Pubblicazioni scientifiche
Successful treatment of life-threatening mycobacteriosis using adjunctive gamma-interferon therapy with genetic analysis.
Abstract No abstract Autori Confalonieri P, Maiocchi S, Salton F, Ruaro B, Rizzardi C, Volpe MC, Licastro D, Braga L, Confalonieri M Rivista IJTLD Open Data di pubblicazione 01/01/2024 Consulta la pubblicazione
Synthesis, characterization, functional testing and ageing analysis of bifunctional Zn-air battery GDEs, based on α-MnO2 nanowires and Ni/NiO nanoparticle electrocatalysts
Abstract Electrically rechargeable alkaline zinc air batteries (RZAB) – currently still at the R&D stage –, have great potential for stationary, as well as prospectively mobile, electrochemical energy storage applications. Their chief appeal is that they are made of abundant, environmentally friendly, intrinsically safe, and cheap materials, with established recycling concepts and auspicious life-cycle costs. One of the key weak points of present-generation RZAB programs is the air gas-diffusion electrode (GDE). In fact, on the one hand, GDE fabrication and testing are generally based on poorly understood protocols, and, on the other hand, performance is challenged by efficiency and durability issues. This work is centred on the fabrication of a novel bifunctional GDE for the air side of RZABs, on the assessment of its electrochemical performance and on the identification of factors impacting its efficiency and durability. The electrocatalysts for oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) are α-MnO2 nanowires and Ni/NiO nanoparticles, respectively. The composition of the active layer was optimized with rotating ring-disc electrode (RRDE) electrocatalysts tests. The GDEs were fabricated by spray-coating an ink, formulated with the electrocatalysts and the PTFE binder in an aqueous matrix. Fabrication and functional performance of GDEs – in pristine form and after ageing under realistic RZAB conditions – are rationalized on the basis of Scanning Electron Microscopy (SEM), Scanning Transmission X-ray SpectroMicroscopy (STXSM) at the Mn l-edge and Transmission Electron Microscopy (TEM) analyses. Imaging and spectral imaging disclosed the morphological and chemical-state evolution, brought about by electrochemical cycling. Special attention was devoted to the understanding of the role played by the presence of zincate in the electrolyte on the performance and ageing of the reversible air electrodes. Autori Salman Y., Waseem S., Alleva A., Banerjee P., Bonanni V., Emanuele E., Ciancio R., Gianoncelli A., Kourousias G., Li Bassi A., Macrelli A, Marini E., Rajak P., Bozzini B., Rivista Electrochimica Acta Data di pubblicazione 20/11/2023 Consulta la pubblicazione
The geometry of hidden representations of large transformer models
Abstract Large transformers are powerful architectures used for self-supervised data analysis across various data types, including protein sequences, images, and text. In these models, the semantic structure of the dataset emerges from a sequence of transformations between one representation and the next. We characterize the geometric and statistical properties of these representations and how they change as we move through the layers.By analyzing the intrinsic dimension (ID) and neighbor composition, we find that the representations evolve similarly in transformers trained on protein language taskand image reconstruction tasks. In the first layers, the data manifold expands, becoming high-dimensional, and then contracts significantly in the intermediate layers. In the last part of the model, the ID remains approximately constant or forms a second shallow peak. We show that the semantic information of the dataset is better expressed at the end of the first peak, and this phenomenon can be observed across many models trained on diverse datasets.Based on our findings, we point out an explicit strategy to identify, without supervision, the layers that maximize semantic content: representations at intermediate layers corresponding to a relative minimum of the ID profile are more suitable for downstream learning tasks. Autori Lucrezia Valeriani, Diego Doimo, Francesca Cuturello, Alessandro Laio, Alessio Ansuini, Alberto Cazzaniga Rivista Advances in Neural Information Processing Systems 36 (NEURIPS 2023) Main Conference Track Data di pubblicazione 21/09/2023 Consulta la pubblicazione
Unveiling the Electronic Structure of Pseudotetragonal WO3 Thin Films
