Search

High-precision surface defect detection in manufacturing is essential for ensuring quality control. Laser triangulation profilometric sensors are key to this process, providing detailed and accurate surface measurements over a line. To achieve a complete and precise surface scan, accurate relative motion between the sensor and the workpiece is required. It is crucial to control the sensor pose to maintain optimal distance and relative orientation to the surface. It is also important to ensure uniform profile distribution throughout the scanning process. This paper presents a novel Reinforcement Learning (RL) based approach to optimize robot inspection trajectories for profilometric sensors. Building upon the Boustrophedon scanning method, our technique dynamically adjusts the sensor position and tilt to maintain optimal orientation and distance from the surface, while also ensuring a consistent profile distance for uniform and high-quality scanning. Utilizing a simulated environment based on the CAD model of the part, we replicate real-world scanning conditions, including sensor noise and surface irregularities. This simulation-based approach enables offline trajectory planning based on CAD models. Key contributions include the modeling of the state space, action space, and reward function, specifically designed for inspection applications using profilometric sensors. We use Proximal Policy Optimization (PPO) algorithm to efficiently train the RL agent, demonstrating its capability to optimize inspection trajectories with profilometric sensors. To validate our approach, we conducted several experiments where a model trained on a specific training piece was tested on various parts in simulation. Also, we conducted a real-world experiment by executing the optimized trajectory, generated offline from a CAD model, to inspect a part using a UR3e robotic arm model.

Approximate nearest neighbor (ANN) search is a key component in many modern machine learning pipelines; recent use cases include retrieval-augmented generation (RAG) and vector databases. Clustering-based ANN algorithms, that use score computation methods based on product quantization (PQ), are often used in industrial-scale applications due to their scalability and suitability for distributed and disk-based implementations. However, they have slower query times than the leading graph-based ANN algorithms. In this work, we propose a new supervised score computation method based on the observation that inner product approximation is a multivariate (multi-output) regression problem that can be solved efficiently by reduced-rank regression. Our experiments show that on modern high-dimensional data sets, the proposed reduced-rank regression (RRR) method is superior to PQ in both query latency and memory usage. We also introduce LoRANN, a clustering-based ANN library that leverages the proposed score computation method. LoRANN is competitive with the leading graph-based algorithms and outperforms the state-of-the-art GPU ANN methods on high-dimensional data sets., Comment: Accepted to NeurIPS 2024

Knee osteoarthritis (OA) is a widespread chronic condition that impairs mobility and diminishes quality of life. Despite the proven benefits of exercise therapy and patient education in managing the OA symptoms pain and functional limitations, these strategies are often underutilized. Personalized outcome prediction models can help motivate and engage patients, but the accuracy of existing models in predicting changes in knee pain remains insufficiently examined. To validate existing models and introduce a concise personalized model predicting changes in knee pain before to after participating in a supervised education and exercise therapy program (GLA:D) for knee OA patients. Our models use self-reported patient information and functional measures. To refine the number of variables, we evaluated the variable importance and applied clinical reasoning. We trained random forest regression models and compared the rate of true predictions of our models with those utilizing average values. We evaluated the performance of a full, continuous, and concise model including all 34, all 11 continuous, and the six most predictive variables respectively. All three models performed similarly and were comparable to the existing model, with R-squares of 0.31-0.32 and RMSEs of 18.65-18.85 - despite our increased sample size. Allowing a deviation of 15 VAS points from the true change in pain, our concise model and utilizing the average values estimated the change in pain at 58% and 51% correctly, respectively. Our supplementary analysis led to similar outcomes. Our concise personalized prediction model more accurately predicts changes in knee pain following the GLA:D program compared to average pain improvement values. Neither the increase in sample size nor the inclusion of additional variables improved previous models. To improve predictions, new variables beyond those in the GLA:D are required.

