Search

Quantitative mass spectrometry has revolutionized proteomics by enabling simultaneous quantification of thousands of proteins. Pooling patient-derived data from multiple institutions enhances statistical power but raises significant privacy concerns. Here we introduce FedProt, the first privacy-preserving tool for collaborative differential protein abundance analysis of distributed data, which utilizes federated learning and additive secret sharing. In the absence of a multicenter patient-derived dataset for evaluation, we created two, one at five centers from LFQ E.coli experiments and one at three centers from TMT human serum. Evaluations using these datasets confirm that FedProt achieves accuracy equivalent to DEqMS applied to pooled data, with completely negligible absolute differences no greater than $\text{$4 \times 10^{-12}$}$. In contrast, -log10(p-values) computed by the most accurate meta-analysis methods diverged from the centralized analysis results by up to 25-27. FedProt is available as a web tool with detailed documentation as a FeatureCloud App., Comment: 52 pages, 16 figures, 12 tables. Last two authors listed are joint last authors

In this paper, we present a data-driven model for estimating optimal rework policies in manufacturing systems. We consider a single production stage within a multistage, lot-based system that allows for optional rework steps. While the rework decision depends on an intermediate state of the lot and system, the final product inspection, and thus the assessment of the actual yield, is delayed until production is complete. Repair steps are applied uniformly to the lot, potentially improving some of the individual items while degrading others. The challenge is thus to balance potential yield improvement with the rework costs incurred. Given the inherently causal nature of this decision problem, we propose a causal model to estimate yield improvement. We apply methods from causal machine learning, in particular double/debiased machine learning (DML) techniques, to estimate conditional treatment effects from data and derive policies for rework decisions. We validate our decision model using real-world data from opto-electronic semiconductor manufacturing, achieving a yield improvement of 2 - 3% during the color-conversion process of white light-emitting diodes (LEDs)., Comment: 30 pages, 10 figures

In autonomous driving, the most challenging scenarios can only be detected within their temporal context. Most video anomaly detection approaches focus either on surveillance or traffic accidents, which are only a subfield of autonomous driving. We present HF$^2$-VAD$_{AD}$, a variation of the HF$^2$-VAD surveillance video anomaly detection method for autonomous driving. We learn a representation of normality from a vehicle's ego perspective and evaluate pixel-wise anomaly detections in rare and critical scenarios., Comment: Daniel Bogdoll and Jan Imhof contributed equally. Accepted for publication at BMVC 2024 RROW workshop

The scale-up of autonomous vehicles depends heavily on their ability to deal with anomalies, such as rare objects on the road. In order to handle such situations, it is necessary to detect anomalies in the first place. Anomaly detection for autonomous driving has made great progress in the past years but suffers from poorly designed benchmarks with a strong focus on camera data. In this work, we propose AnoVox, the largest benchmark for ANOmaly detection in autonomous driving to date. AnoVox incorporates large-scale multimodal sensor data and spatial VOXel ground truth, allowing for the comparison of methods independent of their used sensor. We propose a formal definition of normality and provide a compliant training dataset. AnoVox is the first benchmark to contain both content and temporal anomalies., Comment: Daniel Bogdoll, Iramm Hamdard, and Lukas Namgyu R\"o{\ss}ler contributed equally. Accepted for publication at ECCV 2024 W-CODA workshop

We introduce Adapters, an open-source library that unifies parameter-efficient and modular transfer learning in large language models. By integrating 10 diverse adapter methods into a unified interface, Adapters offers ease of use and flexible configuration. Our library allows researchers and practitioners to leverage adapter modularity through composition blocks, enabling the design of complex adapter setups. We demonstrate the library's efficacy by evaluating its performance against full fine-tuning on various NLP tasks. Adapters provides a powerful tool for addressing the challenges of conventional fine-tuning paradigms and promoting more efficient and modular transfer learning. The library is available via https://adapterhub.ml/adapters., Comment: EMNLP 2023: Systems Demonstrations

Quantized adiabatic transport can occur when a system is slowly modulated over time. In most realizations however, the efficiency of such transport is reduced by unwanted dissipation, back-scattering, and non-adiabatic effects. In this work, we realize a topological adiabatic pump in an electrical circuit network that supports remarkably stable and long-lasting pumping of a voltage signal. We further characterize the topology of our system by deducing the Chern number from the measured edge band structure. To achieve this, the experimental setup makes use of active circuit elements that act as time-variable voltage-controlled inductors., Comment: main (5 pages, 3 figures) plus supplement (8 pages, 4 figures)