Abstract WO3 is a 5d compound that undergoes several structural transitions in its bulk form. Its versatility is well-documented, with a wide range of applications, such as flexopiezoelectricity, electrochromism, gating-induced phase transitions, and its ability to improve the performance of Li-based batteries. The synthesis of WO3 thin films holds promise in stabilizing electronic phases for practical applications. However, despite its potential, the electronic structure of this material remains experimentally unexplored. Furthermore, its thermal instability limits its use in certain technological devices. Here, we employ tensile strain to stabilize WO3 thin films, which we call the pseudotetragonal phase, and investigate its electronic structure using a combination of photoelectron spectroscopy and density functional theory calculations. This study reveals the Fermiology of the system, notably identifying significant energy splittings between different orbital manifolds arising from atomic distortions. These splittings, along with the system’s thermal stability, offer a potential avenue for controlling inter- and intraband scattering for electronic applications. Autori Mazzola F., Hassani H., Amoroso D., Chaluvadi S.K., Fujii J., Polewczyk V., Rajak P., Koegler M. Ciancio R., Partoens B., Rossi G., Vobornik I., Ghosez P. and Orgiani P. Rivista Journal of Physical Chemistry Letters Data di pubblicazione 08/08/2023 Consulta la pubblicazione
The Interaction of Amines with Gold Nanoparticles
Abstract Here, we present an integrated ultra-high-vacuum (UHV) apparatus for the growth of complex materials and heterostructures. The specific growth technique is the Pulsed Laser Deposition (PLD) by means of a dual-laser source based on an excimer KrF ultraviolet and solid-state Nd:YAG infra-red lasers. By taking advantage of the two laser sources—both lasers can be independently used within the deposition chambers—a large number of different materials—ranging from oxides to metals, to selenides, and others—can be successfully grown in the form of thin films and heterostructures. All of the samples can be in situ transferred between the deposition chambers and the analysis chambers by using vessels and holders’ manipulators. The apparatus also offers the possibility to transfer samples to remote instrumentation under UHV conditions by means of commercially available UHV-suitcases. The dual-PLD operates for in-house research as well as user facility in combination with the Advanced Photo-electric Effect beamline at the Elettra synchrotron radiation facility in Trieste and allows synchrotron-based photo-emission as well as x-ray absorption experiments on pristine films and heterostructures. Autori Yanchao Lyu, Lucia Morillas Becerril, Mirko Vanzan, Stefano Corni, Mattia Cattelan, Gaetano Granozzi, Marco Frasconi, Piu Rajak, Pritam Banerjee, Fabrizio Mancin, Paolo Scrimin Rivista Advanced Materials Data di pubblicazione 23/03/2023 Consulta la pubblicazione
Evidence of silicide at the Ni/ β-Si3N4(0001)/Si(111) interface
Abstract We present a study of a sub-nanometre interlayer of crystalline silicon nitride at the Ni/Si interface. We performed transmission electron microscopy measurements complemented by energy dispersive X-ray analysis to investigate to what extent the nitride layer act as a barrier against atom diffusion. The results show that discontinuous silicide areas can form just below the nitride layer, whose composition is compatible with that of the nickel disilicide. The Ni–Si reaction is tentatively attributed to the thermal strain suffered by the interface during the deposition of Ni at low temperature. Autori Piu Rajak, Regina Ciancio, Antonio Caretta, Simone Laterza, Richa Bhardwaj, Matteo Jugovac, Marco Malvestuto, Paolo Moras, Roberto Flammini Rivista Applied Surface Science Data di pubblicazione 13/03/2023 Consulta la pubblicazione
Nd:YAG infrared laser as a viable alternative to excimer laser: YBCO case study
Abstract We report on the growth and characterization of epitaxial YBa2Cu3O7−δ (YBCO) complex oxide thin films and related heterostructures exclusively by Pulsed Laser Deposition (PLD) and using first harmonic Nd:Y3Al5O12 (Nd:YAG) pulsed laser source (λ = 1064 nm). High-quality epitaxial YBCO thin film heterostructures display superconducting properties with transition temperature ∼ 80 K. Compared with the excimer lasers, when using Nd:YAG lasers, the optimal growth conditions are achieved at a large target-to-substrate distance d. These results clearly demonstrate the potential use of the first harmonic Nd:YAG laser source as an alternative to the excimer lasers for the PLD thin film community. Its compactness as well as the absence of any safety issues related to poisonous gas represent a major breakthrough in the deposition of complex multi-element compounds in form of thin films. Autori Sandeep Kumar Chaluvadi, Shyni Punathum Chalil, Federico Mazzola, Simone Dolabella, Piu Rajak, Marcello Ferrara, Regina Ciancio, Jun Fujii, Giancarlo Panaccione, Giorgio Rossi & Pasquale Orgiani Rivista Scientific Reports Data di pubblicazione 08/03/2023 Consulta la pubblicazione
Dual pulsed laser deposition system for the growth of complex materials and heterostructures
