@article{22734, abstract = {Silver selenide (Ag2Se) is a promising near-room-temperature thermoelectric material, but its narrow stoichiometric window and β–α phase transition complicate reproducible microstructure control. Here, we present a mismatch-assisted microstructure engineering strategy in which Ag2Se particles are treated with polyanionic ZnSe complexes and consolidated through the β–α transition to introduce ZnSe nanoprecipitates, Ag2Se/ZnSe interfaces, and local strain fields. The crystallographic mismatch between ZnSe and Ag2Se, together with the Zn2+/Ag+ size difference, amplifies phase-transition-induced deformation and promotes high-density dislocations with periodic strain modulations. This defect architecture suppresses grain coarsening, removes excess Ag, limits Ag-interstitial formation, and reduces lattice thermal conductivity through lattice softening and multiscale phonon scattering. Ag2Se–4%ZnSe nanocomposites achieve a peak zTmax of 1.13 at 369 K and a zTavg of 1.08 from 300 to 380 K, demonstrating mismatch-driven defect engineering through the β–α phase transition as a route for optimizing Ag2Se-based thermoelectrics.}, author = {Liu, Yu and Kleinhanns, Tobias and Spadaro, Maria Chiara and Genç, Aziz and Horta, Sharona and Navita, Navita and Costanzo, Tommaso and Dutkiewicz, Ewelina and Arbiol, Jordi and Hong, Min and Ibáñez, Maria}, issn = {2380-8195}, journal = {ACS Energy Letters}, number = {8}, pages = {5752--5762}, publisher = {American Chemical Society}, title = {{Exploiting mismatch strain and the β–α phase transition for microstructural engineering in thermoelectric Ag2Se}}, doi = {10.1021/acsenergylett.6c01499}, volume = {11}, year = {2026}, } @phdthesis{22694, abstract = {We develop and employ techniques from equivariant algebraic K-theory and related invariants in the context of geometric representation theory, in both arithmetic and topological situations. We showcase the use of such techniques on the affine Grassmannian Gr, a space of fundamental interest in the geometric Langlands program. It is a deep development of mathematics of the last century that many concrete, yet combina- torially complex algebraic problems may be effectively studied through the lens of algebraic geometry. The objects of interest can be often realized as cohomological invariants of algebraic varieties, and good understanding of their geometry sheds light into the original questions. Such techniques have seen immense applications in the Langlands program, where they go under the label of geometric representation theory. One source of powerful invariants in algebraic geometry comes from algebraic K-theory, Hochschild homology, and their relatives. These localizing invariants contain large amount of information, but are quite hard to compute. For this reason, their usage in geometric representation theory has been limited. The aim of this thesis is to showcase how to control such invariants in the situations of interest and use them to obtain new insights. We start by reinterpreting equivariant Hochschild homology in terms of functions on certain fixed-point schemes, which are of independent interest. We compare it to equivariant K-theory via the trace map. We give new computations and comparisons of such invariants of affine Schubert varieties in Gr, including arithmetic situations. We show that they behave much better than expected. We finally utilize this circle of ideas in a purely topological setting. We describe the varying fixed points of the extended torus action on the affine Grassmannian, and use it to compute its equivariant topological K-theory ring. The answer is nontrivial and verifies an outstanding conjecture in the subject. We compare, partly conjecturally, the resulting K-theory ring to the completed center of an integral even hybrid quantum group and its deformed quantum category O. This gives a genuine application of our computations in pure representation theory.}, author = {Löwit, Jakub}, issn = {2663-337X}, pages = {185}, publisher = {Institute of Science and Technology Austria}, title = {{Equivariant K-theory of affine Grassmannians in representation theory and arithmetic}}, doi = {10.15479/AT-ISTA-22694}, year = {2026}, } @article{22693, abstract = {We study torus-equivariant algebraic K-theory of affine Schubert varieties in the perfect affine Grassmannians over Fp. We further compare it to the torus-equivariant Hochschild homology of perfect complexes, which has a geometric description in terms of global functions on certain fixed-point schemes. We prove that Fp-linearly, this comparison is an isomorphism. Our approach is quite constructive, resulting in new computations of these K-theory rings. We establish various structural results for equivariant perfect algebraic K-theory on the way; we believe these are of independent interest.