Abnormal Input Research Articles

Frass deposition from pest outbreaks affects soil organic carbon and its relationship with environmental factors in a deciduous broad-leaved forest

Forest defoliators are one of the major biological disturbances to forest ecosystems. As one of the abnormal nutrient input paths into forest ecosystems, frass deposition from the pest outbreak plays a critical role in regulating soil organic carbon (SOC) in forest ecosystems. However, how frass deposition affects SOC and its fractions in forests remains unclear. Based on a severe outbreak of defoliator in an oak-sweetgum mixed forest in Jigong Mountain in 2014, we compared the difference in SOC between plots with and without frass deposition for 4 consecutive years. The results showed that frass deposition led to a significant increase of 25.1 % in soil microbial biomass C (MBC) and 32.0 % in dissolved organic C (DOC) in 2014, which further escalated to 50.4 % and 50.6 % in the subsequent year (2015), respectively. The response of SOC to frass deposition lagged behind MBC and DOC. Specifically, there was no change in SOC in 2014, but a significant increase (50.9 %) was observed in the subsequent 2–3 years. The positive dependences of MBC and DOC upon fine root biomass were negated under frass deposition, while the relationship between SOC and fine root biomass remained unaffected. Soil organic carbon and DOC showed non-linear responses to frass amount and the changed soil nitrogen content. Our finding that the response of SOC to frass deposition lagged behind soil labile C indicates that SOC exhibits a certain resilience towards forest disturbance. The findings also imply that investigating the long-term impacts of frass deposition on SOC in forests would contribute to the scientific assessment of forest C cycling under disturbance.

A new robust dynamic state estimation approach for power systems with non-Gaussian noise

The Gaussian noise distribution is typically used in dynamic state estimation (DSE) but it is not always true in practice because of abnormal system inputs, impulsive noise and measurement outliers. In this paper, a new robust DSE approach based on a new robust Lp norm based estimator and the cubature Kalman filter (CKF) is developed for power systems with non-Gaussian noise statistics. The Lp norm based estimator is derived from the Lp norm formula and the quadratic formula in order to alleviate the impacts from bad data and outliers. The proposed Lp-CKF DSE approach exhibits good accuracy because a new estimation error covariance is obtained by using the influence function. The robustness of the proposed Lp-CKF DSE approach is verified by performing simulations on a generator in the IEEE 39-bus system.

Corticostriatal causality analysis in children and adolescents with attention-deficit/hyperactivity disorder.

The effective connectivity between the striatum and cerebral cortex has not been fully investigated in attention-deficit/hyperactivity disorder (ADHD). Our objective was to explore the interaction effects between diagnosis and age on disrupted corticostriatal effective connectivity and to represent the modulation function of altered connectivity pathways in children and adolescents with ADHD. We performed Granger causality analysis on 300 participants from a publicly available Attention-Deficit/Hyperactivity Disorder-200 dataset. By computing the correlation coefficients between causal connections between striatal subregions and other cortical regions, we estimated the striatal inflow and outflow connection to represent intermodulation mechanisms in corticostriatal pathways. Interactions between diagnosis and age were detected in the superior occipital gyrus within the visual network, medial prefrontal cortex, posterior cingulate gyrus, and inferior parietal lobule within the default mode network, which is positively correlated with hyperactivity/impulsivity severity in ADHD. Main effect of diagnosis exhibited a general higher cortico-striatal causal connectivity involving default mode network, frontoparietal network and somatomotor network in ADHD compared with comparisons. Results from high-order effective connectivity exhibited a disrupted information pathway involving the default mode-striatum-somatomotor-striatum-frontoparietal networks in ADHD. The interactions detected in the visual-striatum-default mode networks pathway appears to be related to the potential distraction caused by long-term abnormal information input from the retina in ADHD. Higher causal connectivity and weakened intermodulation may indicate the pathophysiological process that distractions lead to the impairment of motion planning function and the inhibition/control of this unplanned motion signals in ADHD.

