Search

The Hopfield model is a paradigmatic model of neural networks that has been analyzed for many decades in the statistical physics, neuroscience, and machine learning communities. Inspired by the manifold hypothesis in machine learning, we propose and investigate a generalization of the standard setting that we name random-features Hopfield model. Here, P binary patterns of length N are generated by applying to Gaussian vectors sampled in a latent space of dimension D a random projection followed by a nonlinearity. Using the replica method from statistical physics, we derive the phase diagram of the model in the limit P,N,D→∞ with fixed ratios α=P/N and α_{D}=D/N. Besides the usual retrieval phase, where the patterns can be dynamically recovered from some initial corruption, we uncover a new phase where the features characterizing the projection can be recovered instead. We call this phenomena the learning phase transition, as the features are not explicitly given to the model but rather are inferred from the patterns in an unsupervised fashion.

Empirical studies on the landscape of neural networks have shown that low-energy configurations are often found in complex connected structures, where zero-energy paths between pairs of distant solutions can be constructed. Here, we consider the spherical negative perceptron, a prototypical nonconvex neural network model framed as a continuous constraint satisfaction problem. We introduce a general analytical method for computing energy barriers in the simplex with vertex configurations sampled from the equilibrium. We find that in the overparametrized regime the solution manifold displays simple connectivity properties. There exists a large geodesically convex component that is attractive for a wide range of optimization dynamics. Inside this region we identify a subset of atypical high-margin solutions that are geodesically connected with most other solutions, giving rise to a star-shaped geometry. We analytically characterize the organization of the connected space of solutions and show numerical evidence of a transition, at larger constraint densities, where the aforementioned simple geodesic connectivity breaks down.

Prostate cancer (PC) is the most frequently diagnosed cancer among adult men, and its incidence is increasing worldwide [...].

We study the binary and continuous negative-margin perceptrons as simple nonconvex neural network models learning random rules and associations. We analyze the geometry of the landscape of solutions in both models and find important similarities and differences. Both models exhibit subdominant minimizers which are extremely flat and wide. These minimizers coexist with a background of dominant solutions which are composed by an exponential number of algorithmically inaccessible small clusters for the binary case (the frozen 1-RSB phase) or a hierarchical structure of clusters of different sizes for the spherical case (the full RSB phase). In both cases, when a certain threshold in constraint density is crossed, the local entropy of the wide flat minima becomes nonmonotonic, indicating a breakup of the space of robust solutions into disconnected components. This has a strong impact on the behavior of algorithms in binary models, which cannot access the remaining isolated clusters. For the spherical case the behavior is different, since even beyond the disappearance of the wide flat minima the remaining solutions are shown to always be surrounded by a large number of other solutions at any distance, up to capacity. Indeed, we exhibit numerical evidence that algorithms seem to find solutions up to the SAT/UNSAT transition, that we compute here using an 1RSB approximation. For both models, the generalization performance as a learning device is shown to be greatly improved by the existence of wide flat minimizers even when trained in the highly underconstrained regime of very negative margins.

Millions of liters of diluted bitumen (dilbit), a crude oil product from Canada's oil sands region, is transported through critical Pacific salmon habitat each day. While the toxicity of the water-soluble fraction of dilbit (WSFd) to early life-stages of salmon is known, quantitative data on life-stage differences in sensitivity to WSFd is missing. To fill this knowledge gap, we exposed two juvenile life-stages of coho salmon (O. kisutch) in parallel to very low (parts per billion), environmentally-relevant concentrations of WSFd for acute (48 h) and sub-chronic (4 wk) durations. The relative sensitivities of the two life-stages (fry and parr) were assessed by comparing the timing and magnitude of biological responses using common organismal and molecular endpoints of crude oil exposure. A significant reduction in body condition occurred in both fry and parr after 4 wk exposure to WSFd. Both life-stages also experienced a concentration-dependent decrease in time-to-loss-of-equilibrium during a hypoxia challenge test at both 48 h and 4 wk of exposure. Although organismal responses were similar, molecular responses were distinct between life-stages. In general, unexposed fry had higher baseline values of hepatic phase I biotransformation indicators than unexposed parr, but induction of EROD activity and cyp1a mRNA expression in response to WSFd exposure was greater in parr than in fry. Neither gst nor hsp70 mRNA expression, markers of phase II biotransformation and cell stress, respectively, were reliably altered by WSFd exposure in either life-stage. Taken together, results of this study do not support differential sensitivities of coho fry and parr to WSFd. All the same, the potential for ontogenic differences in the expression and induction of phase I biotransformation need to be considered because age does matter for these endpoints if they are used as bioindicators of exposure in post-spill impact assessments., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2022. Published by Elsevier B.V.)

