Drift Model Research Articles

The Self‐Calibrating Tilt Accelerometer: A Method for Observing Tilt and Correcting Drift With a Triaxial Accelerometer

AbstractWe present observations from two field deployments of a calibrated tiltmeter that we name the Self‐Calibrating Tilt Accelerometer (SCTA). The tiltmeter is based upon a triaxial quartz crystal accelerometer; the horizontal channels measure tilt and are periodically rotated into the vertical to obtain a measurement of the acceleration of gravity. Changes in the measured total acceleration are ascribed to drift in the vertical channel and used as calibrations for removing that same drift from the tilt time series observed between calibrations. Changes in the span (sensitivity) of the accelerometer channels can also be measured by calibrating them pointing up and down. A 3‐year test on the seafloor at Axial Seamount show that the calibrations are consistent with a linear‐exponential model of drift to a RMS residual of ∼0.5 μg (μrad). The calibrated tilt time series was impacted by platform settling for the first 2 years, but after repositioning the tiltmeter, the calibrated observations were consistent for the final year with the tilt observed on a nearby LILY tiltmeter, within an assumed level of drift for the unconstrained LILY sensor. A separate 15‐month test in a stable vault at Piñon Flat Observatory was complicated by seasonal temperature variations of >5°C; the calibrations are consistent with a linear‐exponential model of drift to ∼2 μg RMS when temperature and temperature time‐derivative dependence is included. Similarly, the calibrated tilt time series was impacted by thermal deformation of the SCTA assembly. A future test in a thermally and tectonically stable borehole will be required to assess the accuracy of the SCTA.

FGTL: Federated Graph Transfer Learning for Node Classification

Unsupervised multi-source domain transfer in federated scenario has become an emerging research direction, which can help unlabeled target domain to obtain the adapted model through source domains under privacy-preserving. However, when local data is graph, the difference of domains (or data heterogeneity) mainly originate from the difference in node attributes and subgraph structures, leading to serious model drift, which is not considered by the existing related algorithms. Currently, there are two challenges in this scenario: (1) The node representations extracted directly through conventional GNNs lack inter-domain generalized and consistent information, making it difficult to apply existing federated learning algorithms. (2) The knowledge of source domains has quality differences, which may lead to negative transfer. To address these issues, we propose a novel two-phase Federated Graph Transfer Learning (FGTL) framework. In the generalization phase, FGTL utilizes local contrastive learning and global context embedding to force node representations to capture the inter-domain generalized and consistent information, lightly alleviating model drift. In the transfer phase, FGTL utilizes consensus knowledge to force the decision bound of classifier to adapt to the target client. In addition, FGTL+ exploits model grouping to make consensus knowledge generation more efficient, further enhancing the scalability of FGTL. Extensive experiments show that FGTL significantly outperforms state-of-the-art related methods, while FGTL+ further enhances privacy protection and reduces both communication and computation overhead.