Stimulated cervical afferent input increases postural instability in older people with chronic neck pain: a cross-sectional study

BackgroundSeveral potential causes can impair balance in older people. The neck torsion maneuver may be useful in demonstrating impaired balance caused by the stimulation of cervical proprioceptive input. Whereas evidence suggests impaired standing balance in older people with chronic neck pain, balance impairment during the neck torsion position and its relationship with clinical characteristics have not yet been investigated in this population. The aims of this study were to investigate whether the neck torsion position could significantly influence balance responses in older people with chronic non-specific neck pain and to determine the relationships between the balance responses and characteristics of neck pain.MethodsSixty-eight older people (34 with chronic non-specific neck pain and 34 controls) participated in the study. Balance was tested using a force plate during comfortable stance with eyes open under four conditions: neutral head on a firm surface, neutral head on a soft surface, neck torsion to left and right on a firm surface and neck torsion to left and right on a soft surface. Balance outcomes were anterior-posterior (AP) and medial-lateral (ML) displacements, sway area and velocity. Characteristics of neck pain were intensity, duration and disability.ResultsOverall, the neck pain group exhibited greater AP and ML displacements, sway area and velocity in the neck torsion position on firm and soft surfaces compared to controls (partial eta squared (η²p) = 0.06–0.15, p < 0.05). The neck pain group also had greater AP displacement, sway area and velocity in the neutral position on a soft surface compared to controls (η²p = 0.09–0.16, p < 0.05). For both groups, the neck torsion position displayed overall greater postural sway compared to the neutral position (η²p = 0.16–0.69, p < 0.05). There were no relationships between the postural sway outcomes and characteristics of neck pain (p > 0.05).ConclusionThe neck torsion maneuver, stimulating the receptors resulted in increased postural sway in older people, with a more pronounced effect in those with neck pain. The study provides evidence supporting the use of neck torsion for assessing impaired balance related to abnormal cervical input in older people with chronic non-specific neck pain.

Unraveling the Twist: Spatial Navigational Challenges in Cervical Dystonia.

Cervical dystonia (CD) is an intricate neurological condition with motor and nonmotor symptoms. These include disruptions in visual perception, self-orientation, visual working memory, and vestibular functions. However, the specific impact of CD on perceiving self-motion direction, especially with isolated visual or vestibular stimuli, remains largely unexplored. This study aimed to examine the effects of CD on linear motion perception, hypothesizing impaired heading discrimination in both vestibular and visual tasks, and that such deficits correlate with the disease severity. We employed a cutting-edge motion platform to precisely control whole-body linear motion. Through repeated two-alternative forced-choice tasks, we assessed vestibular heading direction discrimination. Participants observed dynamic star clouds in immersive virtual reality and indicated their perceived self-motion direction, evaluating visual heading direction discrimination. Sensitivity to direction variations and response accuracy errors were analyzed using robust Gaussian cumulative distribution psychometric functions. Heading perception is impaired in individuals with CD, particularly evident in vestibular heading discrimination. CD severity correlated with elevated thresholds for both vestibular and visual heading discrimination. Surprisingly, lateralized CD did not introduce bias in either system, suggesting widespread disruption over localized effects. Contrary to previous beliefs, our findings challenge the idea that CD-related heading discrimination issues mainly arise from peripheral vestibular effects. Instead, abnormal proprioceptive input from dystonic neck muscles introduces noise into the central mechanism integrating visual, vestibular, and proprioceptive signals. These insights into spatial navigation deficits have implications for future CD research. © 2023 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society. This article has been contributed to by U.S. Government employees and their work is in the public domain in the USA.

PatchCensor: Patch Robustness Certification for Transformers via Exhaustive Testing