In manufacturing, rework refers to an optional step of a production process which aims to eliminate errors or remedy products that do not meet the desired quality standards. Reworking a production lot involves repeating a previous production stage with adjustments to ensure that the final product meets the required specifications. While offering the chance to improve the yield and thus increase the revenue of a production lot, a rework step also incurs additional costs. Additionally, the rework of parts that already meet the target specifications may damage them and decrease the yield. In this paper, we apply double/debiased machine learning (DML) to estimate the conditional treatment effect of a rework step during the color conversion process in opto-electronic semiconductor manufacturing on the final product yield. We utilize the implementation DoubleML to develop policies for the rework of components and estimate their value empirically. From our causal machine learning analysis we derive implications for the coating of monochromatic LEDs with conversion layers., Comment: 22 pages, 15 figures

On-resonance Rydberg atom-based radio-frequency (RF) electric field sensing methods remain limited by the narrow frequency signal detection bands available by resonant transitions. The use of an additional RF tuner field to dress or shift a target Rydberg state can be used to return a detuned signal field to resonance and thus dramatically extend the frequency range available for resonant sensing. Here we investigate three distinct tuning level schemes based on adjacent Rydberg transitions, which are shown to have distinct characteristics and can be controlled with mechanisms based on the tuning field frequency or field strength. We further show that a two-photon Raman feature can be used as an effective tuning mechanism separate from conventional Autler-Townes splitting. We compare our tuning schemes to AC Stark effect-based broadband RF field sensing and show that although the sensitivity is diminished as we tune away from a resonant state, it nevertheless can be used in configurations where there is a low density of Rydberg states, which would result in a weak AC Stark effect.

United Nations Educational, Scientific and Cultural Organization (UNESCO), the United Nations Children's Fund (UNICEF), the World Bank and the Organisation for Economic Co-operation and Development (OECD) have collaborated in the third round of the Survey on National Education Responses to COVID-19 School Closures, administered by the UNESCO Institute for Statistics (UIS) and OECD to Ministry of Education officials. The questions covered four levels of education: preprimary, primary, lower secondary and upper secondary. While the first two rounds of the survey were implemented during the periods May-June and July-October 2020, respectively, the third round was implemented during the period February-June 2021. In total, 143 countries responded to the questionnaire. Thirty-one countries submitted responses to the OECD ("OECD survey") and 112 countries responded to the UIS ("UIS survey"). Seven countries responded to both surveys. In these instances, the more complete set responses were used in analysis.

Procrastination, the irrational delay of tasks, is a common occurrence in online learning. Potential negative consequences include higher risk of drop-outs, increased stress, and reduced mood. Due to the rise of learning management systems and learning analytics, indicators of such behavior can be detected, enabling predictions of future procrastination and other dilatory behavior. However, research focusing on such predictions is scarce. Moreover, studies involving different types of predictors and comparisons between the predictive performance of various methods are virtually non-existent. In this study, we aim to fill these research gaps by analyzing the performance of multiple machine learning algorithms when predicting the delayed or timely submission of online assignments in a higher education setting with two categories of predictors: subjective, questionnaire-based variables and objective, log-data based indicators extracted from a learning management system. The results show that models with objective predictors consistently outperform models with subjective predictors, and a combination of both variable types perform slightly better. For each of these three options, a different approach prevailed (Gradient Boosting Machines for the subjective, Bayesian multilevel models for the objective, and Random Forest for the combined predictors). We conclude that careful attention should be paid to the selection of predictors and algorithms before implementing such models in learning management systems., Comment: 20 pages, 6 tables, 6 photos

We observe a localized cnoidal (LCn) state in an electric circuit network. Its formation derives from the interplay of non-linearity and the topology inherent to a Su-Schrieffer-Heeger (SSH) chain of inductors. Varicap diodes act as voltage-dependent capacitors, and create a non-linear on-site potential. For a sinusoidal voltage excitation around midgap frequency, we show that the voltage response in the non-linear SSH circuit follows the Korteweg-de Vries equation. The topological SSH boundary state which relates to a midgap impedance peak in the linearized limit is distorted into the LCn state in the non-linear regime, where the cnoidal eccentricity decreases from edge to bulk.

Curved spaces play a fundamental role in many areas of modern physics, from cosmological length scales to subatomic structures related to quantum information and quantum gravity. In tabletop experiments, negatively curved spaces can be simulated with hyperbolic lattices. Here we introduce and experimentally realize hyperbolic matter as a paradigm for topological states through topolectrical circuit networks relying on a complex-phase circuit element. The experiment is based on hyperbolic band theory that we confirm here in an unprecedented numerical survey of finite hyperbolic lattices. We implement hyperbolic graphene as an example of topologically nontrivial hyperbolic matter. Our work sets the stage to realize more complex forms of hyperbolic matter to challenge our established theories of physics in curved space, while the tunable complex-phase element developed here can be a key ingredient for future experimental simulation of various Hamiltonians with topological ground states., Comment: 6 pages + methods + supplementary information

Recent advances in operating and manipulating DNA have provided unique experimental possibilities in many fields of DNA research, especially in gene therapy. Researchers have deployed many techniques, experimental and theoretical, to study the DNA structure changes due to external perturbation. It is crucial to understand the structural and dynamical changes in the DNA molecules in a confined state to understand and control the self-assembly of DNA confined in a chamber or nano-channel for various applications. In the current manuscript, we extend the work study the effect of confinement on the thermal stability and the structural properties of duplex DNA. The present work is an extension of our previous research works. For our study we have considered a 1 BNA chain that is confined in a cylindrical geometry. How the geometry of the confinement affects the opening and other structural parameters of DNA molecule is the objective of this manuscript. We have used a statistical model(PBD model) and Molecular dynamics simulations for our purpose.