}, author = {Löwit, Jakub}, issn = {1431-0643}, journal = {Documenta Mathematica}, keywords = {equivariant algebraic K-theory, perfection in positive characteristic, affine Grassmannian, affine Schubert varieties, Dennis trace map, equivariant Hochschild homology, fixed-point schemes, toric varieties}, publisher = {EMS Press}, title = {{Equivariant K-theory, affine Grassmannian and perfection}}, doi = {10.4171/dm/1064}, year = {2026}, } @phdthesis{22684, abstract = {Contact electrification (CE) is a simple yet elusive phenomenon that occurs when two materials come into contact and separate, leaving behind net electrical charge. Despite its ubiquity, the microscopic origin of CE remains unclear. In this thesis, we investigate CE from three complementary perspectives: developing a quantitative method to measure charge at the nanoscale, exploring the dynamic behavior of charge on insulating surfaces, and uncovering the role of mechanical history in forming a triboelectric series. In the first part, we establish a rigorous framework for converting qualitative Kelvin probe force microscopy (KPFM) voltage maps into quantitative charge density distributions. Using finite element method (FEM) simulations, we determine the point-spread function of the KPFM tip–sample geometry and demonstrate that the true surface charge can be reconstructed by numerical deconvolution. This procedure enables the recovery of both the magnitude and sign of charge density with high fidelity, resolving nanoscale features that are otherwise obscured. Applying the method to contact-charged SiO$_2$ surfaces, we show that existing analytical approximations, such as parallel plate or spherical models, can miscalculate charge magnitude by orders of magnitude. Our hybrid FEM/KPFM approach therefore provides a fast and general method to convert qualitative KPFM signals into quantitative charge data, enabling nanoscale charge mapping under realistic experimental conditions. In the second part, we study the temporal stability of CE-induced charges and identify the key material factors that determine whether KPFM can capture meaningful charge patterns. Through time-resolved experiments combining a custom-built transfer system with both microscopic and macroscopic measurements, we demonstrate that only the best insulators, such as SiO$_2$, preserve CE charge long enough for stationary imaging. For less conductive polymers, such as PDMS, charge decays within the duration of a single KPFM scan due to bulk conduction. Using a simple capacitor-based model, we reproduce the observed decay dynamics and confirm that the transferred charge decays characteristic to the sample's bulk conductivity. Further, we always observe homogeneous charge transfer. In the third part, we address the question: can we form a triboelectric series with identical materials? Using controlled repetitive contact experiments, we show that nominally identical materials can progressively order themselves into a triboelectric series, where surfaces with more contact history charge negatively relative to fresher ones. By constructing a minimal model based on this ``contact bias'', we replicate the evolution from random to ordered charging observed in experiments. Supporting surface analyses, including atomic force microscopy, reveal that repeated contact induces nanoscale morphological changes, suggesting a mechanism tightly coupled to mechanical strain. These results highlight the crucial role of surface history and nanoscale mechanics in dictating charge transfer, motivating further exploration of mechanisms such as mechanochemical bond cleavage and flexoelectric polarization.}, author = {Pertl, Felix}, isbn = {978-3-99078-083-1}, issn = {2663-337X}, pages = {107}, publisher = {Institute of Science and Technology Austria}, title = {{Experimental probing of nanoscale charge features and surface morphology changes during tribocharging}}, doi = {10.15479/AT-ISTA-22684}, year = {2026}, } @article{21762, abstract = {Bacteria, like eukaryotes, use conserved cytoskeletal systems for intracellular organization. The plasmid-encoded ParMRC system forms actin-like filaments that segregate low–copy number plasmids. In multicellular cyanobacteria such as Anabaena sp., we found that a chromosomally encoded ParMR system has evolved into a cytoskeletal system named CorMR with a function in cell shape control rather than DNA segregation. Live-cell imaging, in vitro reconstitution, and cryo–electron microscopy revealed that CorM formed dynamically unstable, antiparallel double-stranded filaments that were recruited to the membrane by CorR through an amphipathic helix conserved in multicellular cyanobacteria. CorMR filaments were regulated by MinC, which excluded them from the poles and division plane. Comparative genomics indicated that the repurposing of ParMR and Min systems coevolved with cyanobacterial multicellularity, highlighting the evolutionary plasticity of cytoskeletal systems in bacteria.