In the past few years, Transformer has been widely adopted in many domains and applications because of its impressive performance. Vision Transformer (ViT), a successful and well-known variant, attracts considerable attention from both industry and academia thanks to its record-breaking performance in various vision tasks. However, ViT is also highly nonlinear like other classical neural networks and could be easily fooled by both natural and adversarial perturbations. This limitation could pose a threat to the deployment of ViT in the real industrial environment, especially in safety-critical scenarios. How to improve the robustness of ViT is thus an urgent issue that needs to be addressed. Among all kinds of robustness, patch robustness is defined as giving a reliable output when a random patch in the input domain is perturbed. The perturbation could be natural corruption, such as part of the camera lens being blurred. It could also be a distribution shift, such as an object that does not exist in the training data suddenly appearing in the camera. And in the worst case, there could be a malicious adversarial patch attack that aims to fool the prediction of a machine learning model by arbitrarily modifying pixels within a restricted region of an input image. This kind of attack is also called physical attack, as it is believed to be more real than digital attack. Although there has been some work on patch robustness improvement of Convolutional Neural Network, related studies on its counterpart ViT are still at an early stage as ViT is usually much more complex with far more parameters. It is harder to assess and improve its robustness, not to mention to provide a provable guarantee. In this work, we propose PatchCensor, aiming to certify the patch robustness of ViT by applying exhaustive testing. We try to provide a provable guarantee by considering the worst patch attack scenarios. Unlike empirical defenses against adversarial patches that may be adaptively breached, certified robust approaches can provide a certified accuracy against arbitrary attacks under certain conditions. However, existing robustness certifications are mostly based on robust training, which often requires substantial training efforts and the sacrifice of model performance on normal samples. To bridge the gap, PatchCensor seeks to improve the robustness of the whole system by detecting abnormal inputs instead of training a robust model and asking it to give reliable results for every input, which may inevitably compromise accuracy. Specifically, each input is tested by voting over multiple inferences with different mutated attention masks, where at least one inference is guaranteed to exclude the abnormal patch. This can be seen as complete-coverage testing, which could provide a statistical guarantee on inference at the test time. Our comprehensive evaluation demonstrates that PatchCensor is able to achieve high certified accuracy (e.g., 67.1% on ImageNet for 2%-pixel adversarial patches), significantly outperforming state-of-the-art techniques while achieving similar clean accuracy (81.8% on ImageNet). The clean accuracy is the same as vanilla ViT models. Meanwhile, our technique also supports flexible configurations to handle different adversarial patch sizes by simply changing the masking strategy.

Motor unit firing patterns in older adults with low skeletal muscle mass

Muscular dysfunctions involving a decline in muscle strength are often induced by loss of muscle mass in older adults. Understanding neural activation in older adults in addition to muscular characteristics may be important to prevent such age-related dysfunctions. This study aimed to investigate the difference in motor unit firing patterns between community-dwelling older individuals with normal and low skeletal muscle mass. Sixty-six older adults (62–90 years) performed muscle strength and function tests. On conducting high-density surface electromyography of the vastus lateralis, individual motor unit firing properties were assessed. Individual motor units were divided into three different recruitment threshold groups and their firing rates were compared. The skeletal muscle quantity and quality were assessed using bioimpedance methods and ultrasound images. They were divided into two groups according to sarcopenia criteria: a normal group (n = 39) and presarcopenia group with low skeletal muscle mass but normal physical functions (n = 21). Skeletal muscle mass and muscle thickness were greater and echo intensity was lower in the normal group than presarcopenia group. Motor units in normal older adults fired at different rates with a hierarchy depending on their recruitment threshold, observed as a normal phenomenon. However, motor units in the presarcopenia group fired without showing the hierarchical pattern. The results suggest that older adults with low skeletal muscle mass exhibited an abnormal neural input pattern, in addition to declines in muscle quantity and quality.

Dynamic state estimation of power systems considering maximum correlation entropy and quadratic function

Uncertainties such as abnormal system inputs, strong model nonlinearities, outliers and impulsive noise unavoidably exist in the power system dynamic state estimation (SE) (DSE). The existing Kalman filter-based DSE methods cannot deal with the uncertainties well and can be improved further. In this paper, a robust DSE method based on the maximum correlation entropy (MCE), quadratic function (QF) and the cubature Kalman filter is proposed to reduce the effects caused by the uncertainties. The MCE-QF based estimator is robust and can alleviate the influence of uncertainties. A new SE error covariance is derived by using the robust statistical tool, influence function. The robustness and estimation precision of the proposed DSE method is demonstrated by conducting simulations on a synchronous generator under different cases.

The effect of single-cell knockout of Fragile X Messenger Ribonucleoprotein on synaptic structural plasticity.

Fragile X Syndrome (FXS) is the best-known form of inherited intellectual disability caused by the loss-of-function mutation in a single gene. The FMR1 gene mutation abolishes the expression of Fragile X Messenger Ribonucleoprotein (FMRP), which regulates the expression of many synaptic proteins. Cortical pyramidal neurons in postmortem FXS patient brains show abnormally high density and immature morphology of dendritic spines; this phenotype is replicated in the Fmr1 knockout (KO) mouse. While FMRP is well-positioned in the dendrite to regulate synaptic plasticity, intriguing in vitro and in vivo data show that wild type neurons embedded in a network of Fmr1 KO neurons or glia exhibit spine abnormalities just as neurons in Fmr1 global KO mice. This raises the question: does FMRP regulate synaptic morphology and dynamics in a cell-autonomous manner, or do the synaptic phenotypes arise from abnormal pre-synaptic inputs? To address this question, we combined viral and mouse genetic approaches to delete FMRP from a very sparse subset of cortical layer 5 pyramidal neurons (L5 PyrNs) either during early postnatal development or in adulthood. We then followed the structural dynamics of dendritic spines on these Fmr1 KO neurons by in vivo two-photon microscopy. We found that, while L5 PyrNs in adult Fmr1 global KO mice have abnormally high density of thin spines, single-cell Fmr1 KO in adulthood does not affect spine density, morphology, or dynamics. On the contrary, neurons with neonatal FMRP deletion have normal spine density but elevated spine formation at 1 month of age, replicating the phenotype in Fmr1 global KO mice. Interestingly, these neurons exhibit elevated thin spine density, but normal total spine density, by adulthood. Together, our data reveal cell-autonomous FMRP regulation of cortical synaptic dynamics during adolescence, but spine defects in adulthood also implicate non-cell-autonomous factors.