PI-RADS 3 prostate lesions clinical management is still debated, with high variability among different centers. Identifying clinically significant tumors among PI-RADS 3 is crucial. Radiomics applied to multiparametric MR (mpMR) seems promising. Nevertheless, reproducibility assessment by external validation is required. We retrospectively included all patients with at least one PI-RADS 3 lesion (PI-RADS v2.1) detected on a 3T prostate MRI scan at our Institution (June 2016-March 2021). An MRI-targeted biopsy was used as ground truth. We assessed reproducible mpMRI radiomic features found in the literature. Then, we proposed a new model combining PSA density and two radiomic features (texture regularity (T2) and size zone heterogeneity (ADC)). All models were trained/assessed through 100-repetitions 5-fold cross-validation. Eighty patients were included (26 with GS ≥ 7). In total, 9/20 T2 features (Hector's model) and 1 T2 feature (Jin's model) significantly correlated to biopsy on our dataset. PSA density alone predicted clinically significant tumors (sensitivity: 66%; specificity: 71%). Our model obtained a sensitivity of 80% and a specificity of 76%. Standard-compliant works with detailed methodologies achieve comparable radiomic feature sets. Therefore, efforts to facilitate reproducibility are needed, while complex models and imaging protocols seem not, since our model combining PSA density and two radiomic features from routinely performed sequences appeared to differentiate clinically significant cancers.

Current deep neural networks are highly overparameterized (up to billions of connection weights) and nonlinear. Yet they can fit data almost perfectly through variants of gradient descent algorithms and achieve unexpected levels of prediction accuracy without overfitting. These are formidable results that defy predictions of statistical learning and pose conceptual challenges for nonconvex optimization. In this paper, we use methods from statistical physics of disordered systems to analytically study the computational fallout of overparameterization in nonconvex binary neural network models, trained on data generated from a structurally simpler but "hidden" network. As the number of connection weights increases, we follow the changes of the geometrical structure of different minima of the error loss function and relate them to learning and generalization performance. A first transition happens at the so-called interpolation point, when solutions begin to exist (perfect fitting becomes possible). This transition reflects the properties of typical solutions, which however are in sharp minima and hard to sample. After a gap, a second transition occurs, with the discontinuous appearance of a different kind of "atypical" structures: wide regions of the weight space that are particularly solution dense and have good generalization properties. The two kinds of solutions coexist, with the typical ones being exponentially more numerous, but empirically we find that efficient algorithms sample the atypical, rare ones. This suggests that the atypical phase transition is the relevant one for learning. The results of numerical tests with realistic networks on observables suggested by the theory are consistent with this scenario.

