Archivio della ricerca - Fondazione Bruno Kessler
Not a member yet
21227 research outputs found
Sort by
The Warmup Dilemma: How Learning Rate Strategies Impact Speech-to-Text Model Convergence
Full text linkTraining large-scale models presents challenges not only in terms of resource requirements but also in terms of their convergence. For this reason, the learning rate (LR) is often decreased when the size of a model is increased. Such a simple solution is not enough in the case of speech-to-text (S2T) trainings, where evolved and more complex variants of the Transformer architecture – e.g., Conformer or Branchformer – are used in light of their better performance. As a workaround, OWSM designed a double linear warmup of the LR, increasing it to a very small value in the first phase before updating it to a higher value in the second phase. While this solution worked well in practice, it was not compared with alternative solutions, nor was the impact on the final performance of different LR warmup schedules studied. This paper fills this gap, revealing that i) large-scale S2T trainings demand a sub-exponential LR warmup, and ii) a higher LR in the warmup phase accelerates initial convergence, but it does not boost final performance
Synthetic-lattice Bloch wave dynamics in a single-mode microwave resonator
No full textFrequency-based synthetic dimensions are a promising avenue for expanding the dimensionality of photonic systems. In this work, we show how a tilted synthetic lattice is naturally realized by periodically modulating a single-mode resonator under a coherent monochromatic drive. We theoretically study the Bloch wave dynamics in the tilted synthetic lattice, which gives rise to peculiar features in the spectral distribution of the cavity field. Our predictions are experimentally confirmed using a tunable superconducting microwave resonator
Measurement of off-shell Higgs boson production in the H*→ZZ→4l decay channel using a neural simulation-based inference technique in 13 TeV pp collisions with the ATLAS detector
No full textA measurement of off-shell Higgs boson production in theH∗→ZZ→4ldecay channel is presented. The measurement uses 140 fb-1of proton-proton collisions ats=13TeV collected by the ATLAS detector at the Large Hadron Collider and supersedes the previous result in this decay channel using the same dataset. The data analysis is performed using a neural simulation-based inference method, which builds per-event likelihood ratios using neural networks. The observed (expected) off-shell Higgs boson production signal strength in theZZ→4ldecay channel at 68% CL is0.87-0.54+0.75(1.00-0.95+1.04). The evidence for off-shell Higgs boson production using theZZ→4ldecay channel has an observed (expected) significance of 2.5σ(1.3σ). The expected result represents a significant improvement relative to that of the previous analysis of the same dataset, which obtained an expected significance of 0.5σ. When combined with the most recent ATLAS measurement in theZZ→2l2νdecay channel, the evidence for off-shell Higgs boson production has an observed (expected) significance of 3.7σ(2.4σ). The off-shell measurements are combined with the measurement of on-shell Higgs boson production to obtain constraints on the Higgs boson total width. The observed (expected) value of the Higgs boson width at 68% CL is4.3-1.9+2.7(4.1-3.4+3.5) MeV
Microwave Sensing of Humidity with a Transparent and Flexible Electromagnetic Metasurface
No full textA planar, low-cost, and flexible microwave sensor is introduced, utilizing an electromagnetic metasurface of frequency-selective surface (FSS) type. This sensor concept relies on altering the inter-plate capacitance of the electric inductive-capacitive (ELC) resonators forming the FSS periodic array grid. When a material with varying electrical characteristics is placed between the plates, changes in capacitance and inductance lead to a shift and attenuation of the resonant frequency that are exploited as encoding mechanism for the sensing functionality. The metasurface approach offers the advantage of enhancing the sensor's radar cross section (RCS), which in turn improves either the reading distance or the strength of the sensor response
Chatbot-Based Version of a World Health Organization-Validated Intervention for Stress Management in Patients With Breast Cancer (Self-Help Plus): Protocol for a Pilot Feasibility Study
No full textBackground: Emerging digital tools play an innovative and key role in supporting women's psychological well-being throughout the different stages and challenges of cancer. The development and adoption of digital interventions, including chatbots and virtual coaches within smartphone apps, are increasingly recognized as valuable resources for enhancing women's mental health. Objective: The aim of this paper is to present the research protocol for a pilot study designed as a proof-of-concept investigation. The study evaluates the feasibility, acceptability, and perceived utility of a mobile app delivering an acceptance and commitment therapy-based stress management intervention. The intervention is delivered through ALBA (A Well-Being Assistant), a virtual coach embedded within the TreC (an acronym for cartella clinica del cittadino, meaning "citizen's electronic health record") research platform-a mobile health ecosystem designed to support research and digital health interventions. ALBA guides users through 5 coaching sessions tailored for women undergoing breast cancer (BC) treatment. The chatbot-delivered app is an adaptation of Self-Help Plus, a World Health Organization (WHO)-validated stress management intervention, and is provided in text, audio, and video formats. The intervention's potential impact on participants' psychological well-being is also explored. Methods: A convenience sample size of 50 participants will be identified to meet the study's objectives. Participants will be recruited using a convenience sampling approach from women receiving care at the Breast Unit of the Azienda Provinciale per Servizi Sanitari di Trento. ALBA will interact with the participants for 6 weeks. Specifically, there will be 1 coaching session per week, followed by weekly assigned acceptance and commitment therapy exercises to be performed between sessions. Results: The app is expected to demonstrate high usability and engagement, aligning with the WHO Self-Help Plus protocol. Improvements in psychological well-being and quality of life are anticipated. Data from this pilot will be analyzed using both quantitative and qualitative methods, with a focus on assessing feasibility, acceptability, and perceived utility and usability in supporting women during BC treatment. Conclusions: Existing literature indicates a promising role for new technologies in delivering validated mental health interventions, highlighting the potential of digital interventions to address barriers related to social stigma and seeking assistance. This pilot is expected to provide valuable insights on the potential acceptability and usefulness of providing consistent mobile health psychoeducational support to women throughout the course of BC. International registered report identifier (irrid): PRR1-10.2196/65837