Neural Network, Nonlinear-Fitting, Sliding Mode, Event-Triggered Control under Abnormal Input for Port Artificial Intelligence Transportation Robots

A new control algorithm was designed to solve the problems of actuator physical failure, remote network attack, and sudden change in trajectory curvature when a port’s artificial intelligence-based transportation robots track transportation in a freight yard. First of all, the nonlinear, redundant, saturated sliding surface was designed based on the redundant information of sliding mode control caused by the finite nature of control performance; the dynamic acceleration characteristic of super-twisted sliding mode reaching law was considered to optimize the control high frequency change caused by trajectory mutation; and an improved super-twist reaching law was designed. Then, a nonlinear factor was designed to construct a nonlinear, fault-tolerant filtering mechanism to compensate for the abnormal part of the unknown input that cannot be executed by adaptive neural network reconstruction. On this basis, the finite-time technology and parameter-event-triggered mechanism were combined to reduce the dependence on communication resources. As a result, the design underwent simulation verification to verify its effectiveness and superiority. In the comparative simulation, under a consistent probability of a network attack, the tracking accuracy of the algorithm proposed in this paper was 22.65%, 12.69% and 11.48% higher those that of the traditional algorithms.

DUMA: Dual Mask for Multivariate Time Series Anomaly Detection

As a major category of unsupervised anomaly detection methods for multivariate time series, autoregression-based methods train a predictor to model the normal pattern from only normal time series, and then detect anomalies by prediction error. However, we find that the discrepancy of input data between the training and inference stages is a crucial challenge, which may lead to volatile results when detectors take abnormal time series as input. Furthermore, the correlations among multiple sensors are intricate, where irrelevant sensors may bring noise dependencies. This article proposes an autoregression-based time series anomaly detection method named DUal Masked self-Attention (DUMA). First, we propose a block-mask mechanism to enhance the robustness of the predictor for abnormal input data. Then a max-mask self-attention is proposed to reduce the noise dependencies between irrelevant sensors. Experiments on three cyber-physical systems (CPSs) datasets show that DUMA outperforms the state-of-the-art baseline methods.

Neritic larval fish assemblages across the Texas shelf in the northwestern Gulf of Mexico following Hurricane Harvey

AbstractFall is an important time of year for fish recruitment in the Gulf of Mexico, with many commercially and recreationally important species spawning during this period. Changes in temperature and salinity regimes through seasonally abnormal freshwater input may change spawning patterns and displace larvae. Such an event occurred off the coast of Galveston Bay, Texas, when a freshwater flood plume formed from unprecedented rainfall related to Hurricane Harvey at the end of August 2017. This study investigated the effects of a large‐scale flood plume on larval fish assemblages, collected at nearshore and shelf zones 1 and 2 months after Hurricane Harvey. Samples from 2017 were compared to historical datasets collected by NOAA Fisheries in September (2000–2004 and 2006–2016) and October (2000–2012). Evidence of lowered water temperature and salinity was found in both September and October 2017. Larval community composition changed along cross‐shelf gradients and with increasing distance from shore. We also found evidence of 2017 assemblages differing from historical data, with the September 2017 assemblage more closely resembling those from October historical samples. Observed differences in 2017 assemblages were most likely explained by an earlier onset in fall sciaenid spawning, suggesting that decreases in water temperature occurring during hurricanes have the potential to alter fish spawning patterns.