The success of deep learning has revealed the application potential of neural networks across the sciences and opened up fundamental theoretical problems. In particular, the fact that learning algorithms based on simple variants of gradient methods are able to find near-optimal minima of highly nonconvex loss functions is an unexpected feature of neural networks. Moreover, such algorithms are able to fit the data even in the presence of noise, and yet they have excellent predictive capabilities. Several empirical results have shown a reproducible correlation between the so-called flatness of the minima achieved by the algorithms and the generalization performance. At the same time, statistical physics results have shown that in nonconvex networks a multitude of narrow minima may coexist with a much smaller number of wide flat minima, which generalize well. Here, we show that wide flat minima arise as complex extensive structures, from the coalescence of minima around "high-margin" (i.e., locally robust) configurations. Despite being exponentially rare compared to zero-margin ones, high-margin minima tend to concentrate in particular regions. These minima are in turn surrounded by other solutions of smaller and smaller margin, leading to dense regions of solutions over long distances. Our analysis also provides an alternative analytical method for estimating when flat minima appear and when algorithms begin to find solutions, as the number of model parameters varies.

Background: In free-ranging reptile populations, bacterial, fungal, viral and parasitic pathogens may affect hosts through impairment in movements, thermoregulation, reproduction, survival, and population dynamics. The speckled dwarf tortoise (Chersobius [Homopus] signatus) is a threatened species that is mostly restricted to the Succulent Karoo biome in South Africa, and little information on pathogens of this species is available yet. We derived baseline parameters for five males and five females that were captured to genetically enhance a conservation breeding program in Europe. Upon collection of the tortoises, ticks were removed and identified. Immediately upon arrival in Europe, ocular, nasal, oral and cloacal swabs were taken for viral, bacteriological and mycological examinations. Fecal samples were collected before and 1 month after fenbendazole treatment, and analyzed for parasites. A panel of PCR, aiming to detect herpesviruses, adenoviruses and iridoviruses, was carried out., Results: Samples were negative for viruses, while bacteriological examination yielded detectable growth in 82.5% of the swabs with a mean load of 16 × 10 7  ± 61 × 10 8 colony forming units (CFU) per swab, representing 34 bacterial species. Cloacal and oral swabs yielded higher detectable growth loads than nasal and ocular swabs, but no differences between sexes were observed. Fungi and yeasts (mean load 5 × 10 3  ± 13 × 10 3 CFU/swab) were detected in 25% of the swabs. All pre-treatment fecal samples were positive for oxyurid eggs, ranging from 200 to 2400 eggs per gram of feces, whereas after the treatment a significantly reduced egg count (90-100% reduction) was found in seven out of 10 individuals. One remaining individual showed 29% reduction, and two others had increased egg counts. In five tortoises, Nycthocterus spp. and coccidian oocysts were also identified. Soft ticks were identified as Ornithodoros savignyi., Conclusions: Our baseline data from clinically healthy individuals will help future studies to interpret prevalences of microorganisms in speckled dwarf tortoise populations. The study population did not appear immediately threatened by current parasite presence.

Introduction: The objective of this study was to test Prostate Imaging Reporting and Data System (PI-RADS) classification on multiparametric magnetic resonance imaging (mpMRI) and MRI-derived prostate-specific antigen density (PSAD) in predicting the risk of reclassification in men in active surveillance (AS), who underwent confirmatory or per-protocol follow-up biopsy., Materials and Methods: Three hundred eighty-nine patients in AS underwent mpMRI before confirmatory or follow-up biopsy. Patients with negative (-) mpMRI underwent systematic random biopsy. Patients with positive (+) mpMRI underwent targeted fusion prostate biopsies + systematic random biopsies. Different PSAD cutoff values were tested (< 0.10, 0.10-0.20, ≥ 0.20). Multivariable analyses assessed the risk of reclassification, defined as clinically significant prostate cancer of grade group 2 or more, during follow-up according to PSAD, after adjusting for covariates., Results: One hundred twenty-seven (32.6%) patients had mpMRI(-); 72 (18.5%) had PI-RADS 3, 150 (38.6%) PI-RADS 4, and 40 (10.3%) PI-RADS 5 lesions. The rate of reclassification to grade group 2 PCa was 16%, 22%, 31%, and 39% for mpMRI(-) and PI-RADS 3, 4, and 5, respectively, in case of PSAD < 0.10 ng/mL 2 ; 16%, 25%, 36%, and 44%, in case of PSAD 0.10 to 0.19 ng/mL 2 ; and 25%, 42%, 55%, and 67% in case of PSAD ≥ 0.20 ng/mL 2 . PSAD ≥ 0.20 ng/mL 2 (odds ratio [OR], 2.45; P = .007), PI-RADS 3 (OR, 2.47; P = .013), PI-RADS 4 (OR, 2.94; P < .001), and PI-RADS 5 (OR, 3.41; P = .004) were associated with a higher risk of reclassification., Conclusion: PSAD ≥ 0.20 ng/mL 2 may improve predictive accuracy of mpMRI results for reclassification of patients in AS, whereas PSAD < 0.10 ng/mL 2 may help selection of patients at lower risk of harboring clinically significant prostate cancer. However, the risk of reclassification is not negligible at any PSAD cutoff value, also in the case of mpMRI(-)., (Copyright © 2020 Elsevier Inc. All rights reserved.)

