Gjesteforelesninger og seminarer - Side 10
The efficient redistribution and control of flow is essential in many biological and engineered systems, from our cardiovascular system to plants and soft robotics. Inspired by nature, microfluidic devices with passive valves have been developed to perform a variety of tasks, from cell manipulation to fluid mixing and reaction control, giving rise to the field of soft hydraulics. In this talk, we present two prototypes of passive valves: one harnessing buckling in slender beams and another leveraging snapping in spherical shells, where fluid-induced elastic instabilities are tamed to achieve function. Through a combination of precision desktop-scale experiments, numerical simulations, and theoretical analysis, we rationalize these behaviors and provide design rules based on scaling regimes of material, geometric, and fluid parameters.
In 2008, looking to bound the face vectors of tropical linear spaces, Speyer introduced the g-invariant of a matroid. He proved its coefficients nonnegative for matroids representable in characteristic zero and conjectured this in general. Later, Shaw and Speyer and I reduced the question to positivity of the top coefficient. This talk will overview work being written up with Andy Berget (Western Washington) that proves the conjecture. The first half will be introductory, aiming to define the g-invariant, state the theorem, and compare to other recent work in matroid theory. The second half will look at the proof techniques.
Geometrically, the main ingredient in the proof is a Kempf collapsing constructed from the matroid tautological vector bundles of Berget--Eur--Spink--Tseng, and an initial degeneration thereof. Combinatorially, it is an extension of the definition of external activity to a pair of matroids and a way to compute it using the fan displacement rule on tropical Chern classes. The work of Ardila and Boocher on the closure of a linear space in (P^1)^n is a special case.
Based on J. Kollár and C. Voisin. Flat pushforwards of Chern classes and the smoothability of cycles below the middle dimension. Ann. of Math. (2), 200(2):771–797, 2024.
QOMBINE seminar by Timo Hillmann (Chalmers)
Leiv Rønneberg is a DSTrain Postdoctoral Fellow at the Department of Mathematics and affiliated with The Norwegian Centre for Knowledge-driven Machine Learning (INTEGREAT) at the University of Oslo. Prior to this he was a Research Associate at the MRC Biostatistics Unit at the University of Cambridge, and completed his PhD in Biostatistics at the University of Oslo in 2022. His interests are broadly within the realms of probabilistic machine learning, and connections to Bayesian nonparametrics.
Earthquake, landslide, and glacier instabilities are multi-scale processes that cannot be predicted. When a rock is subjected to large stress, it can develop fractures that grow and connect to form a network. This network of fractures spans the entire rock right before it fails catastrophically. Despite extensive research on rock damage, researchers are still far from being able to predict when a rock will fail. Here, I will show how to use a multi-view convolutional neural network model to identify characteristics of a failing rock. By training a neural network model on images of rock samples exposed to different levels of stress, one can predict how close a rock is to failure. The neural network model outperforms traditional estimates based on fracture density. More importantly, the model provides fundamental insight that may offer precursory information on impending material failure. For instance, according to the model, the angle of the fracture plane relative to the principal loading direction becomes a key factor contributing to failure.
Based on preprints https://arxiv.org/abs/2006.07213 and https://arxiv.org/abs/2406.05382.
QOMBINE Seminar by Chloe Kim (University of Illinois Urbana-Champaign)
QOMBINE seminar by Roy Araiza (University of Illinois at Urbana-Champaign)
Thomas Nagler is a Professor of Computational Statistics \& Data Science at the Department of Statistics, LMU Munich, and a principal investigator at the Munich Center for Machine Learning (MCML).
Previously, he held positions as an assistant professor at TU Delft and Leiden University, following a PhD in Mathematical Statistics from TU Munich.
His research lies at the intersection of mathematical and computational statistics. His group develops statistical methods, establishes theoretical guarantees, and designs scalable algorithms, which are packaged into user-friendly software. He also regularly collaborates with domain experts to tackle challenges in diverse application areas.
C*-seminar by Ulrik Enstad.
Håkon Hoel got his PhD in Numerical Analysis from KTH Stockholm in 2012. He is currently an Associate Professor in Computational Mathematics at the University of Oslo. His main research interests are Monte Carlo methods and numerical methods for stochastic differential equations and data assimilation.
Heat and mass transfer occurs in several natural flows, as well as in many industries. This process is multiscale by nature: the chemistry at the interfaces affects the macroscopic system behavior (e.g. heat transfer coefficients) and the large-scale flow structures determine the details of the local mass and energy fluxes; as example, multicomponent droplet evaporation displays a complex interplay between density, concentration and temperature gradients. The talk will focus on recent advancements in modelling heat and mass transfer in laminar and turbulent flows. We explore various interface capturing methods, including diffuse-interface and sharp-interface approaches, and their coupling with phase-change models to account for interfacial heat and mass exchange. Specifically, we present a formulation based on the diffuse-interface Baer-Nunziato models, where mass transfer is driven by chemical potential disequilibrium. A pressure-based formulation is employed to mitigate time step restrictions while maintaining complex thermodynamic accuracy. Simulations of evaporative flows will be showcased, comparing the performance of diffuse-interface methods with sharp-interface approaches.
Mirror symmetry predicts that the symplectic geometry of a manifold is equivalent to the complex geometry of the mirror manifold. In particular, enumerative mirror symmetry relates quantum cohomology (Gromov-Witten invariants) of a manifold with period integrals of the mirror. A fundamental question in mirror symmetry is to construct mirrors. I will explain Gross-Siebert's intrinsic mirror symmetry which gives a mirror construction using relative quantum cohomology. More precisely, one considers a maximally unipotent monodromy degeneration of a Calabi—Yau manifold and take proj of the degree zero part of a relative quantum cohomology ring associated with the degeneration. Theta functions form a canonical basis of the degree zero part of the relative quantum cohomology ring. Theta functions can be defined in terms of punctured Gromov-Witten invariants of the broken line type. I will explain a variant of intrinsic mirror symmetry using orbifold Gromov-Witten invariants of root stacks, theta functions for general snc pairs and possible applications to questions in mirror symmetry.
Mikko has been working at Natural Resources Institute Finland (Luke) since 2017.
He finished his PhD on spatial point processes in 2022 (University of Jyväskylä).
Since then Mikko has been working on many kinds of remote sensing, forest inventory and forest planning projects.
Also, during 2015-2016 Mikko was working at the National Institute of Health and Welfare as a data manager.
QOMBINE seminar talk by Anna Pachol (University of South-Eastern Norway)
Swimming microorganisms and artificial microswimmers use environmental cues to influence their dynamics, helping them to improve transport, reach specific flow regions, or avoid harmful conditions. Although plankton navigate turbulent flows in nature, it is poorly understood how they do so efficiently. By combining Reinforcement Learning with analytical perturbation theory, we investigate how microswimmers can sense environmental signals and adjust a few control parameters to navigate turbulent flows. This approach helps identify effective navigation strategies for various tasks. We illustrate this with two key tasks: avoiding high-strain regions and navigating against gravity, both essential for plankton in the ocean.
C*-seminar by Nadia Larsen.
Camilla got her PhD in Biostatistics from the University of Cambridge in 2023, where she worked on the development of Bayesian methodology for joint network modelling of high-dimensional data, with applications in autoimmune disease and cancer omics. She is currently a DSTrain Postdoctoral Fellow at The Norwegian Centre for Knowledge-driven Machine Learning Integreat, at the University of Oslo.