Local kernel renormalization as a mechanism for feature learning in overparametrized convolutional neural networks.

Empirical evidence shows that fully-connected neural networks in the infinite-width limit (lazy training) eventually outperform their finite-width counterparts in most computer vision tasks; on the other hand, modern architectures with convolutional layers often achieve optimal performances in the finite-width regime. In this work, we present a theoretical framework that provides a rationale for these differences in one-hidden-layer networks; we derive an effective action in the so-called proportional limit for an architecture with one convolutional hidden layer and compare it with the result available for fully-connected networks. Remarkably, we identify a completely different form of kernel renormalization: whereas the kernel of the fully-connected architecture is just globally renormalized by a single scalar parameter, the convolutional kernel undergoes a local renormalization, meaning that the network can select the local components that will contribute to the final prediction in a data-dependent way.

High Prevalence of High-Risk HPV Among People with and Without HIV: Insights into Risk Factors for Tailored Screening Approaches.

Men who have sex with men (MSM) and people with HIV are at increased risk of anal HPV infection and cancer. This study aimed to assess the prevalence of anal HPV among MSM with HIV (MWH) and without HIV (MWoH), as well as among MSM under and over 35 years. Factors associated with infection from high-risk (HR) HPV were investigated.

Single spike mutation differentiating XBB.1 and XBB.1.5 enhances SARS-CoV-2 cell-to-cell transmission and facilitates serum-mediated enhancement.

Introduction: The ongoing emergence of SARS-CoV-2 variants poses significant challenges to existing therapeutics. The spike (S) glycoprotein is central to both viral entry and cell-to-cell transmission via syncytia formation, a process that confers resistance to neutralizing antibodies. The mechanisms underlying this resistance, particularly in relation to spike-mediated fusion, remain poorly understood.

The Effectiveness of COVID-19 Vaccines During the Pre-Omicron and Omicron Periods: A Retrospective Test-Negative Case-Control Study.

Background: The aim of this study was to estimate the effectiveness of original and bivalent COVID-19 vaccines in reducing COVID-19-associated hospitalizations among the adult population of Turin, Italy.

Methods: We conducted a retrospective, test-negative, case-control study of 5768 adults aged ≥50 years who had symptoms that were consistent with COVID-19-like illness and were admitted to the hospitals of the Turin Health Unit network from 1 January 2021 to 31 January 2023. We evaluated the effectiveness of the vaccines that at the time of the study were authorized in the European Union (original/bivalent BNT162b2; original mRNA-1273; ChAdOx1-S; Ad26.

Breakthrough Rectal Infections After Meningococcal B Vaccination: Microbiological and Clinical Features.

Background: 4CMenB appears to be effective in reducing (Ng) infections. Aims are to assess factors associated with breakthrough rectal Ng after 4CMenB and evaluate clinical and microbiological characteristics of breakthrough infections compared with before vaccination.

Methods: This was a retrospective study of gay, bisexual, and other men who have sex with men (GBMSM) vaccinated with 4CMenB (2 doses) between 2017 and 2023 at the San Raffaele Scientific Institute for Research, Hospitalization and Healthcare (IRCCS San Raffaele Scientific Institute), Milan, Italy, and tested for rectal Ng.