This paper explores decentralized learning in a graph-based setting, where data is distributed across nodes. We investigate a decentralized SGD algorithm that utilizes a random walk to update a global model based on l...
详细信息
ISBN:
(数字)9798350382846
ISBN:
(纸本)9798350382853
This paper explores decentralized learning in a graph-based setting, where data is distributed across nodes. We investigate a decentralized SGD algorithm that utilizes a random walk to update a global model based on local data. Our focus is on designing the transition probability matrix to speed up convergence. While importance sampling can enhance centralized learning, its decentralized counterpart, using the Metropolis-Hastings (MH) algorithm, can lead to the entrapment problem, where the random walk becomes stuck at certain nodes, slowing convergence. To address this, we propose the Metropolis-Hastings with Levy Jumps (MHLJ) algorithm, which incorporates random perturbations (jumps) to overcome entrapment. We theoretically establish the convergence rate and error gap of MHLJ and validate our findings through numerical experiments.
We present a combined study of the mechanical properties of 3D printed scaffolds made by nanocomposite materials based on polycaprolactone (PCL). The geometry and dimensions of the three different systems is the same....
We present a combined study of the mechanical properties of 3D printed scaffolds made by nanocomposite materials based on polycaprolactone (PCL). The geometry and dimensions of the three different systems is the same. Τhe porosity is 50% for all systems. Distributions of von-Mises strains and stresses, and total deformations were obtained through Finite Element Analysis (FEA) for a maximum amount of force applied, in a compressive numerical experiment. Also compressive experiments were performed for both raw and 3D nanoconposite scaffolds.
As the power grid undergoes significant paradigm shift due to the increasing penetration of renewable generation, the ever-growing installation of behind-the-meter (BTM) solar generation in the power grid also has a s...
详细信息
ISBN:
(数字)9798350381832
ISBN:
(纸本)9798350381849
As the power grid undergoes significant paradigm shift due to the increasing penetration of renewable generation, the ever-growing installation of behind-the-meter (BTM) solar generation in the power grid also has a significant impact on nodal loads, posing challenges on transmission operators. Furthermore, increasing frequent and severe extreme weather events intertwine with ubiquitous BTM solar generations and have amplified the challenges of accurately model nodal load profiles, especially under the lack of ground-truth information for verification. To tackle these challenges, this paper introduces a bilevel model that utilizes year-long data (e.g., proxy solar, zonal load, and individual node load profiles) to disaggregate metered profiles into actual demand and BTM solar generation at each transmission node. The proxy solar not only scales the BTM solar generation of individual nodes but also create a compensation term for enhancing performance on days with unexpected extreme weather events. The proposed algorithm is validated with real-world PJM Interconnection data during unexpected events like the recent Winter Storm Elliott. For quantitative evaluations, a novel Score
error
is introduced, which is based on mean percentages and load scales and offers a universal assessment method suitable for all nodes and different data formats (e.g., normalized or raw values).
This paper presents the design of a through-metal acoustic wireless power transfer (AWPT) system using piezoelectric transducers (PT) as transmitter (TX) and receiver (RX). An equivalent circuit model using measured i...
详细信息
ISBN:
(数字)9798350307238
ISBN:
(纸本)9798350307245
This paper presents the design of a through-metal acoustic wireless power transfer (AWPT) system using piezoelectric transducers (PT) as transmitter (TX) and receiver (RX). An equivalent circuit model using measured impedance parameters is used to model the complete PT-to-PT electromechanical conversion. The system is designed by sweeping frequency, load resistance, and compensation design to maximize transferred power. Boundary constraints are presented based on temperature and voltage limits the PTs can withstand. An optimal design staying within the operating bounds is then chosen and tested in hardware. The designed system is fabricated and verified experimentally for varying voltage levels. The prototype system has been tested up to 160 V dc input and 41 W output.
In dynamic inductive power transfer (DIPT) systems for automated guided vehicles (AGVs), the movement of the AGV can significantly impact the coupling coefficient, degrading the output performance. Conventional contro...
详细信息
The flexibility of unmanned aerial vehicles(UAVs)allows them to be quickly deployed to support ground *** reflecting surface(IRS)can reflect the incident signal and form passive beamforming to enhance the signal in th...
详细信息
The flexibility of unmanned aerial vehicles(UAVs)allows them to be quickly deployed to support ground *** reflecting surface(IRS)can reflect the incident signal and form passive beamforming to enhance the signal in the specific *** by the promising benefits of both technologies,we consider a new scenario in this paper where a UAV uses non-orthogonal multiple access to serve multiple users with *** to their distance to the UAV,the users are divided into the close users and remote *** UAV hovers above the close users due to their higher rate requirement,while the IRS is deployed near the remote users to enhance their received *** aim at minimizing the transmit power of UAV by jointly optimizing the beamforming of UAV and the phase shift of IRS while ensuring the decoding ***,the problem is ***,we decompose it into two sub-problems,including the transmit beamforming optimization and phase shift optimization,which are transformed into second-order cone programming and semidefinite programming,*** propose an iterative algorithm to solve the two sub-problems *** results prove the effectiveness of the proposed scheme in minimizing the transmit power of UAV.
Conventional tomographic reconstruction typically depends on centralized servers for both data storage and computation, leading to concerns about memory limitations and data privacy. Distributed reconstruction algorit...
详细信息
In modern software development, Python third-party libraries play a critical role, especially in fields like deep learning and scientific computing. However, API parameters in these libraries often change during evolu...
详细信息
The rapid growth of offshore wind energy needs robust condition monitoring to ensure the reliability and efficiency of direct-drive permanent magnet synchronous generators (PMSGs). Demagnetization, a critical fault in...
详细信息
ISBN:
(数字)9798331520748
ISBN:
(纸本)9798331520755
The rapid growth of offshore wind energy needs robust condition monitoring to ensure the reliability and efficiency of direct-drive permanent magnet synchronous generators (PMSGs). Demagnetization, a critical fault in these generators, can severely impact performance and increase operational costs, especially in harsh marine environments. Traditional diagnostic techniques often struggle to detect subtle magnetic property changes under varying load conditions. This paper presents an innovative framework for demagnetization detection in PMSGs, leveraging the electrical Multi-phase Imbalance Separation Technique (eMIST) combined with machine learning models. Using a dataset from offshore wind turbines, including simulated faulty conditions, the proposed methodology enhances anomaly detection by isolating imbalance signatures and correlating them with key operational parameters. Gaussian Process Regression (GPR) is employed to predict anomalies, offering high accuracy and confidence in fault classification. Results demonstrate the approach's ability to identify early-stage demagnetization, enabling predictive maintenance, reducing downtime, and improving the economic viability of offshore wind farms.
暂无评论