TY - THES AB - The polaron model is a basic model of quantum field theory describing a single particle interacting with a bosonic field. It arises in many physical contexts. We are mostly concerned with models applicable in the context of an impurity atom in a Bose-Einstein condensate as well as the problem of electrons moving in polar crystals. The model has a simple structure in which the interaction of the particle with the field is given by a term linear in the field’s creation and annihilation operators. In this work, we investigate the properties of this model by providing rigorous estimates on various energies relevant to the problem. The estimates are obtained, for the most part, by suitable operator techniques which constitute the principal mathematical substance of the thesis. The first application of these techniques is to derive the polaron model rigorously from first principles, i.e., from a full microscopic quantum-mechanical many-body problem involving an impurity in an otherwise homogeneous system. We accomplish this for the N + 1 Bose gas in the mean-field regime by showing that a suitable polaron-type Hamiltonian arises at weak interactions as a low-energy effective theory for this problem. In the second part, we investigate rigorously the ground state of the model at fixed momentum and for large values of the coupling constant. Qualitatively, the system is expected to display a transition from the quasi-particle behavior at small momenta, where the dispersion relation is parabolic and the particle moves through the medium dragging along a cloud of phonons, to the radiative behavior at larger momenta where the polaron decelerates and emits free phonons. At the same time, in the strong coupling regime, the bosonic field is expected to behave purely classically. Accordingly, the effective mass of the polaron at strong coupling is conjectured to be asymptotically equal to the one obtained from the semiclassical counterpart of the problem, first studied by Landau and Pekar in the 1940s. For polaron models with regularized form factors and phonon dispersion relations of superfluid type, i.e., bounded below by a linear function of the wavenumbers for all phonon momenta as in the interacting Bose gas, we prove that for a large window of momenta below the radiation threshold, the energy-momentum relation at strong coupling is indeed essentially a parabola with semi-latus rectum equal to the Landau–Pekar effective mass, as expected. For the Fröhlich polaron describing electrons in polar crystals where the dispersion relation is of the optical type and the form factor is formally UV–singular due to the nature of the point charge-dipole interaction, we are able to give the corresponding upper bound. In contrast to the regular case, this requires the inclusion of the quantum fluctuations of the phonon field, which makes the problem considerably more difficult. The results are supplemented by studies on the absolute ground-state energy at strong coupling, a proof of the divergence of the effective mass with the coupling constant for a wide class of polaron models, as well as the discussion of the apparent UV singularity of the Fröhlich model and the application of the techniques used for its removal for the energy estimates. AU - Mysliwy, Krzysztof ID - 11473 SN - 2663-337X TI - Polarons in Bose gases and polar crystals: Some rigorous energy estimates ER - TY - THES AB - Because of the increasing popularity of machine learning methods, it is becoming important to understand the impact of learned components on automated decision-making systems and to guarantee that their consequences are beneficial to society. In other words, it is necessary to ensure that machine learning is sufficiently trustworthy to be used in real-world applications. This thesis studies two properties of machine learning models that are highly desirable for the sake of reliability: robustness and fairness. In the first part of the thesis we study the robustness of learning algorithms to training data corruption. Previous work has shown that machine learning models are vulnerable to a range of training set issues, varying from label noise through systematic biases to worst-case data manipulations. This is an especially relevant problem from a present perspective, since modern machine learning methods are particularly data hungry and therefore practitioners often have to rely on data collected from various external sources, e.g. from the Internet, from app users or via crowdsourcing. Naturally, such sources vary greatly in the quality and reliability of the data they provide. With these considerations in mind, we study the problem of designing machine learning algorithms that are robust to corruptions in data coming from multiple sources. We show that, in contrast to the case of a single dataset with outliers, successful learning within this model is possible both theoretically and practically, even under worst-case data corruptions. The second part of this thesis deals with fairness-aware machine learning. There are multiple areas where machine learning models have shown promising results, but where careful considerations are required, in order to avoid discrimanative decisions taken by such learned components. Ensuring fairness can be particularly challenging, because real-world training datasets are expected to contain various forms of historical bias that may affect the learning process. In this thesis we show that data corruption can indeed render the problem of achieving fairness impossible, by tightly characterizing the theoretical limits of fair learning under worst-case data manipulations. However, assuming access to clean data, we also show how fairness-aware learning can be made practical in contexts beyond binary classification, in particular in the challenging learning to rank setting. AU - Konstantinov, Nikola H ID - 10799 KW - robustness KW - fairness KW - machine learning KW - PAC learning KW - adversarial learning SN - 2663-337X TI - Robustness and fairness in machine learning ER - TY - THES AB - Plant growth and development is well known to be both, flexible and dynamic. The high capacity for post-embryonic organ formation and tissue regeneration requires tightly regulated intercellular communication and coordinated tissue polarization. One