Search

The raw data are organized in folders based on the kind of device presented in the paper. The Juppyter notebooks contain the code for analyzing and plotting the data, there is one notebook for each figure of the papaer and supplementary. The scripts used to import the datasets of different kinds are contained in util_scripts, along with some other useful functions for plotting and formatting.

Introduction: A timely diagnosis of osteoporosis is key to reducing its growing clinical and economic burden. Radiofrequency Echographic Multi Spectrometry (REMS), a new diagnostic technology using an ultrasound approach, has been recognized by scientific associations as a facilitator of patients’ care pathway. We aimed at evaluating the costs of REMS vs. the conventional ionizing technology (dual-energy X-ray absorptiometry, DXA) for the diagnosis of osteoporosis from the perspective of the Italian National Health Service (NHS) using a cost-minimization analysis (CMA). Methods: We carried out structured qualitative interviews and a structured expert elicitation exercise to estimate healthcare resource consumption with a purposeful sample of clinical experts. For the elicitation exercise, an Excel tool was developed and, for each parameter, experts were asked to provide the lowest, highest and most likely value. Estimates provided by experts were averaged with equal weights. Unit costs were retrieved using different public sources. Results: Considering the base-case scenario (most likely value), the cost of professionals amounts to €31.9 for REMS and €48.8 for DXA, the cost of instrumental examinations and laboratory tests to €45.1 for REMS and €68.2 for DXA. Overall, in terms of current costs, REMS is associated with a mean saving for the NHS of €40.0 (range: €27.6-71.5) for each patient. Conclusions: REMS is associated with lower direct healthcare costs with respect to DXA. These results may inform policy-makers on the value of the REMS technology in the earlier diagnosis for osteoporosis, and support their decision regarding the reimbursement and diffusion of the technology in the Italian NHS.

The efficient control of a large number of qubits is one of the most challenging aspects for practical quantum computing. Current approaches in solid-state quantum technology are based on brute-force methods, where each and every qubit requires at least one unique control line—an approach that will become unsustainable when scaling to the required millions of qubits. Here, inspired by random-access architectures in classical electronics, we introduce the shared control of semiconductor quantum dots to efficiently operate a two-dimensional crossbar array in planar germanium. We tune the entire array, comprising 16 quantum dots, to the few-hole regime. We then confine an odd number of holes in each site to isolate an unpaired spin per dot. Moving forward, we demonstrate on a vertical and a horizontal double quantum dot a method for the selective control of the interdot coupling and achieve a tunnel coupling tunability over more than 10 GHz. The operation of a quantum electronic device with fewer control terminals than tunable experimental parameters represents a compelling step forward in the construction of scalable quantum technology., QCD/Veldhorst Lab, BUS/TNO STAFF, QCD/Vandersypen Lab, QCD/Scappucci Lab, QN/Veldhorst Lab

Highly uniform quantum systems are essential for the practical implementation of scalable quantum processors. While quantum dot spin qubits based on semiconductor technology are a promising platform for large-scale quantum computing, their small size makes them particularly sensitive to their local environment. Here, we present a method to electrically obtain a high degree of uniformity in the intrinsic potential landscape using hysteretic shifts of the gate voltage characteristics. We demonstrate the tuning of pinch-off voltages in quantum dot devices over hundreds of millivolts that then remain stable at least for hours. Applying our method, we homogenize the pinch-off voltages of the plunger gates in a linear array for four quantum dots, reducing the spread in pinch-off voltages by one order of magnitude. This work provides a new tool for the tuning of quantum dot devices and offers new perspectives for the implementation of scalable spin qubit arrays., QCD/Veldhorst Lab, BUS/TNO STAFF, QCD/Scappucci Lab, QN/Veldhorst Lab