Accurate ocean forecasting systems are vital for understanding marine dynamics, which play a crucial role in environmental management and climate adaptation strategies. Traditional numerical solvers, while effective, are computationally expensive and time-consuming. Recent advancements in machine learning have revolutionized weather forecasting, offering fast and energy-efficient alternatives. Building on these advancements, we introduce SeaCast, a neural network designed for high-resolution, medium-range ocean forecasting. SeaCast employs a graph-based framework to effectively handle the complex geometry of ocean grids and integrates external forcing data tailored to the regional ocean context. Our approach is validated through experiments at a high spatial resolution using the operational numerical model of the Mediterranean Sea provided by the Copernicus Marine Service, along with both numerical and data-driven atmospheric forcings., Comment: 28 pages, 35 figures. Accepted to the Tackling Climate Change with Machine Learning workshop at NeurIPS 2024

With the next generation of big telescopes such as the ELT and SKA it might become possible to measure changes in the expansion rate of the Universe in real time by measuring the change of the redshifts of a large number of galaxies over a period of the order of 10 years. This phenomenon, known as 'redshift drift,' will provide a crucial direct test of cosmological models. The change in redshift is readily explained using the concept of conformal time which is the comoving distance of a galaxy in lightyears. We emphasize that the redshift drift is directly proportional to the average change in the cosmic expansion rate between the time of a galaxy's light emission and its absorption. This phenomenon is illustrated within the framework of the concordance model, the Lambda-CDM model of the universe., Comment: 12 pages, 5 figures

Background: Patient-reported survey data are used to train prognostic models aimed at improving healthcare. However, such data are typically available multi-centric and, for privacy reasons, cannot easily be centralized in one data repository. Models trained locally are less accurate, robust, and generalizable. We present and apply privacy-preserving federated machine learning techniques for prognostic model building, where local survey data never leaves the legally safe harbors of the medical centers. Methods: We used centralized, local, and federated learning techniques on two healthcare datasets (GLA:D data from the five health regions of Denmark and international SHARE data of 27 countries) to predict two different health outcomes. We compared linear regression, random forest regression, and random forest classification models trained on local data with those trained on the entire data in a centralized and in a federated fashion. Results: In GLA:D data, federated linear regression (R2 0.34, RMSE 18.2) and federated random forest regression (R2 0.34, RMSE 18.3) models outperform their local counterparts (i.e., R2 0.32, RMSE 18.6, R2 0.30, RMSE 18.8) with statistical significance. We also found that centralized models (R2 0.34, RMSE 18.2, R2 0.32, RMSE 18.5, respectively) did not perform significantly better than the federated models. In SHARE, the federated model (AC 0.78, AUROC: 0.71) and centralized model (AC 0.84, AUROC: 0.66) perform significantly better than the local models (AC: 0.74, AUROC: 0.69). Conclusion: Federated learning enables the training of prognostic models from multi-center surveys without compromising privacy and with only minimal or no compromise regarding model performance.

In non-relativistic physics the concepts of geometry and topology are usually applied to characterise spatial structures, or structures in momentum space. We introduce the concept of temporal geometry, which encompasses the geometric and topological properties of temporal shapes, i.e. trajectories traced by a tip of a time-dependent vector in vector space. We apply it to vectors of ultrafast electron current or induced polarization in chiral molecules. The central concepts of temporal geometry - curvature and connection - emerge as ubiquitous features of photoexcited, non-equilibrium, chiral electron dynamics. We demonstrate that curvature and connection (i) rely on the interplay of molecular chirality and the polarization properties of light pulses, (ii) can be introduced for multiphoton processes, and (iii) control enantio-sensitive geometric observables via non-equilibrium electronic dynamics excited by tailored laser fields. Our findings may open a way to ultrafast, topologically non-trivial, and enantio-sensitive chemical dynamics., Comment: 32 pages, 5 figures

We use the Stein-Chen method to prove new explicit inequalities for the total variation, Wasserstein and local distances between the distribution of a random diagonal sum of a Bernoulli matrix and a Poisson distribution. Approximation results using a finite signed measure of higher order are given as well. Some of our bounds improve on those in Theorem 4.A of A.D. Barbour, L. Holst and S. Janson (Poisson approximation. Clarendon Press, Oxford, 1992)., Comment: 25 pages