}, author = {Springstein, Benjamin L and Javoor, Manjunath and Megrian, Daniela and Hajdu, Roman and Hanke, Dustin M. and Zens, Bettina and Weiss, Gregor L. and Schur, Florian Km and Loose, Martin}, issn = {1095-9203}, journal = {Science}, number = {6795}, publisher = {AAAS}, title = {{Repurposing of a DNA segregation machinery into a cytoskeletal system controlling cell shape}}, doi = {10.1126/science.aea6343}, volume = {392}, year = {2026}, } @phdthesis{22744, author = {Javoor, Manjunath}, isbn = {978-3-99078-090-9 }, issn = {2663-337X}, keywords = {Actin cytoskeleton, Cell migration, cryo-electron tomography}, pages = {121}, publisher = {Institute of Science and Technology Austria }, title = {{Large-scale imaging of cellular actin networks at single filament resolution using montage cryo-electron tomography}}, doi = {10.15479/AT-ISTA-22744}, year = {2026}, } @phdthesis{22664, abstract = {The widespread adoption of apps like Whatsapp and Signal has translated into billions of people all around the world communicating on a regular basis by making use of services that offer end-to-end encryption and even provide security guarantees when a user's device is compromised. This was made possible by the introduction of the Double Ratchet Algorithm~\cite{double_ratchet}, which was designed for a setting where communication takes place between two parties. However, in practice, many apps offer the possibility of creating groups. The protocols they use to secure communication are inefficient for large group which has the undesireable consequence that the aforementioned apps have established limits on the group size of roughly 1000 users. This has motivated the introduction of the Messaging Layer Security (MLS) standard~\cite{rfc9420} by the IETF which is based on a primitive called Continuous Group Key Agreement (CGKA)~\cite{C:ACDT20}. This primitive allows a group of users to maintain a shared secret key that is frequently rotated by the group members in order to change group membership, achieve forward secrecy (FS) and post compromise security (PCS). Most protocols are based on binary trees where the nodes are associated to a pair formed by public key and a secret key. Each leaf corresponds to one of the group members and a user knows the secret keys associated to nodes along the path from their leaf to the root. When a user wants to update their key material they have to change $ \log(N) $ many keys. This requires uploading $ \log(N) $ many ciphertexts to communicate the new keys to the rest of the group members in a way that respects the tree structure. In this thesis we study how much communication between group members is required in order to add and remove users from a group as well as in order to provide PCS when we consider CGKAs built using standard cryptographic primitives like pseudo-random functions and public-key encryption. Furthermore, we also consider the case of MLS and provide the first lower bound showing that its communication complexity is much worse than previously believed, i.e., it is very far from $ \log(N) $. Finally, we also propose a variant of MLS which provably achieves the same security properties with a much lower communication cost.}, author = {Cueto Noval, Miguel}, isbn = { 978-3-99078-087-9}, issn = {2663-337X}, pages = {187}, publisher = {Institute of Science and Technology Austria}, title = {{Towards efficient secure group messaging}}, doi = {10.15479/AT-ISTA-22664}, year = {2026}, } @article{22771, abstract = {Elucidating reaction mechanisms requires efficient generation of transition states (TSs) and products. Existing diffusion and sequence-based models accelerate parts of this process over traditional string-based methods, but typically still require manual enumeration of either TSs or products, and stochastic diffusion dynamics can be inefficient and hard to control. We introduce MolGEN, a conditional flow-matching framework that uses deterministic optimal transport to map Gaussian priors to chemical distributions. For TS generation, MolGEN improves TS geometry and barrier-height prediction over diffusion models while enabling sub-second sampling. For reaction product generation, it achieves competitive top-k accuracy while preserving mass and electron balance. Using the same backbone for TS and product sampling, MolGEN enables template-free generative exploration of reaction networks without the repeated quantum-chemistry searches required by prior methods. For the γ-ketohydroperoxide decomposition network, it produces more valid TSs than string-based methods using only 12 quantum-chemistry evaluations instead of 1156, and identifies a lower-barrier pathway.