Repository for Publications and Research Data
News from the ETH Library
Recently Added
Dynamic Probit Models With Network Interdependence and Unobserved Heterogeneity With an Application to Multinational-Firm Participation in China
Item type: Journal Article
Egger P.H.; Kesina M. (2026)
In many areas of regional/spatial economics, we are faced with analyzing discrete dynamic choice problems of economic agents interacting with some network structure through general equilibrium, input-output links, or strategic interaction. Stochastic estimating equations for such problems can be difficult for many practitioners to estimate. This paper proposes MCMC estimation approaches based on control functions for estimating dynamic panel probit models where a large number of cross-sectional units is observed over a short period of time, and cross-sectional units feature network interdependencies. The proposed approaches enable accounting for dynamic adjustment and different types of cross-sectional dependence. These features should make the approaches interesting for applications in many empirical contexts. The paper outlines the estimation approaches, illustrates their suitability by simulation examples, and provides an application to study dynamic multinational-firm participation among regionally connected Chinese firms.
Optimizing against Infeasible Inclusions from Data for Semantic Segmentation through Morphology
Item type: Conference Paper
Basu S.; Van Gool L.; Sakaridis C. (2026)
State-of-the-art semantic segmentation models are typically optimized in a data-driven fashion, minimizing solely per-pixel or per-segment classification objectives on their training data. This purely data-driven paradigm often leads to absurd segmentations, especially when the domain of input images is shifted from the one encountered during training. For instance, state-of-the-art models may assign the label "road"to a segment that is included by another segment that is respectively labeled as "sky". However, the ground truth of the existing dataset at hand dictates that such inclusion is not feasible. Our method, Infeasible Semantic Inclusions (InSeIn), first extracts explicit inclusion constraints that govern spatial class relations from the semantic segmentation training set at hand in an offline, data-driven fashion, and then enforces a morphological yet differentiable loss that penalizes violations of these constraints during training to promote prediction feasibility. InSeIn is a light-weight plug-and-play method, constitutes a novel step towards minimizing infeasible semantic inclusions in the predictions of learned segmentation models, and yields consistent and significant performance improvements over diverse state-of-the-art networks across the ADE20K, Cityscapes, and ACDC datasets. Codebase will be made available. Code is available at https://github.com/SHAMIK-97/InSeIn
BitLogic: A Framework for Gradient-Based LUT-Native Neural Networks
Item type: Journal Article
Bührer S.; Plesner A.; Aczel T.; et al. (2026)
Gradient-based LUT-and logic-gate-based neural networks (LUTNet, LogicNets, Dif-fLogic, PolyLUT, NeuraLUT, WARP-LUT, DWN, LILogicNet, LightLUT) replace multiply-accumulate arithmetic with Boolean lookups. The same trained checkpoint deploys to GPU as bitwise ops on bit-packed activations, to FPGA as LUT primitives, and to ASIC as standard-cell gates, all from one code path. Yet each method ships its own training pipeline, encoder, connectivity rule, fan-in, and hardware-reporting convention. The natural practitioner question, which of these choices actually matter for accuracy and which for hardware cost, therefore has no answer in the current literature. We release BitLogic, a unified framework that factors the field into a five-axis design space (encoder, connectivity, fan-in, node parameterization, head) and instantiates every prior method under one shared training and evaluation protocol. The framework deliberately omits method-specific procedures such as calibration, pruning, and thresholding, and all evaluations are limited to two-layer feed-forward networks. Combining the per-axis winners identifies a new best-of-space configuration that outperforms every retrained prior on every (dataset, width) cell in which every compared prior fits the shared budget, across MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100. We evaluate the best-of-space model on all three backends. On MNIST the resulting two-layer network reaches ∼126 MSamples/s on FPGA, ∼15× the throughput of a bit-packed GPU forward path that itself processes 64 samples per 64-bit operation, at four-to-five orders of magnitude less energy.
Item type: Person
Development of All-solid-state All-oxide thin film Li-ion batteries
Item type: Doctoral Thesis
Montazerian, Mohammadhossein (2026)
The rapid miniaturization of electronic devices has created an urgent demand for safe, high-power, and long-lifetime micro-energy storage systems. Thin film lithium-ion batteries represent a promising solution; however, their performance is fundamentally constrained by interfacial instabilities, limited ionic transport, and structural degradation of electrode and electrolyte materials under high charge/discharge rates. This thesis addresses these challenges through the design, fabrication, and electrochemical investigation of oxide-based thin film materials for advanced solid-state thin film lithium-ion batteries.
In chapter three of this work, lithium germanium phosphate (Li4-xGe1-xPxO4, LGPO) thin films are investigated as potential solid-state electrolytes for microbattery applications. LGPO films are deposited by pulsed laser deposition, and their ionic transport properties are systematically studied using in situ electrochemical impedance spectroscopy. The films exhibit room-temperature lithium-ion conductivities on the order of 10-5 S.cm-1, demonstrating their suitability as solid electrolytes for thin film battery architectures. These results provide insight into the structure–transport relationships in LGPO and identify key parameters governing ionic conduction in oxide electrolyte thin films.
In addition, in chapter 4 of this thesis, oxide-based electrode materials are engineered at the thin film level to achieve enhanced electrochemical performance. The role of epitaxial growth, interface quality, and microstructural control is examined in detail, revealing the strong influence of structural coherence and interfacial properties on rate capability and cycling stability.
Finally, in chapter five, fully solid-state thin film battery architectures are explored by integrating oxide electrodes with lithium phosphorus oxynitride (LiPON). The compatibility and electrochemical behavior of different material combinations are evaluated, demonstrating the feasibility of oxide all-solid-state configurations and highlighting key factors governing internal resistance and long-term performance.
Taken together, the results presented in this thesis establish interface engineering as a powerful strategy to stabilize oxide thin film electrodes and electrolytes, enabling safe, high-rate, and long-lifetime thin film lithium-ion batteries. This work lays the foundation for future all oxide, all-solid-state microbattery architectures suitable for next-generation microelectronic and energy-autonomous systems.