We say of an isolated macroscopic quantum system in a pure state $\psi$ that it is in macroscopic thermal equilibrium if $\psi$ lies in or close to a suitable subspace $\mathcal{H}_{eq}$ of Hilbert space. It is known that every initial state $\psi_0$ will eventually reach macroscopic thermal equilibrium and stay there most of the time ("thermalize") if the Hamiltonian is non-degenerate and satisfies the appropriate version of the eigenstate thermalization hypothesis (ETH), i.e., that every eigenvector is in macroscopic thermal equilibrium. Shiraishi and Tasaki recently proved the ETH for a certain perturbation $H_\theta$ of the Hamiltonian $H_0$ of $N\gg 1$ free fermions on a one-dimensional lattice. The perturbation is needed to remove the high degeneracies of $H_0$. Here, we point out that also for degenerate Hamiltonians, all $\psi_0$ thermalize if the ETH holds for every eigenbasis, and we prove that this is the case for $H_0$. On top of that and more generally, we develop another strategy of proving thermalization, inspired by the fact that there is one eigenbasis of $H_0$ for which ETH can be proven more easily and with smaller error bounds than for the others. This strategy applies to arbitrarily small generic perturbations $H$ of $H_0$ and to arbitrary spatial dimensions. In fact, we consider any given $H_0$, suppose that the ETH holds for some but not necessarily every eigenbasis of $H_0$, and add a small generic perturbation, $H=H_0+\lambda V$ with $\lambda\ll 1$. Then, although $H$ (which is non-degenerate) may still not satisfy the ETH, we show that nevertheless (i) every $\psi_0$ thermalizes for most perturbations $V$, and more generally, (ii) for any subspace $\mathcal{H}_\nu$ (such as corresponding to a non-equilibrium macro state), most perturbations $V$ are such that most $\psi_0$ from $\mathcal{H}_\nu$ thermalize., Comment: 42 pages LaTeX, no figures; v2 minor improvements throughout the paper, further references added

We introduce an information theoretic criterion for Bayesian network structure learning which we call quotient normalized maximum likelihood (qNML). In contrast to the closely related factorized normalized maximum likelihood criterion, qNML satisfies the property of score equivalence. It is also decomposable and completely free of adjustable hyperparameters. For practical computations, we identify a remarkably accurate approximation proposed earlier by Szpankowski and Weinberger. Experiments on both simulated and real data demonstrate that the new criterion leads to parsimonious models with good predictive accuracy., Comment: Accepted to AISTATS 2018

This study explores the impact of unemployment benefits on employment quality, job stability, and tenure within complex labour market dynamics. Given the macroeconomic consequences of changes in unemployment benefits, including their impact on employment rates and output growth, we develop a closed macroeconomic model that integrates heterogeneous households and adaptive firms and incorporates real-world entry-exit market mechanisms. The model considers personal values, social norms, and social network formation among workers as we examine the role of social contacts in mediating the effects of unemployment benefits on job-matching quality and labour market outcomes. We simulate the model across various scenarios where unemployment benefit schemes differ in level and/or duration. Our results suggest that extending the duration of unemployment benefits does not necessarily improve job-matching quality. Longer benefits may indeed reduce the effectiveness of social networks in job finding, indicating that social contacts play a key role in labour market dynamics., Comment: 23 pages, 11 figures, 4 tables, 3 appendices

An isoperimetric inequality on the Hamming cube for exponents $\beta\ge 0.50057$ is proved, achieving equality on any subcube. This was previously known for $\beta\ge \log_2(3/2)\approx 0.585$. Improved bounds are also obtained at the critical exponent $\beta=0.5$, including a bound that is asymptotically sharp for small subsets. A key ingredient is a new Bellman-type function involving the Gaussian isoperimetric profile which appears to be a good approximation of the true envelope function. Verification uses computer-assisted proofs and interval arithmetic. Applications include progress towards a conjecture of Kahn and Park as well as sharp Poincar\'e inequalities for Boolean-valued functions near $L^1$., Comment: 46 pages, 9 figures; accompanying code at https://github.com/roos-j/dir24-isoperim