}, author = {Tuo, Ping and Chen, Jiale and Li, Ju}, issn = {2041-1723}, journal = {Nature Communications}, publisher = {Springer Nature}, title = {{Flow matching for reaction pathway generation}}, doi = {10.1038/s41467-026-75654-w}, volume = {17}, year = {2026}, } @misc{22852, abstract = {This Research Data contains supplemental videos for Chapter 4 "Designing bistable nanostructures for target behavior" of my PhD Thesis "Biological functionality without biochemistry: designing nanomachines for target behavior". Supplemental video 1: Video showing the transition pathway of a bistable nanostructure with sphere-based arms, corresponding to the Machine in Scenario 4. Supplemental video 2: Video showing the transition pathway of the Source in Scenario 1. The arm tips change sides during the transition, demonstrating that the arms pass through each other. Supplemental video 3: Video showing the transition pathway of a fully polyhedral hinge structure with unconstrained arms. Note that we only show the ends of the arms. Supplemental video 4: Video showing the transition pathway of the coupled energy-delivery reaction of a Machine (gray) and a Source (blue) nanostructure for the optimized parameters in Scenario 3. Note that we only show the ends of the arms.}, author = {Ehrmann, Andreas}, keywords = {functional nanostructures, bistability, target behavior, transition pathway}, publisher = {Institute of Science and Technology Austria}, title = {{Supplemental videos for Designing bistable nanostructures for target behavior}}, doi = {10.15479/AT-ISTA-22852}, year = {2026}, } @article{22772, abstract = {The cytoplasmic antiviral sensor MDA5 is activated by double-stranded RNAs. Endogenous double-stranded RNAs are modified by the A-to-I RNA-editing ADAR family to prevent activation of MDA5. In vivo, cytoplasmic ADAR1p150 is critically required to suppress MDA5 activation, yet the editing signature of all ADAR isoforms is strongly overlapping in mice. Further, it is not clear how A-to-I modifications in dsRNA prevent MDA5 activation. Here we show that 3′ UTRs harboring inverted repeats activate MDA5 in vitro and in cells. In vitro editing by either ADAR isoform leads to editing at overlapping hotspot regions and prevents MDA5 activation in vitro and in cells. Remarkably, only inosines introduced by RNA editing are capable of suppressing MDA5 activation, while replacing guanosines with inosines during in vitro transcription has no impact on MDA5 activation. A comparison of inosines introduced by ADAR1p150 in vitro, in cells, and in vivo suggests that a small number of A-to-I conversions may be critically required to suppress MDA5 activation. As those critical editing events are predominantly altering A:U basepairs into I:U wobble basepairs, we suggest that the helical distortion introduced by those wobble pairs may prevent MDA5 polymerization and thus downstream activation of the type I interferon response.}, author = {Varada, Rajagopal and Leuchtenberger, Alina F. and Vesely, Cornelia and Kaczmarek, Beata M and Mansouri Khosravi, Hamid and Mandl, Therese C. and Milanovic, Katarina and Honarmand Tamizkar, Kasra and Rajendra, Vinod and Senoner, Hannes and Steinbichl, Linda and Borojevic, Marija and Sombke, Andy and Schmidt, Katy and Eckhard, Margret and Hofacker, Ivo L. and Walkley, Carl and Heraud-Farlow, Jacki E. and Picardi, Ernesto and Bernecky, Carrie A and Jantsch, Michael F.}, issn = {1362-4962}, journal = {Nucleic Acids Research}, number = {16}, publisher = {Oxford University Press}, title = {{Distinguishing self from non-self RNA by editing-specific inosine patterns}}, doi = {10.1093/nar/gkag845}, volume = {54}, year = {2026}, } @article{22818, abstract = {Physical processes at play in the genesis and evolution of tropical cyclones are conducive to the formation of warm cores at their centres. When the warm anomaly is particularly large in the upper troposphere, it is referred to as a high-level warm core. Previous works documented the generation of high-level warm cores as a consequence of stratospheric air intrusion into the troposphere induced by the upper-level dynamics of a tropical cyclone. However, little attention has been given to their effects on the storm's subsequent evolution. It has been suggested that the presence of a high-level warm core can have opposite effects on tropical cyclone intensity: both strengthening and weakening have been described as possible consequences of its formation. In this study, we examine the role of high-level warm cores in the intensification and dissipation processes of tropical cyclones, as reproduced in numerical models of different complexities, namely the model “System for Atmospheric Modeling” (SAM) run