Machine learning applications of network security enhancement: review

Machine learning (ML) is being used to improve intrusion detection mechanisms and identification in cyber security. Network data volume scaling (with the help of Machine learning) — Automated analysis and pattern recognition for large amounts of network-data, thereby detection of anomalies / potentially malicious activities that escape current rule-based techniques. By training ML models on historical data, these models can learn benign network behavior as well the anomalies in them that may result from malicious activities. The purpose of this essay/report is to begin taking a look under the hood at how ML can be used for security threat detection and analysis in-networking on an ongoing basis. This paper covers the automation of ML algorithms to enhance network security. Given the current state of electronic threats and their evolution, traditional security methods are typically insufficient. ML can analyze large volumes of data, learn patterns from it and this makes it suitable to complement network defense mechanisms. It then discusses different ML applications in network cybersecurity: intrusion detection, anomaly detection, spam and malware analysis; which the paper characterizes. It analyzes the potential, benefits and constrains of major ML methods in network security like supervised learning; unsupervised learning and reinforcement-learning. Finally, this paper represents recent progress in the use and impact of ML techniques along with case studies.The paper discusses the existing difficulties in the field such as the necessity for datasets and the vulnerability of machine learning models to adversarial attacks.T he paper also highlights avenues for exploration by focusing on developing scalable security solutions based on machine learning that are resilient and flexible.The goal of this examination is to offer both researchers and industry professionals valuable perspectives into the opportunities and obstacles linked to utilizing machine learning, in the domain of network security. ML methods have potential to improve network security by addressing the challenges posed by the increasing cyber threats that traditional security measures struggle to combat effectively over time. One key strength of ML lies in its capacity to analyze datasets and identify intricate patterns efficiently. The research paper delves into applications of ML in enhancing network security. The list covers security tools like Intrusion Detection Systems (IDS) examining malware and phishing attempts as well as anomalies in network activity and user behavior analysis (UEBA). The study explores both supervised and unsupervised learning methods. How they are used for quick threat detection and response in real time scenarios.You will find case studies and recent developments that showcase the implementation and effectiveness of these strategies.In addition the article delves into the obstacles linked to using machine learning techniques in network security including the necessity, for datasets, The paper's goal is to give an enlightening summary to scholars and practitioners about how machine learning can be applied to network security in order to provide solutions that are robust, adaptive, and scalable. To this end, it touches on several relevant aspects. One is the threat posed by adversarial attacks on the sorts of models that are likely to be used in this context. Another is the imperative, deriving from both adversarial threat and model drift, that models needed in this context be available in a form usable for continuous update. Keywords: Network Security, machine learning (ML), Intrusion Detection Systems (IDS), entity behavior analytics (UEBA).

Real-Time Tracking Target System Based on Kernelized Correlation Filter in Complicated Areas.

The achievement of rapid and reliable image object tracking has long been crucial and challenging for the advancement of image-guided technology. This study investigates real-time object tracking by offering an image target based on nuclear correlation tracking and detection methods to address the challenge of real-time target tracking in complicated environments. In the tracking process, the nuclear-related tracking algorithm can effectively balance the tracking performance and running speed. However, the target tracking process also faces challenges such as model drift, the inability to handle target scale transformation, and target length. In order to propose a solution, this work is organized around the following main points: this study dedicates its first part to the research on kernelized correlation filters (KCFs), encompassing model training, object identification, and a dense sampling strategy based on a circulant matrix. This work developed a scale pyramid searching approach to address the shortcoming that a KCF cannot forecast the target scale. The tracker was expanded in two stages: the first stage output the target's two-dimensional coordinate location, and the second stage created the scale pyramid to identify the optimal target scale. Experiments show that this approach is capable of resolving the target size variation problem. The second part improved the KCF in two ways to meet the demands of a long-term object tracking task. This article introduces the initial object model, which effectively suppresses model drift. Secondly, an object detection module is implemented, and if the tracking module fails, the algorithm is redirected to the object detection module. The target detection module utilizes two detectors, a variance classifier and a KCF. Finally, this work includes trials on object tracking experiments and subsequent analysis of the results. Initially, this research provides a tracking algorithm assessment system, including an assessment methodology and the collection of test videos, which helped us to determine that the suggested technique outperforms the KCF tracking method. Additionally, the implementation of an evaluation system allows for an objective comparison of the proposed algorithm with other prominent tracking methods. We found that the suggested method outperforms others in terms of its accuracy and resilience.

Thermal calibration for triaxial gyroscope of MEMS-IMU based on segmented systematic method

With the progress of micro electromechanical system (MEMS) technology, the performance of MEMS inertial measurement unit (IMU) composed of gyroscopes and accelerometers has been improved. Among the inertial sensors, MEMS triaxial gyroscope plays an important role in attitude estimation, navigation and positioning of intelligent mobile terminals such as unmanned aerial vehicles and unmanned vehicles. However, the measured values of low and medium cost MEMS triaxial gyroscopes are mainly affected by temperature (or thermal effect) and random errors. As results, the drift errors correlated with temperature will reduce application accuracy of MEMS triaxial gyroscope. The traditional calibration method for thermal drift errors relies on the expensive equipment, such as turntable and the temperature control system, which increases the cost of calibration. Therefore, an effective thermal calibration method that is available when using low- or high-cost tools for MEMS triaxial gyroscope will be meaningful for majority users. Hence, this paper analyzed and established the thermal drift model of MEMS triaxial gyroscope, and proposed a segmented systematic calibration method based on 24 states according to this model. Simulation shows that the proposed method can obtain the results approaching the real parameter thermal drift curves. Calibration experiments and parameters test show that the max errors of attitude calculation are reduced to 10% of the results using the original parameters, which indicates that the effectiveness of the proposed method.