Freeze-and-mutate: abnormal sample identification for DL applications through model core analysis

Deep learning (DL) applications, representing an emerging form of new software, are gaining increasing popularity by their intelligent and adaptive services. However, their service reliability depends highly on the prediction accuracy of their internally-integrated DL models. In practice, DL models are often observed to suffer from ill predictions upon abnormal inputs (e.g., adversarial attacking samples, out-of-distribution (OOD) samples, and etc.), and this could easily lead to unexpected behaviors or even catastrophic consequences (e.g., system crash). One promising way to guard the application reliability is to reveal such abnormal inputs in time before they are fed to the DL models integrated in the concerned applications. Then remedy actions (e.g., discarding or fixing these inputs) can be done to protect applications from acting abnormally. Existing work addressed this revealing problem by either making sample distance-comparison based analysis or generating sufficient model mutants for comparative analysis. However, such treatments caused a restricted focus on samples only, while overlooking the DL models themselves, or had to analyze massive mutants, incurring non-negligible overheads to applications. In this article, we propose a novel approach, NetChopper, to conducting a core analysis on the target DL model, and then partitioning it into two parts, one associating closely with the training knowledge being the model core (expected to be important and thus stable), and the other being the remaining part (expected to be immaterial and thus changeable). Based on such partitioning, NetChopper proceeds to preserve (or freeze) the model core, but mutate the remaining part to produce only a small number of model mutants. Later, NetChopper becomes able to reveal abnormal inputs from normal ones by exploiting these model-relevant and light-weight mutants only. We experimentally evaluated NetChopper by widely-used DL subjects (e.g., MNIST+LeNet4, and CIFAR10+VGG16) and typical abnormal inputs (e.g., adversarial and OOD samples). The results reported NetChopper ’s promising AUROC scores in revealing the abnormal degrees of inputs, generally and stably outperforming, or comparably effective as, state-of-the-art techniques (e.g., mMutant, Surprise, and Mahalanobis), and also confirmed its high effectiveness and efficiency (with only marginal online overhead).

This Month in Anesthesiology

This Month in Anesthesiology

Fault Location and Classification for Distribution Systems Based on Deep Graph Learning Methods

Accurate and timely fault diagnosis is of great significance for the safe operation and power supply reliability of distribution systems. However, traditional intelligent methods limit the use of the physical structures and data information of power networks. To this end, this study proposes a fault diagnostic model for distribution systems based on deep graph learning. This model considers the physical structure of the power network as a significant constraint during model training, which endows the model with stronger information perception to resist abnormal data input and unknown application conditions. In addition, a special spatiotemporal convolutional block is utilized to enhance the waveform feature extraction ability. This enables the proposed fault diagnostic model to be more effective in dealing with both fault waveform changes and the spatial effects of faults. In addition, a multi-task learning framework is constructed for fault location and fault type analysis, which improves the performance and generalization ability of the model. The IEEE 33-bus and IEEE 37-bus test systems are modeled to verify the effectiveness of the proposed fault diagnostic model. Finally, different fault conditions, topological changes, and interference factors are considered to evaluate the anti-interference and generalization performance of the proposed model. Experimental results demonstrate that the proposed model outperforms other state-of-the-art methods.

Suboccipital Muscles, Forward Head Posture, and Cervicogenic Dizziness.

Dizziness or vertigo can be caused by dysfunction of the vestibular or non-vestibular systems. The diagnosis, treatment, and mechanism of dizziness or vertigo caused by vestibular dysfunction have been described in detail. However, dizziness by the non-vestibular system, especially cervicogenic dizziness, is not well known. This paper explained the cervicogenic dizziness caused by abnormal sensory input with references to several studies. Among head and neck muscles, suboccipital muscles act as stabilizers and controllers of the head. Structural and functional changes of the suboccipital muscles can induce dizziness. Especially, myodural bridges and activation of trigger point stimulated by abnormal head posture may be associated with cervicogenic dizziness.

The Role of Retinal Dysfunction in Myopia Development.

Myopia is a refractive disorder arising from a mismatch between refractive power and relatively long axial length of the eye. With its dramatically increasing prevalence, myopia has become a pervasive social problem. It is commonly accepted that abnormal visual input acts as an initiating factor of myopia. As the first station to perceive visual signals, the retina plays an important role in myopia etiology. The retina is a fine-layered structure with multitudinous cells, processing intricate visual signals via numerous molecular pathways. Accordingly, dopaminergic mechanisms, contributions of rod and cone photoreceptors, myopic structural changes of retinal pigment epithelium (RPE) and neuro-retinal layers have all suggested a vital role of retinal dysfunction in myopia development. Herein, we separately discuss myopia-related retinal dysfunction and current dilemmas by different levels, from molecules to cells, with the hope that the comprehensive delineation could contribute to a better understanding of myopia etiology, indicate novel therapeutic targets, and inspire future studies.