The co-integration of spin, superconducting, and topological systems is emerging as an exciting pathway for scalable and high-fidelity quantum information technology. High-mobility planar germanium is a front-runner semiconductor for building quantum processors with spin-qubits, but progress with hybrid superconductor-semiconductor devices is hindered by the difficulty in obtaining a superconducting hard gap, that is, a gap free of subgap states. Here, we address this challenge by developing a low-disorder, oxide-free interface between high-mobility planar germanium and a germanosilicide parent superconductor. This superconducting contact is formed by the thermally-activated solid phase reaction between a metal, platinum, and the Ge/SiGe semiconductor heterostructure. Electrical characterization reveals near-unity transparency in Josephson junctions and, importantly, a hard induced superconducting gap in quantum point contacts. Furthermore, we demonstrate phase control of a Josephson junction and study transport in a gated two-dimensional superconductor-semiconductor array towards scalable architectures. These results expand the quantum technology toolbox in germanium and provide new avenues for exploring monolithic superconductor-semiconductor quantum circuits towards scalable quantum information processing., QCD/Scappucci Lab, QRD/Kouwenhoven Lab, QCD/Veldhorst Lab, BUS/TNO STAFF, QN/Veldhorst Lab

Multitemporal hyperspectral unmixing (MTHU) is a fundamental tool in the analysis of hyperspectral image sequences. It reveals the dynamical evolution of the materials (endmembers) and of their proportions (abundances) in a given scene. However, adequately accounting for the spatial and temporal variability of the endmembers in MTHU is challenging, and has not been fully addressed so far in unsupervised frameworks. In this work, we propose an unsupervised MTHU algorithm based on variational recurrent neural networks. First, a stochastic model is proposed to represent both the dynamical evolution of the endmembers and their abundances, as well as the mixing process. Moreover, a new model based on a low-dimensional parametrization is used to represent spatial and temporal endmember variability, significantly reducing the amount of variables to be estimated. We propose to formulate MTHU as a Bayesian inference problem. However, the solution to this problem does not have an analytical solution due to the nonlinearity and non-Gaussianity of the model. Thus, we propose a solution based on deep variational inference, in which the posterior distribution of the estimated abundances and endmembers is represented by using a combination of recurrent neural networks and a physically motivated model. The parameters of the model are learned using stochastic backpropagation. Experimental results show that the proposed method outperforms state of the art MTHU algorithms.

State inference and parameter learning in sequential models can be successfully performed with approximation techniques that maximize the evidence lower bound to the marginal log-likelihood of the data distribution. These methods may be referred to as Dynamical Variational Autoencoders, and our specific focus lies on the deep Kalman filter. It has been shown that the ELBO objective can oversimplify data representations, potentially compromising estimation quality. Tighter Monte Carlo objectives have been proposed in the literature to enhance generative modeling performance. For instance, the IWAE objective uses importance weights to reduce the variance of marginal log-likelihood estimates. In this paper, importance sampling is applied to the DKF framework for learning deep Markov models, resulting in the IW-DKF, which shows an improvement in terms of log-likelihood estimates and KL divergence between the variational distribution and the transition model. The framework using the sampled DKF update rule is also accommodated to address sequential state and parameter estimation when working with highly non-linear physics-based models. An experiment with the 3-space Lorenz attractor shows an enhanced generative modeling performance and also a decrease in RMSE when estimating the model parameters and latent states, indicating that tighter MCOs lead to improved state inference performance., Comment: To appear in the Proceedings of the Asilomar Conference on Signals, Systems, and Computers, October 2023, 5 pages, 3 figures, 3 tables

Although considerable effort has been dedicated to improving the solution to the hyperspectral unmixing problem, non-idealities such as complex radiation scattering and endmember variability negatively impact the performance of most existing algorithms and can be very challenging to address. Recently, deep learning-based frameworks have been explored for hyperspectral umixing due to their flexibility and powerful representation capabilities. However, such techniques either do not address the non-idealities of the unmixing problem, or rely on black-box models which are not interpretable. In this paper, we propose a new interpretable deep learning method for hyperspectral unmixing that accounts for nonlinearity and endmember variability. The proposed method leverages a probabilistic variational deep-learning framework, where disentanglement learning is employed to properly separate the abundances and endmembers. The model is learned end-to-end using stochastic backpropagation, and trained using a self-supervised strategy which leverages benefits from semi-supervised learning techniques. Furthermore, the model is carefully designed to provide a high degree of interpretability. This includes modeling the abundances as a Dirichlet distribution, the endmembers using low-dimensional deep latent variable representations, and using two-stream neural networks composed of additive piecewise-linear/nonlinear components. Experimental results on synthetic and real datasets illustrate the performance of the proposed method compared to state-of-the-art algorithms.