Enhancing Maneuverability in a Variable Wheelbase Wheeled Mobile Robot Through Dynamic Steering Curvature Control

ABSTRACTVariable wheelbase wheeled mobile robot (VW‐WMR) is capable of maneuvering flexibly and traversing on rough and soil terrains within confined spaces. While the steering radius of the robot model can be robustly changed by the variable wheelbase length, a challenge is posed in accurately tracking a predefined trajectory through the alteration of wheelbase length. A dynamic steering curvature (DSC) control method is proposed in this work to overcome this challenge, which is achieved by two different approaches utilizing the variable wheelbase length and manipulating drifting motions. First, to enable flexible trajectory adjustments, a 3D Ackermann kinematics model, incorporating the lifting motion of the robot box, is developed to control steering curvature by the changes in wheelbase length. Second, to achieve flexible movement for the inner and outer curves of the originally planned curvilinear trajectory, Ackermann drift models are presented by the establishment of two sequence instantaneous centers. Furthermore, the control performance of DSC is validated through simulation experiments on the ROSTDyn Vortex platform, using a six‐wheeled VW‐WMR named HIT‐MRII robot. The effectiveness of DSC for strategically altering the robot's motion trajectory is demonstrated by the results, which show the relation between the changed trajectory position and the variation in wheelbase length, and the variation in the radii of the instantaneous centers (ICs) in the Ackermann drift model, respectively. In addition, the high maneuverability of the robot using the Ackermann model is proven by physical experiments during steering motions on soil terrain. A comprehensive advantage is demonstrated by the results via Ackermann models, which include shorter runtimes, moderate travel distances, and moderate variations in force on the wheels, compared to different velocity, crabbing motion, and equivalent bicycle models.

High-resolution regional sea-ice model based on the discrete element method with boundary conditions from a large-scale model for ice drift

Abstract Understanding sea-ice dynamics at the floe scale is crucial to comprehend polar climate systems. While continuum models are commonly used to simulate large-scale sea-ice dynamics, they have limitations in accurately representing sea-ice behaviour at small scales. DEMs, on the other hand, are well-suited for modelling the behaviour of individual ice floes but face limitations due to computational constraints. To address the limitations of both approaches while combining their strengths, we explored the feasibility of using a DEM within a continuum model, where the latter provides boundary conditions for a rectangular high-resolution DEM domain. This paper presents a feasibility study where a discrete model of a 100 × 100 km2 icefield was created using high-resolution optical satellite imagery. Sea-ice dynamics were simulated in the DEM considering environmental forces and integrating large-scale ice-drift velocities as boundary conditions. Model predictions were compared with satellite observations for ice drift and deformation parameters. This numerical approach showed potential for offering accurate, high-resolution predictions of sea ice, particularly in coastal areas and near islands, and may find applications in ice navigation and climate studies. However, further development of the DEM, along with upgrades to the coupled ocean models providing data for the ice component, may be necessary. Additionally, challenges remain to develop a two-way coupling between the DEM and a continuum model, which may be needed to improve the accuracy of large-scale simulations.

Overreaction and underreaction to new information and the directional forecast of exchange rates

Overreaction and underreaction to new information may cause, respectively, sequences and reversals in exchange rate returns. Empirical evidence is that the sign of the difference between the probabilities of a sequence and a reversal predicts the sign of the next day exchange rate return out-of-sample better than the random walk without drift model. Since the effect of overreaction of exchange rates to unexpected news dies out over longer time spans, the directional predictability becomes weaker for the weekly exchange rates. The trading strategy exploiting the directional predictability of the exchange rates generates tangible profits compared to the non-trading and buy-and-hold strategies.

