Ugo Moschini
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Ugo Moschini has a total of 57 co-authors. Here are the top co-authors by number of shared publications:
Michael H. F. Wilkinson
Johann Bernoulli Institute, University of Groningen, Groningen, The Netherlands
4 shared publicationsScott C. Trager
Kapteyn Astronomical Institute, University of Groningen, P.O. Box 407, 9700 AK Groningen, The Netherlands
3 shared publications- Valentina Pasquale
Italian Institute of Technology (IIT)
3 shared publications - Curtis J. M. Sharma
Delft University of Technology (TU Delft)
2 shared publications - Delwen Louise Franzen
QUEST Center for Responsible Research, Berlin Institute of Health at Charité - Universitätsmedizin Berlin, Berlin, Germany
2 shared publications - Elda Osmenaj
Italian Institute of Technology (IIT)
2 shared publications
Latest Publications
- Open Access
- 28 January 2025
- Published by Edinburgh University Library in International Journal of Digital Curation
- Vol. 19 (1) , 6
FAIR Principles Implementation in ML/AI - Findings from Skills4EOSC Delphi Study
Implementing the FAIR (Findable, Accessible, Interoperable, and Reusable) principles for scientific data management in machine learning (ML) and artificial intelligence (AI) offers numerous benefits, including higher model reliability, more collaborative research, and greater reproducibility. Despite these advantages, there is a lack of clear, practical guidelines for improving the FAIRness of ML/AI outputs, especially the models. To address this gap, Skills4EOSC Task 6.3.3a of Work Package 6: Professional Networks for Lifelong Learning conducted a Delphi Study to gather expert consensus on implementing FAIR principles in ML/AI model development. A Delphi study, involving two rounds of surveys followed by an online meeting, was conducted. In the first round, ML/AI experts from Europe and beyond rated suggested FAIR practices and proposed additional ones. The second round involved feedback and re-evaluation of these practices. The final meeting included detailed discussions on the Top 10 practices for FAIR principles implementation in ML/AI. The resulting Top 10 practices aim to provide guidelines for researchers and data management professionals to implement FAIR principles in ML/AI.
- 28 January 2025
- Published by Edinburgh University Library in International Journal of Digital Curation
- Vol. 19 (1) , 6
Open AccessFAIR Principles Implementation in ML/AI - Findings from Skills4EOSC Delphi Study
Implementing the FAIR (Findable, Accessible, Interoperable, and Reusable) principles for scientific data management in machine learning (ML) and artificial intelligence (AI) offers numerous benefits, including higher model reliability, more collaborative research, and greater reproducibility. Despite these advantages, there is a lack of clear, practical guidelines for improving the FAIRness of ML/AI outputs, especially the models. To address this gap, Skills4EOSC Task 6.3.3a of Work Package 6: Professional Networks for Lifelong Learning conducted a Delphi Study to gather expert consensus on implementing FAIR principles in ML/AI model development. A Delphi study, involving two rounds of surveys followed by an online meeting, was conducted. In the first round, ML/AI experts from Europe and beyond rated suggested FAIR practices and proposed additional ones. The second round involved feedback and re-evaluation of these practices. The final meeting included detailed discussions on the Top 10 practices for FAIR principles implementation in ML/AI. The resulting Top 10 practices aim to provide guidelines for researchers and data management professionals to implement FAIR principles in ML/AI.
- Open Access
- 28 January 2025
- Published by Edinburgh University Library in International Journal of Digital Curation
- Vol. 19 (1) , 6
FAIR Principles Implementation in ML/AI - Findings from Skills4EOSC Delphi Study
Implementing the FAIR (Findable, Accessible, Interoperable, and Reusable) principles for scientific data management in machine learning (ML) and artificial intelligence (AI) offers numerous benefits, including higher model reliability, more collaborative research, and greater reproducibility. Despite these advantages, there is a lack of clear, practical guidelines for improving the FAIRness of ML/AI outputs, especially the models. To address this gap, Skills4EOSC Task 6.3.3a of Work Package 6: Professional Networks for Lifelong Learning conducted a Delphi Study to gather expert consensus on implementing FAIR principles in ML/AI model development. A Delphi study, involving two rounds of surveys followed by an online meeting, was conducted. In the first round, ML/AI experts from Europe and beyond rated suggested FAIR practices and proposed additional ones. The second round involved feedback and re-evaluation of these practices. The final meeting included detailed discussions on the Top 10 practices for FAIR principles implementation in ML/AI. The resulting Top 10 practices aim to provide guidelines for researchers and data management professionals to implement FAIR principles in ML/AI.
