations and interaction mechanisms with automatic analysis steps. In this talk, I will present novel visualization methods that allow for an interactive comparative analysis of such large and complex data stemming from simulation ensembles. The methods are applied for the analysis of data stemming from biology, climate research, astrophysics, and medicine.

Additional material: Flyer

Speaker: Julius Parulek (Adjunct associate professor at UIB)

Place: Room 3137, 3rd floor, Høyteknologisenteret (data blokk)

Time: Friday April 10, 2015, from 10.15am to 11.15am

Abstract:
In situ analysis of streamed acoustic data is crucial for sub sea environmental monitoring tasks. One of the crucial applications is represented by detection of oil and gas leakages from subsea installations. On the software level, the monitoring system is based on semi-automatic and also on fully-automatic data analysis pipeline, which incorporate challenges related to big data analysis, visualization and machine learning. In this presentation, I will aim mainly at the role of visualization and machine learning in the detection of gas plumes.

Additional material: Flyer

Speaker: Barbora Kozlikova (Assistant Professor at Masaryk University in Brno)

Place: Room 3137, 3rd floor, Høyteknologisenteret (data blokk)

Time: Friday March 13, 2015, from 10.15am to 11.15am

Abstract:
The talk aims to present the background of the CAVER project which is dedicated to analysis and visualization of proteins. Our algorithms are focusing namely on the detection and classification of inner void space. This long-term project is baed on a tight cooperation with the group of biochemists. In the talk I'll concentrate on the main principle and future challenges of this project.

Additional material: Flyer

Speaker: Erlend Hodneland (MedViz PostDoc at CMR.no)

Place: Room 3137, 3rd floor, Høyteknologisenteret (data blokk)

Time: Friday February 06, 2015, from 10.15am to 11.15am

Abstract:
Quantitative analysis of DCE-MRI data from the kidney has the potential to become a clinically highly useful imaging modality for the non-invasive investigation of organ function. There are major challenges related to perfusion imaging of the human kidney that I want to address. The processing pipeline has several components: choice of MR sequence, breathing pattern, conversion to tracer concentrations, motion correction, segmentation of kidney and arterial input function (AIF), and compartment modelling for estimating the filtration and perfusion. The reproducibility between repeated MR scans is high, but a comparison to clinical standards for kidney filtration reveals that the MR estimates for filtration are still not within the range for active clinical use. In this presentation I will address the challenges related to the processing pipeline.

Additional material: Flyer