under idealized conditions and the model Nonhydrostatic ICosahedral Atmospheric Model run under realistic conditions following the DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains summer protocol. Our results confirm the hypothesis of a stratospheric origin behind the formation of high-level warm cores. Their initial role is shown to be an enhancement of storm intensity, by a lowering of the hydrostatic sea-level pressure (HSLP) associated with the presence of warm air aloft. However, as the warm anomaly intensifies and extends to lower levels, it also increases static stability in the air column, with the consequence of inhibiting convection and ultimately contributing to cyclone dissipation. These findings suggest that high-level warm cores play a dual role in the tropical cyclone life cycle, providing a stabilizing mechanism that can limit cyclone strength and longevity.}, author = {Davin, Andrea and Charinti, Giousef Alexandros and Muller, Caroline J and Polesello, Andrea and Pasquero, Claudia}, issn = {1477-870X}, journal = {Quarterly Journal of the Royal Meteorological Society}, publisher = {Wiley}, title = {{Stratospheric influence on tropical cyclone evolution}}, doi = {10.1002/qj.70280}, year = {2026}, } @phdthesis{21863, abstract = {Atoms and photons, two things so different but yet so alike. The former, the building block of matter, something we learn about in school and imagine it as some tiny marbles encircled by other tinier marbles. The latter, an electromagnetic wave, a light particle or an excitation of the electromagnetic field. Quantum mechanics tells us about the properties of these two entities. And even if it sounds, looks and writes counter-intuitive, it has proven right for over a century now. In this work, I elaborate on how we tested the laws of quantum mechanics and how we used them learn more about the tiny building blocks of nature and the fields they use to talk to each other. The atoms we use, are artificial. Superconducting qubits, small electrical circuits with quantized energy levels behave like electrons that transition between different orbitals in an atom. One of the qubits' advantages, is also a big disadvantage. We design the circuits' energy levels and fabricate them in a cleanroom. This allows for arbitrary spaced energy levels but in contrast to real atoms, prevents two superconducting qubits from being alike. Still, this qubit platform is one of the frontrunners for future quantum computing technology and testing fundamental physics due to their scalability. We interface superconducting qubits, which operate in the GHz regime, with microwave photons. We use 3D aluminum cavities as mediators between qubits and photons. The cavities allow for non-destructive readout of the qubit state, they shield the qubits from noise at the qubit frequency and they give us an easy way to frequency-tune these joint systems. We need to operate superconducting qubits and their cavities at millikelvin temperatures in dilution refrigerators. At higher temperatures, superconductivity suffers and even worse, the environment is filled with thermal noise photons. This poses a fundamental limitation on the scalability of superconducting qubit devices. Also connecting multiple devices in different fridges does not work over room temperature links because the microwave photons used for this purpose will be covered in noise and the quantum information they carry, will be unusable. Infrared photons do not suffer from this noise problem since there are close to zero thermal noise photons at their frequencies at room temperature. We cannot simply interface superconducting devices with optical photons due their frequency mismatch and the destructive effect of optical photons on superconductors. Therefore, we use microwave-to-optics transducers that allow to convert microwave photons into optical ones and vice-versa. The transducers that we use are macroscopic electro-optic transducers using the Pockels effect in a disk-shaped Lithium Niobate whispering gallery mode resonator. By using a strong optical pump, photons from the two frequency domains experience a beam-splitter interaction and get converted from one to the other. We measure the generated optical photons using elaborate optical setups, optical heterodyning and single photon detectors to gain knowledge about the qubit state or the converted microwave photons. Bridging the microwave and the optical world allows us to take advantage of both of their strengths but it also requires deep knowledge about both of their working principles. In this work, we describe two experiments that our group conducted to showcase the opportunities that arise from interfacing superconducting qubits with optical photons but also the pitfalls, one may encounter on the