This paper presents an analytic model of one dimensional magnetostriction. We show how specific assumptions regarding the symmetry of key micromagnetic energies (magnetocrystalline, magnetoelastic, and Zeeman) reduce a general three-dimensional statistical mechanics model to a one-dimensional form with an exact solution. We additionally provide a useful form of the analytic equations to help ensure numerical accuracy. Numerical results show that the model maintains accuracy over a large range of applied magnetic fields and stress conditions extending well outside those produced in standard laboratory conditions. A comparison to experimental data is performed for several magnetostrictive materials. The model is shown to accurately predict the behavior of Terfenol-D, while two compositions of Galfenol are modeled with varying accuracy. To conclude we discuss what conditions facilitate the description of materials with cubic crystalline anisotropy as transversely isotropic, to achieve peak model performance.

The Laplace operator encodes the behavior of physical systems at vastly different scales, describing heat flow, fluids, as well as electric, gravitational, and quantum fields. A key input for the Laplace equation is the curvature of space. Here we discuss and experimentally demonstrate that the spectral ordering of Laplacian eigenstates for hyperbolic (negatively curved) and flat two-dimensional spaces has a universally different structure. We use a lattice regularization of hyperbolic space in an electric-circuit network to measure the eigenstates of a "hyperbolic drum", and in a time-resolved experiment we verify signal propagation along the curved geodesics. Our experiments showcase both a versatile platform to emulate hyperbolic lattices in tabletop experiments, and a set of methods to verify the effective hyperbolic metric in this and other platforms. The presented techniques can be utilized to explore novel aspects of both classical and quantum dynamics in negatively curved spaces, and to realise the emerging models of topological hyperbolic matter., Comment: Main text and methods (8 pages, 4 figures) + supplementary information (13 pages, 8 figures). Supplementary data available at https://doi.org/10.3929/ethz-b-000503548. (v2) updated with accepted version

We demonstrate the use of multiple atomic-level Rydberg-atom schemes for continuous frequency detection of radio frequency (RF) fields. Resonant detection of RF fields by electromagnetically-induced transparency and Autler-Townes (AT) in Rydberg atoms is typically limited to frequencies within the narrow bandwidth of a Rydberg transition. By applying a second field resonant with an adjacent Rydberg transition, far-detuned fields can be detected through a two-photon resonance AT splitting. This two-photon AT splitting method is several orders of magnitude more sensitive than off-resonant detection using the Stark shift. We present the results of various experimental configurations and a theoretical analysis to illustrate the effectiveness of this multiple level scheme. These results show that this approach allows for the detection of frequencies in continuous band between resonances with adjacent Rydberg states.

We propose an original application of screening methods using machine learning to detect collusive groups of firms in procurement auctions. As a methodical innovation, we calculate coalition-based screens by forming coalitions of bidders in tenders to flag bid-rigging cartels. Using Swiss, Japanese and Italian procurement data, we investigate the effectiveness of our method in different countries and auction settings, in our cases first-price sealed-bid and mean-price sealed-bid auctions. We correctly classify 90\% of the collusive and competitive coalitions when applying four machine learning algorithms: lasso, support vector machine, random forest, and super learner ensemble method. Finally, we find that coalition-based screens for the variance and the uniformity of bids are in all the cases the most important predictors according the random forest.

Adding to the literature on the data-driven detection of bid-rigging cartels, we propose a novel approach based on deep learning (a subfield of artificial intelligence) that flags cartel participants based on their pairwise bidding interactions with other firms. More concisely, we combine a so-called convolutional neural network for image recognition with graphs that in a pairwise manner plot the normalized bid values of some reference firm against the normalized bids of any other firms participating in the same tenders as the reference firm. Based on Japanese and Swiss procurement data, we construct such graphs for both collusive and competitive episodes (i.e when a bid-rigging cartel is or is not active) and use a subset of graphs to train the neural network such that it learns distinguishing collusive from competitive bidding patterns. We use the remaining graphs to test the neural network's out-of-sample performance in correctly classifying collusive and competitive bidding interactions. We obtain a very decent average accuracy of around 90% or slightly higher when either applying the method within Japanese, Swiss, or mixed data (in which Swiss and Japanese graphs are pooled). When using data from one country for training to test the trained model's performance in the other country (i.e. transnationally), predictive performance decreases (likely due to institutional differences in procurement procedures across countries), but often remains satisfactorily high. All in all, the generally quite high accuracy of the convolutional neural network despite being trained in a rather small sample of a few 100 graphs points to a large potential of deep learning approaches for flagging and fighting bid-rigging cartels.