This paper focuses on robust stability and $H_\infty$ performance analyses of hybrid continuous/discrete time linear multi-rate control systems in the presence of parametric uncertainties. These affect the continuous-time plant in a rational way which is then modeled as a Linear Fractional Transformation (LFT). Based on a zero-order-hold (ZOH) LFT discretization process at the cost of bounded quantifiable approximations, and then using LFT-preserving down-sampling operations, a single-rate discrete-time closed-loop LFT model is derived. Interestingly, for any step inputs, and any admissible values of the uncertain parameters, the outputs of this model cover those of the initial hybrid multi-rate closed-loop system at every sampling time of the slowest control loop. Such an LFT model, which also captures the discretization errors, can then be used to evaluate both robust stability and guaranteed $H_\infty$ performance with a $\mu$-based approach. The proposed methodology is illustrated on a realistic and easily reproducible example inspired by the validation of multi-rate attitude control systems., Comment: 7 pages, 8 figures

Since BCS theory of superconductivity is non-linear, it is difficult to study superconducting properties analytically. There is a more tractable linear criterion which determines a temperature $T_l$ below which the system is superconducting. Here, we observe that there is a similar linear criterion which gives a temperature $T_u$ above which no superconductivity occurs. We provide examples of translation invariant systems where $T_u>T_l$ as well as systems where $T_u=T_l$. Furthermore, we estimate $T_u$ for half-spaces and show that it is exponentially small in the weak coupling limit, exhibiting the same asymptotics as the critical temperature for full space., Comment: 20 pages, 1 figure

Real-world applications of machine learning models are often subject to legal or policy-based regulations. Some of these regulations require ensuring the validity of the model, i.e., the approximation error being smaller than a threshold. A global metric is generally too insensitive to determine the validity of a specific prediction, whereas evaluating local validity is costly since it requires gathering additional data.We propose learning the model error to acquire a local validity estimate while reducing the amount of required data through active learning. Using model validation benchmarks, we provide empirical evidence that the proposed method can lead to an error model with sufficient discriminative properties using a relatively small amount of data. Furthermore, an increased sensitivity to local changes of the validity bounds compared to alternative approaches is demonstrated., Comment: 40th Conference on Uncertainty in Artificial Intelligence

Diffusion models are becoming defector generative models, which generate exceptionally high-resolution image data. Training effective diffusion models require massive real data, which is privately owned by distributed parties. Each data party can collaboratively train diffusion models in a federated learning manner by sharing gradients instead of the raw data. In this paper, we study the privacy leakage risk of gradient inversion attacks. First, we design a two-phase fusion optimization, GIDM, to leverage the well-trained generative model itself as prior knowledge to constrain the inversion search (latent) space, followed by pixel-wise fine-tuning. GIDM is shown to be able to reconstruct images almost identical to the original ones. Considering a more privacy-preserving training scenario, we then argue that locally initialized private training noise $\epsilon$ and sampling step t may raise additional challenges for the inversion attack. To solve this, we propose a triple-optimization GIDM+ that coordinates the optimization of the unknown data, $\epsilon$ and $t$. Our extensive evaluation results demonstrate the vulnerability of sharing gradient for data protection of diffusion models, even high-resolution images can be reconstructed with high quality.

The oscillating magnetic field produced by unbalanced currents in radio-frequency ion traps induces transition frequency shifts and sideband transitions that can be harmful to precision spectroscopy experiments. Here, we describe a methodology, based on two-photon spectroscopy, for determining both the strength and direction of rf-induced magnetic fields without modifying any DC magnetic bias field or changing any trap RF power. The technique is readily applicable to any trapped-ion experiment featuring narrow linewidth transitions., Comment: 13 pages, 6 figures

While diffusion models effectively generate remarkable synthetic images, a key limitation is the inference inefficiency, requiring numerous sampling steps. To accelerate inference and maintain high-quality synthesis, teacher-student distillation is applied to compress the diffusion models in a progressive and binary manner by retraining, e.g., reducing the 1024-step model to a 128-step model in 3 folds. In this paper, we propose a single-fold distillation algorithm, SFDDM, which can flexibly compress the teacher diffusion model into a student model of any desired step, based on reparameterization of the intermediate inputs from the teacher model. To train the student diffusion, we minimize not only the output distance but also the distribution of the hidden variables between the teacher and student model. Extensive experiments on four datasets demonstrate that our student model trained by the proposed SFDDM is able to sample high-quality data with steps reduced to as little as approximately 1%, thus, trading off inference time. Our remarkable performance highlights that SFDDM effectively transfers knowledge in single-fold distillation, achieving semantic consistency and meaningful image interpolation.