Introduction & Objectives: We tested the role of multiparametric magnetic resonance imaging (mpMRI) in disease reclassification and whether the combination of mpMRI and clinicopathological variables could represent the most accurate approach to predict the risk of reclassification during active surveillance., Materials & Methods: Three-hundred eighty-nine patients (pts) underwent mpMRI and subsequent confirmatory or follow-up biopsy according to the Prostate Cancer Research International Active Surveillance (PRIAS) protocol. Pts with negative (-) mpMRI underwent systematic random biopsy. Pts with positive (+) mpMRI [Prostate Imaging Reporting and Data System, version 2 (PI-RADS-V2) score ≥3] underwent targeted + systematic random biopsies. Multivariate analyses were used to create three models predicting the probability of reclassification [International Society of Urological Pathology ≥ Grade Group 2 (GG2)]: a basic model including only clinical variables (age, prostate-specific antigen density, and number of positive cores at baseline), an Magnetic resonance imaging (MRI) model including only the PI-RADS score, and a full model including both the previous ones. The predictive accuracy (PA) of each model was quantified using the area under the curve., Results: mpMRI negative (-) was recorded in 127 (32.6%) pts; mpMRI positive (+) was recorded in 262 pts: 72 (18.5%) had PI-RADS 3, 150 (38.6%) PI-RADS 4, and 40 (10.3%) PI-RADS 5 lesions. At a median follow-up of 12 months, 125 pts (32%) were reclassified to GG2 prostate cancer. The rate of reclassification to GG2 prostate cancer was 17%, 35%, 38%, and 52% for mpMRI (-), PI-RADS 3, 4, and 5, respectively ( P  < 0.001). The PA was 69% and 64% in the basic and MRI models, respectively. The full model had the best PA of 74%: older age ( P  = 0.023; Odds ratio (OR) = 1.040), prostate-specific antigen density ( P  = 0.037; OR = 1.324), number of positive cores at baseline ( P  = 0.001; OR = 1.441), and PI-RADS 3, 4, and 5 (overall P  = 0.001; OR = 2.458, 3.007, and 3.898, respectively) were independent predictors of reclassification., Conclusions: Disease reclassification increased according to the PI-RADS score increase, at confirmatory or follow-up biopsy. However, a no-negligible rate of reclassification was found also in cases of mpMRI (-). The combination of mpMRI and clinicopathological variables still represents the most accurate approach to pts on active surveillance., Competing Interests: All authors have no conflict of interest to declare., (© 2020 Asian Pacific Prostate Society. Published by Elsevier B.V.)