Utilizing multi-point temperature sensing to evaluate the low frequency noise of phasemeter for intersatellite laser interferometer.

High-precision phasemeters are a key technology in intersatellite laser interferometers used for detecting gravitational waves (GWs) in space. As the core of the readout system, the phasemeter must operate in the bandwidth of 5-25MHz, and its resolution needs to reach the order of μrad/Hz at mHz. It presents significant challenges to electronic signal processing technology. To investigate the primary noise source in the low-frequency band, a mathematical model of thermal drift to phase noise was established, and a multi-point temperature sensing scheme for critical electronic components was proposed. In particular, we evaluated a phasemeter based on a commercial platform and assessed the thermal drift noise according to the proposed model. This study identifies and explains the effects of temperature linear drift and overcorrection in components, demonstrating that thermal drift noise is the main noise source for the phasemeter at frequencies from 0.1 to 1 mHz. In addition, the proposed scheme is universal in its applicability and may be implemented in any circuit for the evaluation of temperature effects on the components of interest.

Augmenting human-guided progressive learning with machine vision systems for robust surface defect detection

Machine vision systems commonly utilize Convolutional Neural Networks (CNNs) for in-line surface defect detection of manufactured components. The prediction abilities of vision-based inspection systems deteriorate with time as the defect detection model trained on fixed image datasets fails to accommodate deviations. This paper proposes a human-guided progressive learning approach that systematically imparts learning of new features to the CNN-powered vision-based defect detection system. The approach augments the surface defect detection model with human intelligence, using an intuitive user interface to address model drift. The human expert monitors the trained model performance under specific conditions leading to the change of characteristics during implementation, identifies misclassifications, and initiates re-training. The algorithm accumulates misclassified data till a pre-defined threshold level is reached or a human expert terminates inspection. The misclassified results merge with the original datasets for progressive re-training using a strategy similar to the base model development. The present work utilizes pre-trained CNN Efficientnet-b0 to develop the surface defect detection model for tapered roller inspection through transfer learning. It is concluded that the progressive re-training improves defect detection performance and reduces misclassifications. The Matthews Correlation Coefficient (MCC) score, derived from the confusion matrix, showed improvement from 0.6 to 0.82 after four iterations. A cross-model benchmarking study is also performed to show the versatility of the proposed approach. The present work demonstrated that the human-guided progressive learning approach can provide adaptability to vision-based surface defect detection utilizing deep learning algorithms and enhance system performance during real-time implementation.

A hybrid approach for skillful multiseasonal prediction of winter North Pacific blocking

Wintertime atmospheric blocking often brings adverse environmental and socioeconomic impacts through its accompanying temperature and precipitation extremes. However, due to the chaotic nature of the extratropical atmospheric circulation and the challenges in simulating blocking, the skillful seasonal prediction of blocking remains elusive. In this study, we leverage both observational data and seasonal hindcasts from a state-of-the-art seasonal prediction system to investigate the prediction skill of North Pacific wintertime blocking frequency and its linkage to downstream cold extremes. The observational results show that North Pacific blocking has a local maximum over the central North Pacific Ocean and that the occurrence of North Pacific blocking drives significant cold anomalies over northwestern North America within a week, which are both well reproduced by the model. The model skillfully predicts the western North Pacific blocking frequency near the subtropical jet exit region at the shortest forecast lead, but skill drops off rapidly with lead time partly due to model drift in the background flow. To overcome this rapid drop in skill, we develop a linear hybrid dynamical-statistical model that uses the forecasted Niño 3.4 index and upstream precipitation as predictors and that maintains significant forecast skill of high-latitude North Pacific blocking up to 7 lead months in advance. Our results indicate that an improvement in the seasonal prediction skill of winter North Pacific blocking frequency may be achieved by the enhanced representation of the links among sea surface temperature anomalies, tropical convection, and the ensuing tropical-extratropical interaction that initiates North Pacific blocking.