Effect of electroacupuncture on the expressions of TRPV1, P2X3 receptors in bladder of rats with interstitial cystitis

To observe the effect of electroacupuncture (EA) at "Ciliao" (BL 32) and "Huiyang" (BL 35) on the pain, urodynamic and the expressions of transient receptor poteintial vanilloid 1 (TRPV1) and P2X3 receptors in bladder of rats with interstitial bladder (IC), and to explore the possible mechanism on EA for IC. A total of 24 Wistar female rats were randomly divided into a blank group, a model group and an EA group, 8 rats in each group. In the model group and the EA group, IC model was established by intraperitoneal injection of cyclophosphamide by 150 mg/kg at once. EA was applied at "Ciliao" (BL 32) and "Huiyang" (BL 35) in the EA group for 20 min, with continuous wave, 30 Hz in frequency, once a day for 3 consecutive days. Mechanical pain threshold of bladder and urodynamic indexes (first urination time, bladder effective volume and urination pressure) were observed after model establishment and after intervention, the expressions of TRPV1 and P2X3 receptors in the bladder were detected by Western blot. After model establishment, the mechanical pain threshold of bladder was decreased in the model group and the EA group compared with that in the blank group (P<0.01). After intervention, the mechanical pain threshold of bladder in the model group was lower than the blank group (P<0.01), and that in the EA group was higher than the model group (P<0.01). The urodynamic of the rats in the blank group was normal, obvious abnormal contraction during the filling period of bladder was found in the rats of the model group, while no abnormal contraction during the filling period was found in the rats of the EA group. After model establishment, in the model group and the EA group, the first urination time was earlier than the blank group (P<0.01), while bladder effective volume and urination pressure were lower than the blank group (P<0.01). After intervention, in the model group, the first urination time was earlier than the blank group (P<0.01), while bladder effective volume and urination pressure were lower than the blank group (P<0.05); in the EA group, the first urination time was later than the model group (P<0.05), while bladder effective volume and urination pressure were higher than the model group (P<0.05). Compared with the blank group, the protein expressions of TRPV1 and P2X3 receptors in bladder were up-regulated in the model group (P<0.01); compared with the model group, the protein expressions of TRPV1 and P2X3 receptors in bladder were down-regulated in the EA group (P<0.05). EA can relieve bladder pain and improve urodynamic in IC rats. The mechanism may be related to the down-regulation on the expressions of TRPV1 and P2X3 receptors and the further inhibition on the abnormal input of bladder signal.

Anomaly Detection in Multi-Host Environment Based on Federated Hypersphere Classifier

Detecting anomalous inputs is essential in many mission-critical systems in various domains, particularly cybersecurity. In particular, deep neural network-based anomaly detection methods have been successful for anomaly detection tasks with the recent advancements in deep learning technology. Nevertheless, the existing methods have considered somewhat idealized problems where it is enough to learn a single detector based on a single dataset. In this paper, we consider a more practical problem where multiple hosts in an organization collect their input data, while data sharing among the hosts is prohibitive due to security reasons, and only a few of them have experienced abnormal inputs. Furthermore, the data distribution of the hosts can be skewed; for example, a particular type of input can be observed by a limited subset of hosts. We propose the federated hypersphere classifier (FHC), which is a new anomaly detection method based on an improved hypersphere classifier suited for running in the federated learning framework to perform anomaly detection in such an environment. Our experiments with image and network intrusion detection datasets show that our method outperforms the state-of-the-art anomaly detection methods trained in a host-wise fashion by learning a consensus model as if we have accessed the input data from all hosts but without communicating such data.

Critical Periods in Vision Revisited.

For four decades, investigations of the biological basis of critical periods in the developing mammalian visual cortex were dominated by study of the consequences of altered early visual experience in cats and nonhuman primates. The neural deficits thus revealed also provided insight into the origin and neural basis of human amblyopia that in turn motivated additional studies of humans with abnormal early visual input. Recent human studies point to deficits arising from alterations in all visual cortical areas and even in nonvisual cortical regions. As the new human data accumulated in parallel with a near-complete shift toward the use of rodent animal models for the study of neural mechanisms, it is now essential to review the human data and the earlier animal data obtained from cats and monkeys to infer general conclusions and to optimize future choice of the most appropriate animal model.