way. In the first experiment, we managed to all-optically read out a superconducting qubit. We show that the assignment fidelity, the probability that a measurement of the qubit state matches the prepared state, is close to equal for all-optical, microwave-to-optics and conventional microwave readout. We show T1 and T2 measurements for all three readout types and give an analysis of the noise caused by the optics. Finally, we show that the infrared light does not affect the qubit performance in a negative way but that the heating it causes does. This is an important insight that we used in the next experiment. The second experiment is the upconversion of itinerant single microwave photons to the optical domain. We show that we can generate single microwave photons from a qubit-cavity system. We upconvert these single photons, measure them with a single photon detector and reconstruct their shape. By conducting a single photon Rabi measurement, we show correlations between the microwave and the optical domain. And by thorough signal-to-noise measurements and noise analysis, we find that we can generate single infrared photons with high signal-to-noise ratio 5.1 and low transducer added noise (<0.012 quanta). We show that this measurement creates a path towards entanglement of a superconducting qubit and an optical photon and what parameters need to be improved to achieve it. Additionally, this experiment is a proof of principle for an on-demand infrared single photon source. More generally, it allows to link microwave quantum technology in general to the optical domain.}, author = {Werner, Thomas}, issn = {2663-337X}, keywords = {Superconducting qubits, Quantum optics, Single photons and quantum effects, Nonlinear optics}, pages = {97}, publisher = {Institute of Science and Technology Austria}, title = {{Interfacing superconducting qubits with optical photons}}, doi = {10.15479/AT-ISTA-21863}, year = {2026}, } @unpublished{21870, abstract = {Superconducting qubits are a leading candidate for utility-scale quantum computing due to their fast gate speeds and steadily decreasing error rates. The requirement for millikelvin operating temperatures, however, creates a significant scaling bottleneck. Modular architectures using optical fiber links could bridge separate cryogenic nodes, but superconducting circuits do not have coherent optical transitions and microwave-to-optical conversion has not been shown for any non-classical photon state. In this work, we demonstrate the on-demand generation and tomographic reconstruction of itinerant single microwave photons at 8.9 GHz from a superconducting qubit. We upconvert this non-Gaussian state with a transducer added noise below 0.012 quanta and count the converted telecom photons at 193.4 THz with a signal-to-noise ratio of up to 5.1$\pm$1.1. We characterize the trade-offs between throughput and noise, and establish a viable path toward heralded entanglement distribution and gate teleportation. Looking ahead, these results empower existing superconducting devices to take a key role in distributed quantum technologies and heterogeneous quantum systems.}, author = {Werner, Thomas and Riyazi, Erfan and Hawaldar, Samarth and Sahu, Rishabh and Arnold, Georg M and Paul Falthansl-Scheinecker, Paul Falthansl-Scheinecker and Naranjo, Jennifer A. Sánchez and Loi, Dante and Kapoor, Lucky N. and Zemlicka, Martin and Qiu, Liu and Militaru, Andrei and Fink, Johannes M}, booktitle = {arXiv}, title = {{Electro-optic conversion of itinerant Fock states}}, doi = {10.48550/arXiv.2602.00928}, year = {2026}, } @phdthesis{22745, author = {Michalik, David}, issn = {2663-337X}, pages = {182}, publisher = {Institute of Science and Technology Austria}, title = {{Mechanistic insights into MDA5 selectivity and regulation}}, doi = {10.15479/AT-ISTA-22745}, year = {2026}, } @article{22912, abstract = {Deep convection from tropical cyclones (TCs) can reach the height of the tropopause and as such an interaction between the upper troposphere and lower stratosphere is likely to occur. Such interactions have been reported in both numerical and observational studies, mainly showing that subsidence from the stratosphere into the eye of an intensifying storm leads to a high-level warm core. However, the effect of this upper-level warming on the intensity of the TCs is not yet well understood. In this study, we show that subsiding air from the stratosphere beyond subsidence in the eye is the reason for the upper-level warming in idealized simulations. We further show that it is possible to quantify the effects of the upper-level warming on the potential intensity of the TC. Finally, we conclude that overshooting convection into the stratosphere causes the observed subsidence, as both become more pronounced with increasing sea-surface temperatures (SST).