We study force gradient sensing by a coherently driven mechanical resonator with phase-sensitive detection of motion via the two-tone backaction evading measurement of cavity optomechanics. The response of the cavity to two coherent pumps is solved by numerical integration of the classical equations of motion, showing an extended region of monotonic response. We use Floquet theory to model the fluctuations, which rise only slightly above that of the usual backaction evading measurement in the presence of the mechanical drive. Our analysis indicates that this sensing technique is advantageous for applications such as Atomic Force Microscopy.

The Lightning Network, a payment channel network with a market cap of over 192M USD, is designed to resolve Bitcoin's scalability issues through fast off-chain transactions. There are multiple Lightning Network client implementations, all of which conform to the same textual specifications known as BOLTs. Several vulnerabilities have been manually discovered, but to-date there have been few works systematically analyzing the security of the Lightning Network. In this work, we take a foundational approach to analyzing the security of the Lightning Network with the help of formal methods. Based on the BOLTs' specifications, we build a detailed formal model of the Lightning Network's single-hop payment protocol and verify it using the Spin model checker. Our model captures both concurrency and error semantics of the payment protocol. We then define several security properties which capture the correct intermediate operation of the protocol, ensuring that the outcome is always certain to both channel peers, and using them we re-discover a known attack previously reported in the literature along with a novel attack, referred to as a Payout Race. A Payout Race consists of a particular sequence of events that can lead to an ambiguity in the protocol in which innocent users can unwittingly lose funds. We confirm the practicality of this attack by reproducing it in a local testbed environment., Comment: 16 pages, 7 figures, to appear at ACM CCS 2024

Low-frequency qubit control wires require non-trivial thermal anchoring and low-pass filtering. The resulting electron temperature serves as a quality benchmark for these signal lines. In this technical note, we make use of a primary electron thermometry technique, using a Coulomb blockade thermometer, to establish the electron temperature in the millikelvin regime. The experimental four-probe measurement setup, the data analysis, and the measurement limitations are discussed in detail. We verify the results by also using another electron thermometry technique, based on a superconductor-insulator-normal metal junction. Our comparison of signal lines with QDevil's QFilter to unfiltered signal lines demonstrates that the filter significantly reduces both the rms noise and electron temperature, which is measured to be 22 $\pm$ 1 mK., Comment: 8 pages, 4 figures, technical note

Probabilistic world models increase data efficiency of model-based reinforcement learning (MBRL) by guiding the policy with their epistemic uncertainty to improve exploration and acquire new samples. Moreover, the uncertainty-aware learning procedures in probabilistic approaches lead to robust policies that are less sensitive to noisy observations compared to uncertainty unaware solutions. We propose to combine trajectory sampling and deep Gaussian covariance network (DGCN) for a data-efficient solution to MBRL problems in an optimal control setting. We compare trajectory sampling with density-based approximation for uncertainty propagation using three different probabilistic world models; Gaussian processes, Bayesian neural networks, and DGCNs. We provide empirical evidence using four different well-known test environments, that our method improves the sample-efficiency over other combinations of uncertainty propagation methods and probabilistic models. During our tests, we place particular emphasis on the robustness of the learned policies with respect to noisy initial states.

It has been shown that unclocked, recurrent networks of Boolean gates in FPGAs can be used for low-SWaP reservoir computing. In such systems, topology and node functionality of the network are randomly initialized. To create a network that solves a task, weights are applied to output nodes and learning is achieved by adjusting those weights with conventional machine learning methods. However, performance is often limited compared to networks where all parameters are learned. Herein, we explore an alternative learning approach for unclocked, recurrent networks in FPGAs. We use evolutionary computation to evolve the Boolean functions of network nodes. In one type of implementation the output nodes are used directly to perform a task and all learning is via evolution of the network's node functions. In a second type of implementation a back-end classifier is used as in traditional reservoir computing. In that case, both evolution of node functions and adjustment of output node weights contribute to learning. We demonstrate the practicality of node function evolution, obtaining an accuracy improvement of ~30% on an image classification task while processing at a rate of over three million samples per second. We additionally demonstrate evolvability of network memory and dynamic output signals.