of the most important drivers for patterning and polarity in plant development is the phytohormone auxin. Auxin has the unique characteristic to establish polarized channels for its own active directional cell to cell transport. This fascinating phenomenon is called auxin canalization. Those auxin transport channels are characterized by the expression and polar, subcellular localization of PIN auxin efflux carriers. PIN proteins have the ability to dynamically change their localization and auxin itself can affect this by interfering with trafficking. Most of the underlying molecular mechanisms of canalization still remain enigmatic. What is known so far is that canonical auxin signaling is indispensable but also other non-canonical signaling components are thought to play a role. In order to shed light into the mysteries auf auxin canalization this study revisits the branches of auxin signaling in detail. Further a new auxin analogue, PISA, is developed which triggers auxin-like responses but does not directly activate canonical transcriptional auxin signaling. We revisit the direct auxin effect on PIN trafficking where we found that, contradictory to previous observations, auxin is very specifically promoting endocytosis of PIN2 but has no overall effect on endocytosis. Further, we evaluate which cellular processes related to PIN subcellular dynamics are involved in the establishment of auxin conducting channels and the formation of vascular tissue. We are re-evaluating the function of AUXIN BINDING PROTEIN 1 (ABP1) and provide a comprehensive picture about its developmental phneotypes and involvement in auxin signaling and canalization. Lastly, we are focusing on the crosstalk between the hormone strigolactone (SL) and auxin and found that SL is interfering with essentially all processes involved in auxin canalization in a non-transcriptional manner. Lastly we identify a new way of SL perception and signaling which is emanating from mitochondria, is independent of canonical SL signaling and is modulating primary root growth. AU - Gallei, Michelle C ID - 11626 SN - 2663-337X TI - Auxin and strigolactone non-canonical signaling regulating development in Arabidopsis thaliana ER - TY - THES AB - The complex yarn structure of knitted and woven fabrics gives rise to both a mechanical and visual complexity. The small-scale interactions of yarns colliding with and pulling on each other result in drastically different large-scale stretching and bending behavior, introducing anisotropy, curling, and more. While simulating cloth as individual yarns can reproduce this complexity and match the quality of real fabric, it may be too computationally expensive for large fabrics. On the other hand, continuum-based approaches do not need to discretize the cloth at a stitch-level, but it is non-trivial to find a material model that would replicate the large-scale behavior of yarn fabrics, and they discard the intricate visual detail. In this thesis, we discuss three methods to try and bridge the gap between small-scale and large-scale yarn mechanics using numerical homogenization: fitting a continuum model to periodic yarn simulations, adding mechanics-aware yarn detail onto thin-shell simulations, and quantitatively fitting yarn parameters to physical measurements of real fabric. To start, we present a method for animating yarn-level cloth effects using a thin-shell solver. We first use a large number of periodic yarn-level simulations to build a model of the potential energy density of the cloth, and then use it to compute forces in a thin-shell simulator. The resulting simulations faithfully reproduce expected effects like the stiffening of woven fabrics and the highly deformable nature and anisotropy of knitted fabrics at a fraction of the cost of full yarn-level simulation. While our thin-shell simulations are able to capture large-scale yarn mechanics, they lack the rich visual detail of yarn-level simulations. Therefore, we propose a method to animate yarn-level cloth geometry on top of an underlying deforming mesh in a mechanics-aware fashion in real time. Using triangle strains to interpolate precomputed yarn geometry, we are able to reproduce effects such as knit loops tightening under stretching at negligible cost. Finally, we introduce a methodology for inverse-modeling of yarn-level mechanics of cloth, based on the mechanical response of fabrics in the real world. We compile a database from physical tests of several knitted fabrics used in the textile industry spanning diverse physical properties like stiffness, nonlinearity, and anisotropy. We then develop a system for approximating these mechanical responses with yarn-level cloth simulation, using homogenized shell models to speed up computation and adding some small-but-necessary extensions to yarn-level models used in computer graphics. AU - Sperl, Georg ID - 12358 SN - 2663-337X TI - Homogenizing yarn simulations: Large-scale mechanics, small-scale detail, and quantitative fitting ER - TY - THES AB - In this Thesis, I study composite quantum impurities with variational techniques, both inspired by machine learning as well as fully analytic. I supplement this with exploration of other applications of machine learning, in particular artificial neural networks, in many-body physics. In Chapters 3 and 4, I study quasiparticle systems with variational approach. I derive a Hamiltonian describing the angulon quasiparticle in the presence of a magnetic field. I apply analytic variational treatment to this Hamiltonian. Then, I introduce a variational approach for non-additive systems, based on artificial neural networks. I exemplify this approach on the example of the polaron quasiparticle (Fröhlich Hamiltonian). In Chapter 5, I continue using artificial neural networks, albeit in a different setting. I apply artificial neural networks to detect phases from snapshots of two types physical systems. Namely, I study Monte Carlo snapshots of multilayer classical spin models as well as molecular dynamics maps of colloidal systems. The main type of networks that I use here are convolutional neural networks, known for their applicability to image data. AU - Rzadkowski, Wojciech ID - 10759 SN - 2663-337X TI - Analytic and machine learning approaches to composite quantum impurities ER -