- 28 January 2025
- Published by Edinburgh University Library in International Journal of Digital Curation
- Vol. 19 (1) , 6
Open AccessFAIR Principles Implementation in ML/AI - Findings from Skills4EOSC Delphi Study
Implementing the FAIR (Findable, Accessible, Interoperable, and Reusable) principles for scientific data management in machine learning (ML) and artificial intelligence (AI) offers numerous benefits, including higher model reliability, more collaborative research, and greater reproducibility. Despite these advantages, there is a lack of clear, practical guidelines for improving the FAIRness of ML/AI outputs, especially the models. To address this gap, Skills4EOSC Task 6.3.3a of Work Package 6: Professional Networks for Lifelong Learning conducted a Delphi Study to gather expert consensus on implementing FAIR principles in ML/AI model development. A Delphi study, involving two rounds of surveys followed by an online meeting, was conducted. In the first round, ML/AI experts from Europe and beyond rated suggested FAIR practices and proposed additional ones. The second round involved feedback and re-evaluation of these practices. The final meeting included detailed discussions on the Top 10 practices for FAIR principles implementation in ML/AI. The resulting Top 10 practices aim to provide guidelines for researchers and data management professionals to implement FAIR principles in ML/AI.
- Open Access
- 21 March 2024
- Published by Springer Nature in Nature Medicine
- Vol. 30 (5) , 1220-1221
Understanding the provenance and quality of methods is essential for responsible reuse of FAIR data
- 21 March 2024
- Published by Springer Nature in Nature Medicine
- Vol. 30 (5) , 1220-1221
Open AccessUnderstanding the provenance and quality of methods is essential for responsible reuse of FAIR data
- Open Access
Establishing a core set of open science practices in biomedicine: a modified Delphi study
Background Mandates and recommendations related to embedding open science practices within the research lifecycle are increasingly common. Few stakeholders, however, are monitoring compliance to their mandates or recommendations. It is necessary to monitor the current state of open science to track changes over time and to identify areas to create interventions to drive improvements. Monitoring open science practices requires that they are defined and operationalized. Involving the biomedical community, we sought to reach consensus on a core set of open science practices to monitor at biomedical research institutions. Methods and Findings To establish consensus in a structured and systematic fashion, we conducted a modified 3-round Delphi study. Participants in Round 1 were 80 individuals from 20 biomedical research institutions that exhibit interest in or actively support open science. Participants were research administrators, researchers, specialists in dedicated open science roles, and librarians. In Rounds 1 and 2, participants completed an online survey evaluating a set of potential open science practices that could be important and meaningful to monitor in an automated institutional open science dashboard. Participants voted on the inclusion of each item and provided a rationale for their choice. We defined consensus as 80% agreement. Between rounds, participants received aggregated voting scores for each item and anonymized comments from all participants, and were asked to re-vote on items that did not reach consensus. For Round 3, we hosted two half- day virtual meetings with 21 and 17 participants respectively to discuss and vote on all items that had not reached consensus after Round 2. Ultimately, participants reached consensus to include a 19 open science practices. Conclusions A group of international stakeholders used a modified Delphi process to agree upon open science practices to monitor in a proposed open science dashboard for biomedical institutions. The core set of 19 open science practices identified by participants will form the foundation for institutional dashboards that display compliance with open science practices. They will now be assessed and tested for automatic inclusion in terms of technical feasibility. Using user-centered design, participating institutions will be involved in creating a dashboard prototype, which can then be implemented to monitor rates of open science practices at biomedical institutions. Our methods and approach may also transfer to other research settings–other disciplines could consider using our consensus list as a starting point for agreement upon a discipline-specific set of open science practices to monitor. The findings may also be of broader value to the development of policy, education, and interventions.
Open AccessEstablishing a core set of open science practices in biomedicine: a modified Delphi study
Background Mandates and recommendations related to embedding open science practices within the research lifecycle are increasingly common. Few stakeholders, however, are monitoring compliance to their mandates or recommendations. It is necessary to monitor the current state of open science to track changes over time and to identify areas to create interventions to drive improvements. Monitoring open science practices requires that they are defined and operationalized. Involving the biomedical community, we sought to reach consensus on a core set of open science practices to monitor at biomedical research institutions. Methods and Findings To establish consensus in a structured and systematic fashion, we conducted a modified 3-round Delphi study. Participants in Round 1 were 80 individuals from 20 biomedical research institutions that exhibit interest in or actively support open science. Participants were research administrators, researchers, specialists in dedicated open science roles, and librarians. In Rounds 1 and 2, participants completed an online survey evaluating a set of potential open science practices that could be important and meaningful to monitor in an automated institutional open science dashboard. Participants voted on the inclusion of each item and provided a rationale for their choice. We defined consensus as 80% agreement. Between rounds, participants received aggregated voting scores for each item and anonymized comments from all participants, and were asked to re-vote on items that did not reach consensus. For Round 3, we hosted two half- day virtual meetings with 21 and 17 participants respectively to discuss and vote on all items that had not reached consensus after Round 2. Ultimately, participants reached consensus to include a 19 open science practices. Conclusions A group of international stakeholders used a modified Delphi process to agree upon open science practices to monitor in a proposed open science dashboard for biomedical institutions. The core set of 19 open science practices identified by participants will form the foundation for institutional dashboards that display compliance with open science practices. They will now be assessed and tested for automatic inclusion in terms of technical feasibility. Using user-centered design, participating institutions will be involved in creating a dashboard prototype, which can then be implemented to monitor rates of open science practices at biomedical institutions. Our methods and approach may also transfer to other research settings–other disciplines could consider using our consensus list as a starting point for agreement upon a discipline-specific set of open science practices to monitor. The findings may also be of broader value to the development of policy, education, and interventions.