Job Unfair: An Investigation of Gender and Occupational Bias in Free-Form Text Completions by LLMs
Full text linkDisentangling how gender and occupations are encoded by LLMs is crucial to identify possible biases and prevent harms, especially given the widespread use of LLMs in sensitive domains such as human resources.In this work, we carry out an in-depth investigation of gender and occupational biases in English and Italian as expressed by 9 different LLMs (both base and instruction-tuned). Specifically, we focus on the analysis of sentence completions when LLMs are prompted with job-related sentences including different gender representations. We carry out a manual analysis of 4,500 generated texts over 4 dimensions that can reflect bias, we propose a novel embedding-based method to investigate biases in generated texts and, finally, we carry out a lexical analysis of the model completions. In our qualitative and quantitative evaluation we show that many facets of social bias remain unaccounted for even in aligned models, and LLMs in general still reflect existing gender biases in both languages. Finally, we find that models still struggle with gender-neutral expressions, especially beyond English
Neural Network-Based Temporal Ensembling of Water Depth Estimates Derived from SuperDove Images
Full text linkCubeSats provide a wealth of high-frequency observations at a meter-scale spatial resolution. However, most current methods of inferring water depth from satellite data consider only a single image. This approach is sensitive to the radiometric quality of the data acquired at that particular instant in time, which could be degraded by various confounding factors, such as sun glint or atmospheric effects. Moreover, using single images in isolation fails to exploit recent improvements in the frequency of satellite image acquisition. This study aims to leverage the dense image time series from the SuperDove constellation via an ensembling framework that helps to improve empirical (regression-based) bathymetry retrieval. Unlike previous studies that only ensembled the original spectral data, we introduce a neural network-based method that instead ensembles the water depths derived from multi-temporal imagery, provided the data are acquired under steady flow conditions. We refer to this new approach as NN-depth ensembling. First, every image is treated individually to derive multitemporal depth estimates. Then, we use another NN regressor to ensemble the temporal water depths. This step serves to automatically weight the contribution of the bathymetric estimates from each time instance to the final bathymetry product. Unlike methods that ensemble spectral data, NN-depth ensembling mitigates against propagation of uncertainties in spectral data (e.g., noise due to sun glint) to the final bathymetric product. The proposed NN-depth ensembling is applied to temporal SuperDove imagery of reaches from the American, Potomac, and Colorado rivers with depths of up to 10 m and evaluated against in situ measurements. The proposed method provided more accurate and robust bathymetry retrieval than single-image analyses and other ensembling approaches
Exploration at the high-energy frontier: ATLAS Run 2 searches investigating the exotic jungle beyond the Standard Model
No full textThis report presents a comprehensive collection of searches for new physics performed by the ATLAS Collaboration during the Run 2 period of data taking at the Large Hadron Collider, from 2015 to 2018, corresponding to about 140 fb−1 of √s = 13 TeV proton–proton collision data. These searches cover a variety of beyond-the-standard model topics such as dark matter candidates, new vector bosons, hidden-sector particles, leptoquarks, or vector-like quarks, among others. Searches for supersymmetric particles or extended Higgs sectors are explicitly excluded as these are the subject of separate reports by the Collaboration. For each topic, the most relevant searches are described, focusing on their importance and sensitivity and, when appropriate, highlighting the experimental techniques employed. In addition to the description of each analysis, complementary searches are compared, and the overall sensitivity of the ATLAS experiment to each type of new physics is discussed. Summary plots and statistical combinations of multiple searches are included whenever possible
A SPAD-Based Optical Encoder
No full textThis paper reports on a proof-of-concept SPAD-based (Single Photon Avalanche Diode) optical encoder. The work aims at demonstrating the advantages of SPADs over photodiodes, which are typically used in the current optical position measuring systems. In addition to their high sensitivity and high speed, SPADs allow fully digital signal processing, offering a large system flexibility and scalability toward advanced CMOS technologies. Preliminary tests have been carried out using an array of 100 x 100 SPADs, coupled with an optical Gray-coded disk and a laser diode. Binary frame sequences were acquired and processed off-line through a lightweight algorithm to reproduce the disc code. The described algorithm aims at being integrated in the same chip of the sensor to speed up the signal processing chain, thus allowing high rotation speeds to be achieved. Experimental results are reported, together with future work and conclusions