}, author = {Charinti, Giousef Alexandros and Davin, Andrea and Polesello, Andrea and Muller, Caroline J and Pasquero, Claudia}, issn = {1944-8007}, journal = {Geophysical Research Letters}, number = {17}, publisher = {Wiley}, title = {{Impact of upper-level warming on tropical cyclone intensity}}, doi = {10.1029/2025GL121307}, volume = {53}, year = {2026}, } @phdthesis{21198, abstract = {In recent years there has been a massive increase in the amount of data generated in a decentralized manner. Ever more powerful edge devices, such as smartphones, have become ubiquitous in most societies on earth. Through text typed, photos taken and apps used, these devices, which we refer to as clients, generate enormous amounts of high quality and complex data. Moreover, the nature of these devices means the data they generate is often sensitive and privacy concerns prevent it being gathered and stored in a central location. This presents a challenge to the modern machine learning paradigm that requires central access to large amounts of data. Federated learning (FL) has emerged as one of the answers to this problem. Rather than bringing the data to the model, FL sends the model to the data. Model training takes place on device, with periodically synchronized updates, allowing data to remain locally stored. While this approach offers significant privacy advantages it comes with its own set of unique challenges. These include: data heterogeneity, the notion that different devices generate data in distinct ways which can negatively impact training dynamics; systems heterogeneity, meaning that different devices may have differing hardware specifications; high communication costs, which are induced by the repeated transferring of models over the network and low device computational power, which limits the use of larger models on device. In this thesis we present a range of methods for federated learning. We focus primarily on the challenge of data heterogeneity, though the methods presented are designed to be well adapted to the other challenges of a federated setting, such as the constraints of limited compute and communication overhead. We first present a method for explicitly modeling client data heterogeneity. The approach formulates clients as samples from a certain probability distribution and infers the parameters of this distribution from the available training clients. This learned distribution then represents the heterogeneity present among the clients and can be sampled from in order to create new simulated clients that are similar to the real clients we have observed so far. Following this we present two methods for directly dealing with data heterogeneity through personalization. Highly heterogeneous client data distributions can mean that learning a single global model becomes suboptimal, and some form of personalization of models to each individual client is required. Our approaches are based around hypernetworks, which we use to generate personalized model parameters without the need for additional training or finetuning. In the first approach we focus on generating full parameterizations of client models using learned embeddings of client data and labels, with a hypernetwork located on the central server. In the second approach we address the more challenging scenario where we want to generate a personalized model for a client without any label information. The hypernetwork is trained to generate a low dimensional representation of a client’s personalized model parameters, allowing it to be transferred to and run on the client devices. In our final presented method, we change our focus and rather than aim to directly address the challenge of data heterogeneity, we instead ensure we are unaffected by it. This is done in the context of k-means clustering and we present a method for federated clustering with a focus on added privacy guarantees.}, author = {Scott, Jonathan A}, issn = {2663-337X}, pages = {158}, publisher = {Institute of Science and Technology Austria}, title = {{Data heterogeneity and personalization in federated learning}}, doi = {10.15479/AT-ISTA-21198}, year = {2026}, } @inproceedings{21717, abstract = {Robust Markov Decision Processes (RMDPs) generalize classical MDPs that consider uncertainties in transition probabilities by defining a set of possible transition functions. An objective is a set of runs (or infinite trajectories) of the RMDP, and the value for an objective is the maximal probability that the agent can guarantee against the adversarial environment. We consider (a) reachability objectives, where given a target set of states, the goal is to eventually arrive at one of them; and (b) parity objectives, which are a canonical representation for ω-regular objectives. The qualitative analysis problem asks whether the objective can be ensured with probability 1. In this work, we study the qualitative problem for reachability and parity objectives on RMDPs without making any assumption over the structures of the RMDPs, e.g., unichain or aperiodic. Our contributions are twofold. We first present efficient algorithms with oracle access to uncertainty sets that solve qualitative problems of reachability and parity objectives. We then report experimental results demonstrating the effectiveness of our oracle-based approach on classical RMDP examples from the literature scaling up to thousands of states.