We employ electric circuit networks to study topological states of matter in non-Hermitian systems enriched by parity-time symmetry $\mathcal{PT}$ and chiral symmetry anti-$\mathcal{PT}$ ($\mathcal{APT}$). The topological structure manifests itself in the complex admittance bands which yields excellent measurability and signal to noise ratio. We analyze the impact of $\mathcal{PT}$ symmetric gain and loss on localized edge and defect states in a non-Hermitian Su--Schrieffer--Heeger (SSH) circuit. We realize all three symmetry phases of the system, including the $\mathcal{APT}$ symmetric regime that occurs at large gain and loss. We measure the admittance spectrum and eigenstates for arbitrary boundary conditions, which allows us to resolve not only topological edge states, but also a novel $\mathcal{PT}$ symmetric $\mathbb{Z}_2$ invariant of the bulk. We discover the distinct properties of topological edge states and defect states in the phase diagram. In the regime that is not $\mathcal{PT}$ symmetric, the topological defect state disappears and only reemerges when $\mathcal{APT}$ symmetry is reached, while the topological edge states always prevail and only experience a shift in eigenvalue. Our findings unveil a future route for topological defect engineering and tuning in non-Hermitian systems of arbitrary dimension., Comment: 4 pages, 3 figures, 6 pages supplement (separate file)

Context. The electrostatic potential of a spacecraft, VS, is important for the capabilities of in situ plasma measurements. Rosetta has been found to be negatively charged during most of the comet mission and even more so in denser plasmas. Aims. Our goal is to investigate how the negative VS correlates with electron density and temperature and to understand the physics of the observed correlation. Methods. We applied full mission comparative statistics of VS, electron temperature, and electron density to establish VS dependence on cold and warm plasma density and electron temperature. We also used Spacecraft-Plasma Interaction System (SPIS) simulations and an analytical vacuum model to investigate if positively biased elements covering a fraction of the solar array surface can explain the observed correlations. Results. Here, the VS was found to depend more on electron density, particularly with regard to the cold part of the electrons, and less on electron temperature than was expected for the high flux of thermal (cometary) ionospheric electrons. This behaviour was reproduced by an analytical model which is consistent with numerical simulations. Conclusions. Rosetta is negatively driven mainly by positively biased elements on the borders of the front side of the solar panels as these can efficiently collect cold plasma electrons. Biased elements distributed elsewhere on the front side of the panels are less efficient at collecting electrons apart from locally produced electrons (photoelectrons). To avoid significant charging, future spacecraft may minimise the area of exposed bias conductors or use a positive ground power system., Comment: 12 pages, 17 figures, Accepted for publication in A&A

We propose a new method for flagging bid rigging, which is particularly useful for detecting incomplete bid-rigging cartels. Our approach combines screens, i.e. statistics derived from the distribution of bids in a tender, with machine learning to predict the probability of collusion. As a methodological innovation, we calculate such screens for all possible subgroups of three or four bids within a tender and use summary statistics like the mean, median, maximum, and minimum of each screen as predictors in the machine learning algorithm. This approach tackles the issue that competitive bids in incomplete cartels distort the statistical signals produced by bid rigging. We demonstrate that our algorithm outperforms previously suggested methods in applications to incomplete cartels based on empirical data from Switzerland.

Lake ice, as part of the Essential Climate Variable (ECV) lakes, is an important indicator to monitor climate change and global warming. The spatio-temporal extent of lake ice cover, along with the timings of key phenological events such as freeze-up and break-up, provide important cues about the local and global climate. We present a lake ice monitoring system based on the automatic analysis of Sentinel-1 Synthetic Aperture Radar (SAR) data with a deep neural network. In previous studies that used optical satellite imagery for lake ice monitoring, frequent cloud cover was a main limiting factor, which we overcome thanks to the ability of microwave sensors to penetrate clouds and observe the lakes regardless of the weather and illumination conditions. We cast ice detection as a two class (frozen, non-frozen) semantic segmentation problem and solve it using a state-of-the-art deep convolutional network (CNN). We report results on two winters ( 2016 - 17 and 2017 - 18 ) and three alpine lakes in Switzerland. The proposed model reaches mean Intersection-over-Union (mIoU) scores >90% on average, and >84% even for the most difficult lake. Additionally, we perform cross-validation tests and show that our algorithm generalises well across unseen lakes and winters., Comment: Accepted for ISPRS Congress 2020, Nice, France