Secular and modulator-specific drifts in the predictive performance of a rapid lung function decline algorithm: a cystic fibrosis patient registry study

PurposePredicting the onset of pulmonary exacerbation for people with cystic fibrosis (CF) is critical to identify those at high risk in advance and allows timely clinical intervention. The FEV1-indicated exacerbation signal (FIES) has been proposed by the CF Learning Network to identify pulmonary exacerbation events, but temporal predictive performance of this marker is unknown.MethodsTo assess the epidemiologic impact of secular trends in CF care and treatments on algorithms aimed at prediction of pulmonary exacerbation, we evaluated prediction model “drift” using a longitudinal retrospective cohort of 29,476 individuals from the U.S. CF Foundation Patient Registry (2011–2018). The FIES marker was defined for each clinical encounter using relative drops in lung function, measured as forced expiratory volume in 1 s of % predicted (FEV1pp). We formed predictive probabilities for each FIES event via a target function for a predictive algorithm of FEV1pp. Models were trained using 2 year data windows with prediction performance tested on subsequent years.ResultsYear-over-year FIES event discrimination remained consistent (area under the receiver-operator characteristic curve, AUC > 80%). Annual drift was < 1% with actual drop in AUC ~ 3% in 6 years. The model performed best in individuals with moderate-to-low baseline FEV1pp (AUC range: 84–85%). For relatively higher FEV1pp levels, discrimination was slightly lower with higher AUC variability.ConclusionWhile model fit improved by accounting for modulator therapy, prediction performance was similar. Most impactful model drifts occurred after 2012 coinciding with increased ivacaftor modulator use. Additional evaluation will be necessary given recent, broad uptake of highly effective modulators.

Memristive behaviour of Al/rGO-CdS/FTO device at different temperatures: A MATLAB-integrated study

In the present work, a comprehensive study is carried out to investigate the memristive behaviour of reduced graphene oxide (rGO) conjugated cadmium sulphide quantum dot (CdS QD) nanocomposites (rGO-CdS), offering insights into their dynamic response under varying thermal conditions. The study integrates experimental analysis with MATLAB simulation to provide a detailed understanding of the complex interplay between different operating temperature (300K, 350K, 400K, and 450K) and memristive behavior in rGO-CdS nanocomposites. Different structural and chemical characterizations were carried out which confirms the formation of rGO-CdS nanocomposites. A sandwich structured device was fabricated with the synthesized rGO-CdS nanocomposites using Aluminum (Al) as top and Fluorine doped tin oxide (FTO) as bottom electrode. The influence of operating temperature on hysteresis behaviour of the fabricated Al/rGO-CdS/FTO device was investigated using Keithley 2450 source meter by sweeping a direct current (dc) voltage (−5 V → 5 V → −5 V). Notably, we observe a positive temperature coefficient in the device current, with maximum and minimum recorded current of |1.29mA| and |0.66mA| at 450K and 300K respectively. The current-voltage (I-V) behavior observed in the device reveals that in the low resistance state (LRS), conduction is dominated by bulk-limited mechanisms. However, in the high resistance state (HRS), conduction involves contributions from both Schottky barriers and the Pool-Frenkel effect. A MATLAB based linear drift model was used to simulate the device responses at different temperatures using the experimental data. The study provides first comprehensive analysis of temperature dependent hysteresis behaviour of Al/rGO-CdS/FTO device, integrating MATLAB simulation to glean valuable insights into its operation and possible applications as memristive material across different temperature regimes.