Abstract Here, we present an integrated ultra-high-vacuum (UHV) apparatus for the growth of complex materials and heterostructures. The specific growth technique is the Pulsed Laser Deposition (PLD) by means of a dual-laser source based on an excimer KrF ultraviolet and solid-state Nd:YAG infra-red lasers. By taking advantage of the two laser sources—both lasers can be independently used within the deposition chambers—a large number of different materials—ranging from oxides to metals, to selenides, and others—can be successfully grown in the form of thin films and heterostructures. All of the samples can be in situ transferred between the deposition chambers and the analysis chambers by using vessels and holders’ manipulators. The apparatus also offers the possibility to transfer samples to remote instrumentation under UHV conditions by means of commercially available UHV-suitcases. The dual-PLD operates for in-house research as well as user facility in combination with the Advanced Photo-electric Effect beamline at the Elettra synchrotron radiation facility in Trieste and allows synchrotron-based photo-emission as well as x-ray absorption experiments on pristine films and heterostructures. Autori Orgiani P.; Chaluvadi S.K.; Chalil, S. Punathum; Mazzola F.; Jana A.; Dolabella S.; Rajak P.; Ferrara M.; Benedetti D.; Fondacaro A.; Salvador F.; Ciancio R.; Fujii J.; Panaccione G.; Vobornik I.; Rossi G. Rivista Review of Scientific Instruments Data di pubblicazione 06/03/2023 Consulta la pubblicazione
Speeding‐up pruning for Artificial Neural Networks: Introducing Accelerated Iterative Magnitude Pruning
Abstract: In recent years, Artificial Neural Networks (ANNs) pruning has become the focal point of many researches, due to the extreme overparametrization of such models. This has urged the scientific world to investigate methods for the simplification of the structure of weights in ANNs, mainly in an effort to reduce time for both training and inference. Frankle and Carbin [1], and later Renda, Frankle, and Carbin [2] introduced and refined an iterative pruning method which is able to effectively prune the network of a great portion of its parameters with little to no loss in performance. On the downside, this method requires a large amount of time for its application, since, for each iteration, the network has to be trained for (almost) the same amount of epochs of the unpruned network. In this work, we show that, for a limited setting, if targeting high overall sparsity rates, this time can be effectively reduced for each iteration, save for the last one, by more than 50%, while yielding a final product (i.e., final pruned network) whose performance is comparable to the ANN obtained using the existing method. Autori: Marco Zullich, Eric Medvet; Felice Andrea Pellegrino; Alessio Ansuini Rivista: 2020 25th International Conference on Pattern Recognition (ICPR) Data di pubblicazione: 05/05/2021 Consulta la pubblicazione
Investigating Similarity Metrics for Convolutional Neural Networks in the Case of Unstructured Pruning
Abstract: Deep Neural Networks (DNNs) are essential tools of modern science and technology. The current lack of explainability of their inner workings and of principled ways to tame their architectural complexity triggered a lot of research in recent years. There is hope that, by making sense of representations in their hidden layers, we could collect insights on how to reduce model complexity—without performance degradation—by pruning useless connections. It is natural then to ask the following question: how similar are representations in pruned and unpruned models? Even small insights could help in finding principled ways to design good lightweight models, enabling significant savings of computation, memory, time and energy. In this work, we investigate empirically this problem on a wide spectrum of similarity measures, network architectures and datasets. We find that the results depend critically on the similarity measure used and we discuss briefly the origin of these differences, concluding that further investigations are required in order to make substantial advances. Autori: Alessio Ansuini, Eric Medvet, Felice Andrea Pellegrino, Marco Zullich Rivista: International Conference on Pattern Recognition Applications and Methods (ICPRAM) Data di pubblicazione: 23/12/2020 Consulta la pubblicazione
Hierarchical nucleation in deep neural networks
Abstract Deep convolutional networks (DCNs) learn meaningful representations where data that share the same abstract characteristics are positioned closer and closer. Understanding these representations and how they are generated is of unquestioned practical and theoretical interest. In this work we study the evolution of the probability density of the ImageNet dataset across the hidden layers in some stateof-the-art DCNs. We find that the initial layers generate a unimodal probability density getting rid of any structure irrelevant for classification. In subsequent layers density peaks arise in a hierarchical fashion that mirrors the semantic hierarchy of the concepts. Density peaks corresponding to single categories appear only close to the output and via a very sharp transition which resembles the nucleation process of a heterogeneous liquid. This process leaves a footprint in the probability density of the output layer where the topography of the peaks allows reconstructing the semantic relationships of the categories. Autori Diego Doimo, Aldo Glielmo, Alessio Ansuini, Alessandro Laio Rivista NIPS’20: Proceedings of the 34th International Conference on Neural Information Processing Systems Data di pubblicazione 06/12/2020 Consulta la pubblicazione