- Open Access
- 30 March 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 40 (3) , 513-526
A Hybrid Shared-Memory Parallel Max-Tree Algorithm for Extreme Dynamic-Range Images
Max-trees, or component trees, are graph structures that represent the connected components of an image in a hierarchical way. Nowadays, many application fields rely on images with high-dynamic range or floating point values. Efficient sequential algorithms exist to build trees and compute attributes for images of any bit depth. However, we show that the current parallel algorithms perform poorly already with integers at bit depths higher than 16 bits per pixel. We propose a parallel method combining the two worlds of flooding and merging max-tree algorithms. First, a pilot max-tree of a quantized version of the image is built in parallel using a flooding method. Later, this structure is used in a parallel leaf-to-root approach to compute efficiently the final max-tree and to drive the merging of the sub-trees computed by the threads. We present an analysis of the performance both on simulated and actual 2D images and 3D volumes. Execution times are about 20× better than the fastest sequential algorithm and speed-up goes up to 30-40 on 64 threads.
- 30 March 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 40 (3) , 513-526
Open AccessA Hybrid Shared-Memory Parallel Max-Tree Algorithm for Extreme Dynamic-Range Images
Max-trees, or component trees, are graph structures that represent the connected components of an image in a hierarchical way. Nowadays, many application fields rely on images with high-dynamic range or floating point values. Efficient sequential algorithms exist to build trees and compute attributes for images of any bit depth. However, we show that the current parallel algorithms perform poorly already with integers at bit depths higher than 16 bits per pixel. We propose a parallel method combining the two worlds of flooding and merging max-tree algorithms. First, a pilot max-tree of a quantized version of the image is built in parallel using a flooding method. Later, this structure is used in a parallel leaf-to-root approach to compute efficiently the final max-tree and to drive the merging of the sub-trees computed by the threads. We present an analysis of the performance both on simulated and actual 2D images and 3D volumes. Execution times are about 20× better than the fastest sequential algorithm and speed-up goes up to 30-40 on 64 threads.
- Open Access
- 1 September 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1046-1050
Improving background estimation for faint astronomical object detection
Estimation of the background is an essential step in automated extraction of faint, extended objects from large-scale, optical surveys in astronomy. In this paper we present an improvement on the background estimation method of a commonly used tool in this field: Source Extractor (SEx-tractor). We show that the original method suffers from bias caused by presence of extended sources, and present an alternative which greatly reduces this effect, leading to much better preservation of faint extended structures.
- 1 September 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1046-1050
Open AccessImproving background estimation for faint astronomical object detection
Estimation of the background is an essential step in automated extraction of faint, extended objects from large-scale, optical surveys in astronomy. In this paper we present an improvement on the background estimation method of a commonly used tool in this field: Source Extractor (SEx-tractor). We show that the original method suffers from bias caused by presence of extended sources, and present an alternative which greatly reduces this effect, leading to much better preservation of faint extended structures.
- Open Access
- 26 August 2015
- Published by Springer Nature
- p. 121-133
Parallel 2D Local Pattern Spectra of Invariant Moments for Galaxy Classification
- 26 August 2015
- Published by Springer Nature
- p. 121-133
Open AccessParallel 2D Local Pattern Spectra of Invariant Moments for Galaxy Classification
- Open Access
- 1 January 2015
- Published by Springer Nature
- p. 669-680
Viscous-Hyperconnected Attribute Filters: A First Algorithm
- 1 January 2015
- Published by Springer Nature
- p. 669-680
Open AccessViscous-Hyperconnected Attribute Filters: A First Algorithm
- Open Access
- 1 January 2015
- Published by Springer Nature
- p. 157-168
Improved Detection of Faint Extended Astronomical Objects Through Statistical Attribute Filtering
- 1 January 2015
- Published by Springer Nature
- p. 157-168
Open AccessImproved Detection of Faint Extended Astronomical Objects Through Statistical Attribute Filtering
- Open Access
- 1 January 2013
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 484-495
Mask Connectivity by Viscous Closings: Linking Merging Galaxies without Merging Double Stars
- 1 January 2013
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 484-495
Open AccessMask Connectivity by Viscous Closings: Linking Merging Galaxies without Merging Double Stars