Guest lectures and seminars - Page 6
The frictional behavior of surfaces is a problem of great scientific and practical significance. Recent progress in molecular scale modeling allows us to determine the coefficient of friction for nanoscale surfaces from first principles using molecular dynamics modeling. However, inverse design, that is, designing surfaces with specific frictional propeties is still a complex and largely unsolved challenge in part due to the enormous space of possible surface configurations. Here, we demonstrate how we can use physical forward modeling to find the frictional properties of a set of surfaces that can serve as a training set to design machine learning models. In this talk, we demonstrate both discriminative and generative models for frictional surface design and analyze what physical principles the machine learning models have learned in this process.
his talk discusses a nonparametric inference framework for occupation time curves derived from wearable device data. Such curves provide the total time a subject maintains activity above a given level as a function of that level. Taking advantage of the monotonicity and smoothness properties of these curves, we develop a likelihood ratio approach to construct confidence bands for mean occupation time curves. An extension to fitting concurrent functional regression models is also developed. Application to wearable device data from an ongoing study of an experimental gene therapy for mitochondrial DNA depletion syndrome will be discussed. Based on joint work with Hsin-Wen Chang (Academia Sinica).
During this cold winter in Oslo, we certainly all have experienced the unsteady nature of friction and the sudden loss of stress bearing capacity that initiates catastrophic sliding. A similar kind of frictional rupture arises at the onset of a wide variety of natural disasters that includes earthquakes, landslides, as well as snow avalanches. In this talk, I will focus on the incipient stage of these catastrophic events that is characterized by the local nucleation of failure and its rapid propagation towards intact regions of the material. I will discuss how the analogy to fracture mechanics can be exploited to describe the dynamics of these rupture fronts and develop quantitative models to characterize the onset of failure in geomaterials.
Donaldson-Thomas and Pandharipande-Thomas theory are two approaches to counting curves on projective threefolds in terms of their moduli spaces of sheaves. An important special case in understanding the DT/PT correspondence the equivariant geometry of affine three-space with the natural coordinate action of the rank 3 torus. I will show how one can use new wall-crossing techniques to prove the equivariant K-theoretic DT/PT correspondence in this situation, which was previously known only in the Calabi-Yau limit.
This is part of an ongoing project with Felix Thimm and Henry Liu in which we aim to prove wall-crossing for virtual enumerative invariants associated to equivariant CY3 geometries by extending a vertex algebra formalism for wall-crossing developed by Joyce.
Bone stress injuries affect athletic populations who undertake activities in which bones are repeatedly loaded. In order to understand and reduce the risk of bone stress injuries, we need to quantify the loading experienced by the bones during activities such as running. Bone loading is difficult to quantify as the magnitudes of stress are influenced to a large extent by the magnitude of muscular forces acting on the bone. Musculoskeletal modelling, ranging from very simple to very complex approaches, can be used to estimate the internal loading experienced by the bone during running. This has allowed us to explore factors such as speed, slope and step length and their influence on bone loading during running. However, in order to truly understand risk of stress injuries this needs to be taken out of the lab and in-field. This talk will consider what we know and the limitations to current understanding.