}, author = {Asadi, Ali and Chatterjee, Krishnendu and Kafshdar Goharshadi, Ehsan and Karrabi, Mehrdad and Shafiee, Ali}, booktitle = {Proceedings of the 40th AAAI Conference on Artificial Intelligence}, issn = {2374-3468}, location = {Singapore, Singapore}, number = {43}, pages = {36137--36145}, publisher = {Association for the Advancement of Artificial Intelligence}, title = {{Qualitative analysis of ω-regular objectives on robust MDPs}}, doi = {10.1609/aaai.v40i43.40931}, volume = {40}, year = {2026}, } @article{21746, abstract = {As vertebrates transitioned from water to land, locomotion shifted from undulatory swimming to limb-based movement. How spinal circuits and their cell types evolved to support this transition remains unclear. We leverage frog metamorphosis, which recapitulates this transition within a single organism, to define how spinal circuits generate aquatic versus terrestrial motor patterns. At swim stages, spinal architecture is uniform, with a transcriptionally and anatomically homogeneous motor and interneurons. As limbs develop and their movement complexifies, spinal circuits expand in neuron number and subtype diversity. This expansion is most pronounced for V1 inhibitory neurons, which increase ∼70-fold and diversify into transcriptionally distinct subtypes. Disrupting transcription factors defining emerging motor and V1 populations reveals molecular segregation between swim and limb circuits, highlighting the role of subtype diversity in motor coordination. A multifold increase in inhibitory neuron diversity thus underlies the tail-to-limb locomotor transition, providing a framework for spinal circuit adaptation during vertebrate evolution.}, author = {Vijatovic, David and Toma, Florina Alexandra and Ignatyev, Y and Harrington, Zoe P and Sommer, Christoph M and Hauschild, Robert and Smits, Matthijs Geert and Dalla Vecchia, Marco and Trevisan, Alexandra J. and Chapman, Phillip and Julseth, Mara and Brenner-Morton, Susan and Gabitto, Mariano I. and Dasen, Jeremy S. and Bikoff, Jay B. and Sweeney, Lora Beatrice Jaeger}, issn = {2211-1247}, journal = {Cell Reports}, number = {4}, publisher = {Elsevier}, title = {{Multifold increase in spinal inhibitory cell types with emergence of limb movement}}, doi = {10.1016/j.celrep.2026.117227}, volume = {45}, year = {2026}, } @inproceedings{22146, abstract = {We study differentially private model training with stochastic gradient descent under learning rate scheduling and correlated noise. Although correlated noise, in particular via matrix factorizations, has been shown to improve accuracy, prior theoretical work focused primarily on the prefix-sum workload. That workload assumes a constant learning rate, whereas in practice learning rate schedules are widely used to accelerate training and improve convergence. We close this gap by deriving general upper and lower bounds for a broad class of learning rate schedules in both single- and multi-epoch settings. Building on these results, we propose a learning-rate-aware factorization that achieves improvements over prefix-sum factorizations under both MaxSE and MeanSE error metrics. Our theoretical analysis yields memory-efficient constructions suitable for practical deployment, and experiments on CIFAR-10 and IMDB datasets confirm that schedule-aware factorizations improve accuracy in private training.}, author = {Kalinin, Nikita and Andersson, Joel D}, booktitle = {7th Symposium on Foundations of Responsible Computing}, isbn = {9783959774192}, issn = {1868-8969}, keywords = {differential privacy, machine learning, matrix factorization}, location = {Cambridge, MA; United States}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik}, title = {{Learning rate scheduling with matrix factorization for private training}}, doi = {10.4230/LIPIcs.FORC.2026.2}, volume = {368}, year = {2026}, } @phdthesis{22667, author = {Vijatovic, David}, isbn = {978-3-99078-082-4}, issn = {2663-337X}, pages = {172}, publisher = {Institute of Science and Technology Austria}, title = {{Dissecting the molecular and functional basis of motor control in the frog Xenopus laevis}}, doi = {10.15479/AT-ISTA-22667}, year = {2026}, }