Steric Sea Level Rise and Relationships with Model Drift and Water Mass Representation in GFDL CM4 and ESM4

Abstract Density-driven steric seawater changes are a leading-order contributor to global mean sea level rise. However, inter-model differences in the magnitude and spatial patterns of steric sea level rise exist at regional scales and often emerge during the spin-up and pre-industrial control integrations of climate models. Steric sea level results from an eddy-permitting climate model, GFDL-CM4, are compared with a lower resolution counterpart, GFDL-ESM4. The results from both models are examined through basin-scale heat budgets and watermass analysis, and we compare the patterns of ocean heat uptake, redistribution, and sea level differ in ocean-only (i.e. OMIP) and coupled climate configurations. After correcting for model drift, both GFDL-CM4 and GFDL-ESM4 simulate nearly equivalent ocean heat content change and global sea level rise during the historical period. However, the GFDL-CM4 model exhibits as much as a 40% increase in surface ocean heat uptake in the Southern Ocean and subsequent increases in horizontal export to other ocean basins after bias correction. The results suggest regional differences in the processes governing Southern Ocean heat export, such as the formation of AAIW, SPMW, and gyre transport between the two models, and that sea level changes in these models cannot be fully bias-corrected. Since the process-level differences between the two models are evident in the preindustrial control simulations of both models, these results suggest that the control simulations are important for identifying and correcting sea-level related model biases.

Integrating Habitat Suitability and Larval Drift Modeling for Spawning‐To‐Nursery Functional Habitat Connectivity Analysis in Rivers

AbstractHabitat suitability modeling is a commonly used methodology to plan and assess in‐stream habitat enhancement in rivers, such as for the key fish life stages spawning and juvenile development. However, their use only allows modeling the spatial distribution of habitats, but not their connectivity. By integrating micro‐scale habitat modeling and a larval drift model, we assess the functional connectivity between spawning and larval nursery habitats for four rheophilic and litophilic fish species in a channelized and hydropower impacted reach of the lower Inn River (Bavaria, Germany), in which two restoration measures, a bypass channel and an island side‐channel system, have been constructed to improve longitudinal connectivity and habitat conditions. The study aims to (a) map spawning and larval nursery habitats, (b) quantify their connectivity, and (c) optimize functional habitat connectivity through an alternative bypass channel location. Results show that the channel's morphological complexity influences quantity and quality of available spawning and nursery habitats and their connectivity. Despite the presence of nursery habitats across the analyzed channel, the slow lateral larval dispersion during drift limits their accessibility to only the left river bank, making only 33% of available habitats useable. The degree of functional connectivity and hence the percentage of useable habitats can however be increased up to 95.3 % when considering different spatial configurations of habitats, which were explored in two alternative restoration scenarios. The results demonstrate the importance of considering functional habitat connectivity in habitat assessments and restoration planning.

A novel temperature drift compensation method based on LSTM for NMR sensor

Nuclear Magnetic Resonance (NMR) sensors have become one of the best choices for angular velocity sensors in future Inertial Navigation Systems (INS) due to their small size and high precision. The performance of NMR sensors can be affected by temperature changes, resulting in temperature drift. Therefore, temperature compensation is essential. Accurate temperature compensation is challenging due to the lack of direct temperature observations in the core sensitive area (cell) and the complex temperature drift model. Considering the time correlation of temperatures inside and outside the cell, a new temperature drift compensation method based on the Long-Short Term Memory (LSTM) network is proposed. The memory characteristics of the LSTM enable it to fully utilize current and previous data to learn complex models. The main contributions of this study are as follows: (1) the mechanism of temperature drift in NMR sensors was analyzed; (2) a multi-time scale input–output model for temperature changes, temperature change rates, and ambient temperature versus sensor output bias was established; (3) a temperature drift model identification method based on LSTM was designed. Experimental results show that, compared to traditional polynomial fitting and memory-free methods, the proposed method improves temperature compensation accuracy by 26.0% to 47.0%. This method effectively reduces the adverse impact of temperature changes on the NMR sensor.

A new ship tracing technology from oil spills based on multi-source data

Oil spill from ship can cause serious pollution to the Marine environment, but it is very difficult to find and confirm the troublemaker. In order to determine the oil spill ship, this paper proposes a new method to trace the source of ship oil spills and find the suspected ship that spills oil based on SAR imagery, AIS data and related marine environment data. First, we filter AIS data based on position of oil spill areas on remote sensing imagery and convert oil spill areas into trajectory points. Secondly, based on the Lagrangian particle motion model, a bidirectional drift model is proposed to calculate the average similarity between the forward and backward drift results. Finally, the most likely oil spill ship is determined according to the average similarity results. The results of the case study show that the method is effective and practical.