Objectives: To evaluate the diagnostic accuracy of multiparametric magnetic resonance imaging in the detection of prostate cancer, according to Prostate Imaging Reporting and Data System, and the usefulness of combining clinical parameters to improve patients' risk assessment., Methods: Overall, 201 patients underwent multiparametric magnetic resonance imaging investigation with a 3-T magnet and a 32-channel body coil based on triplanar high-resolution T2-weighted, diffusion-weighted and T1-weighted dynamic contrast-enhanced imaging before, during and after intravenous administration of paramagnetic contrast agent. Random transrectal ultrasound-guided biopsy was carried out for all eligible patients. If a Prostate Imaging Reporting and Data System ≥3 lesion was present, a targeted biopsy with magnetic resonance imaging-transrectal ultrasound fusion-guided system was carried out., Results: Sensitivity, specificity, positive predictive value and negative predictive value of Prostate Imaging Reporting and Data System ≥3 lesions for the detection of prostate cancer were 65.1%, 54.9%, 43.1% and 75.0% respectively, with an accuracy of 64.2% (55.1-72.7%). At uni- and multivariate analysis, age ≥70 years and prostate-specific antigen density ≥0.15 ng/mL/mL were significantly associated with prostate cancer. A new risk model named "modified Prostate Imaging Reporting and Data System" was created considering age and prostate-specific antigen density in addition to the Prostate Imaging Reporting and Data System score showing an improved correlation with prostate cancer compared with the Prostate Imaging Reporting and Data System alone (area under curve 71.4%, 95% confidence interval 62.2-80.5 vs area under curve 62.6%, 95% confidence interval 52.1-73; P ≤ 0.0001)., Conclusions: The accuracy of Prostate Imaging Reporting and Data System alone in the diagnosis of prostate cancer might be suboptimal, whereas a novel risk model based on the combination of multiparametric magnetic resonance imaging data with clinical parameters could offer higher discrimination and improve the ability of diagnosing clinically significant disease., (© 2020 The Japanese Urological Association.)

Colibacillosis is the most frequent bacterial disease in avian species and antimicrobials are the main weapon to reduce incidence and mortality associated to it. However, indiscriminate use of antibiotics may lead to therapy failure and economic losses for the breeder. The aims of this study were to, determine the antibiotic resistance of Escherichia coli isolates, evaluate the correlation between E. coli isolation and systems of breeding included in this study, and identify the avian pathogenic E.coli (APEC) amongst the E. coli strains isolated. A total of 51 E. coli strains were subjected to antimicrobial susceptibility test and they were screened for the presence of virulence genes through PCR. Resistance was most frequently detected against ampicillin and nalidixic acid meanwhile E. coli isolates showed less resistance to the cephalosporins. Overall, 40% of the isolates showed resistance to at least three or more antimicrobials and 16/51 isolates were defined APEC strains. The virulence genes iucD, cvi/cva, irp2 and iss were detected from all 16 APEC strains. The virulence genes tsh, vat, papC, and astA were detected from 11, 7, 5 and 3 APEC strains, respectively. Results demonstrated the importance of studies on APEC and antibiotic resistance genes in Italy, and it was shown that the systems of breeding might influence the antibiotic resistanc.

Objective: This study aimed at estimating the patient effective dose during whole-body low-dose multi detector computed tomography with a scanner Philips Brillance 64, and to compare it with those reported in literature for the same procedure and with the dose of the conventional X-ray examination in our institution., Materials and Methods: WBLDMDCT effective dose was evaluated for 29 MM patients, using Dose Length Product values. Conventional X-ray examination dose was estimated using a Rando Phantom and Dose Area Product indexes. ICRP Publication 103 based coefficients were used., Results: Mean WBLDMDCT effective dose values-estimated using sex-specific conversion coefficients and body weight factors-were 3.6 and 2.8 mSv for females and males, respectively. The effective dose for the conventional method was 1.2 mSv for Rando phantom., Conclusion: The WBLDMDCT effective dose in our institution is consistent with the values reported in previous studies. Such a dose is about 2.5- to 3-fold higher than the mean radiation dose of the conventional X-ray study. Nevertheless, considering the improved diagnostic accuracy of the CT investigation, the comfort of the patient and the old age of the MM population, dose/quality ratio can be considered favourable.