Asymptotic profiles of positive steady states in a reaction–diffusion benthic–drift model

AbstractIn this paper, we investigate a reaction–diffusion–advection benthic–drift model, where the population is divided into two interacting groups: individuals dispersing in the drift zone and individuals living in the benthic zone. For different growth types of the benthic population, we present asymptotic profiles of positive steady states in three cases: (i) large advection; (ii) small diffusion of the drift population; and (iii) large diffusion of the drift population. We prove that in case (i) both the benthic and drift individuals concentrate only at the downstream end; in case (ii), both benthic and drift population reside inhomogeneously in , stay away from the upstream end , and concentrate only at the downstream ; and in case (iii), the drift species distributes evenly on the entire habitat and the benthic species distributes inhomogeneously throughout the habitat. The result supplements the dynamical behaviors of benthic–drift models developed in earlier works and is also of its own interest.

Data-free knowledge distillation via generator-free data generation for Non-IID federated learning

Data heterogeneity (Non-IID) on Federated Learning (FL) is currently a widely publicized problem, which leads to local model drift and performance degradation. Because of the advantage of knowledge distillation, it has been explored in some recent work to refine global models. However, these approaches rely on a proxy dataset or a data generator. First, in many FL scenarios, proxy dataset do not necessarily exist on the server. Second, the quality of data generated by the generator is unstable and the generator depends on the computing resources of the server. In this work, we propose a novel data-Free knowledge distillation approach via generator-Free Data Generation for Non-IID FL, dubbed as FedF2DG. Specifically, FedF2DG requires only local models to generate pseudo datasets for each client, and can generate hard samples by adding an additional regularization term that exploit disagreements between local model and global model. Meanwhile, FedF2DG enables flexible utilization of computational resources by generating pseudo dataset locally or on the server. And to address the label distribution shift in Non-IID FL, we propose a Data Generation Principle that can adaptively control the label distribution and number of pseudo dataset based on client current state, and this allows for the extraction of more client knowledge. Then knowledge distillation is performed to transfer the knowledge in local models to the global model. Extensive experiments demonstrate that our proposed method significantly outperforms the state-of-the-art FL methods and can serve as plugin for existing Federated Learning methds such as FedAvg, FedProx, etc, and improve their performance.

Asynchronous abundance fluctuations can drive giant genotype frequency fluctuations.

Large stochastic population abundance fluctuations are ubiquitous across the tree of life1-7, impacting the predictability of population dynamics and influencing eco-evolutionary outcomes. It has generally been thought that these large abundance fluctuations do not strongly impact evolution, as the relative frequencies of alleles in the population will be unaffected if the abundance of all alleles fluctuate in unison. However, we argue that large abundance fluctuations can lead to significant genotype frequency fluctuations if different genotypes within a population experience these fluctuations asynchronously. By serially diluting mixtures of two closely related E. coli strains, we show that such asynchrony can occur, leading to giant frequency fluctuations that far exceed expectations from models of genetic drift. We develop a flexible, effective model that explains the abundance fluctuations as arising from correlated offspring numbers between individuals, and the large frequency fluctuations result from even slight decoupling in offspring numbers between genotypes. This model accurately describes the observed abundance and frequency fluctuation scaling behaviors. Our findings suggest chaotic dynamics underpin these giant fluctuations, causing initially similar trajectories to diverge exponentially; subtle environmental changes can be magnified, leading to batch correlations in identical growth conditions. Furthermore, we present evidence that such decoupling noise is also present in mixed-genotype S. cerevisiae populations. We demonstrate that such decoupling noise can strongly influence evolutionary outcomes, in a manner distinct from genetic drift. Given the generic nature of asynchronous fluctuations, we anticipate that they are widespread in biological populations, significantly affecting evolutionary and ecological dynamics.