FLaMAS: Federated Learning Based on a SPADE MAS

Rincon, J. A.; Julián, Vicente; Carrascosa, Carlos

doi:10.3390/app12073701

Cited by 10 publications

(5 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on these parameters, it is possible to identify the tool that best suits the user's needs, finding the right compromise between ease of use and efficiency. Given their inherent distributed nature, ABMs lend themselves well to analyzing phenomena from the ever more attractive distributed computing domains such as federated learning [121][122][123][124] and blockchain systems [125][126][127][128]. An interesting perspective worth exploring in future work could be assessing the current ABM tool landscape's availability and suitability in such domains to inform possible feature development in this context.…”

Section: Discussionmentioning

confidence: 99%

Experimenting with Agent-Based Model Simulation Tools

et al. 2022

View full text Add to dashboard Cite

Agent-based models (ABMs) are one of the most effective and successful methods for analyzing real-world complex systems by investigating how modeling interactions on the individual level (i.e., micro-level) leads to the understanding of emergent phenomena on the system level (i.e., macro-level). ABMs represent an interdisciplinary approach to examining complex systems, and the heterogeneous background of ABM users demands comprehensive, easy-to-use, and efficient environments to develop ABM simulations. Currently, many tools, frameworks, and libraries exist, each with its characteristics and objectives. This article aims to guide newcomers in the jungle of ABM tools toward choosing the right tool for their skills and needs. This work proposes a thorough overview of open-source general-purpose ABM tools and offers a comparison from a two-fold perspective. We first describe an off-the-shelf evaluation by considering each ABM tool’s features, ease of use, and efficiency according to its authors. Then, we provide a hands-on evaluation of some ABM tools by judging the effort required in developing and running four ABM models and the obtained performance.

show abstract

Section: Discussionmentioning

confidence: 99%

Experimenting with Agent-Based Model Simulation Tools

et al. 2022

View full text Add to dashboard Cite

show abstract

“…where D = c D c represents for the whole training dataset over devices subset C and |D| = ∑ |C| c=1 |D c | denotes the total number of the data samples. To solve the above-distributed optimization problem, an incomplete list of studies have offered their solutions [2,[20][21][22][23][24][25][26]. In [2], the FederatedAveraging (FedAvg) is first advocated to combine local stochastic gradient descent (SGD) on each device with a server that performs model averaging.…”

Section: Federated Learningmentioning

confidence: 99%

“…In [24], a layers-wise Federated Matched Averaging (FedMA) is proposed for convolutional neural networks (CNNs) and long-short-term memory (LSTM) to address the data heterogeneity. In [25], the authors propose Federated Learning Based on a SPADE MAS (FLaMAS), which designs a multi-agent system to enable flexibility and dynamism in FL. In [26], a Federated Learning-Based Graph Convolutional Network (FedGCN) is proposed to process non-Euclidean data.…”

Section: Federated Learningmentioning

confidence: 99%

DACFL: Dynamic Average Consensus-Based Federated Learning in Decentralized Sensors Network

Chen

Zhu

et al. 2022

Sensors

View full text Add to dashboard Cite

Federated Learning (FL) is a privacy-preserving way to utilize the sensitive data generated by smart sensors of user devices, where a central parameter server (PS) coordinates multiple user devices to train a global model. However, relying on centralized topology poses challenges when applying FL in a sensors network, including imbalanced communication congestion and possible single point of failure, especially on the PS. To alleviate these problems, we devise a Dynamic Average Consensus-based Federated Learning (DACFL) for implementing FL in a decentralized sensors network. Different from existing studies that replace the model aggregation roughly with neighbors’ average, we first transform the FL model aggregation, which is the most intractable in a decentralized topology, into the dynamic average consensus problem by treating a local training procedure as a discrete-time series.We then employ the first-order dynamic average consensus (FODAC) to estimate the average model, which not only solves the model aggregation for DACFL but also ensures model consistency as much as possible. To improve the performance with non-i.i.d data, each user also takes the neighbors’ average model as its next-round initialization, which prevents the possible local over-fitting. Besides, we also provide a basic theoretical analysis of DACFL on the premise of i.i.d data. The result validates the feasibility of DACFL in both time-invariant and time-varying topologies and declares that DACFL outperforms existing studies, including CDSGD and D-PSGD, in most cases. Take the result on Fashion-MNIST as a numerical example, with i.i.d data, our DACFL achieves 19∼34% and 3∼10% increases in average accuracy; with non-i.i.d data, our DACFL achieves 30∼50% and 0∼10% increases in average accuracy, compared to CDSGD and D-PSGD.

show abstract

“…In this way, a centralized collective model is constructed from the contributions of all participating nodes. Similar to our perspective, Rincon et al [18] address this issue by introducing an MAS that forms a flexible and dynamic federated learning framework, allowing for the easy addition of nodes to the system, namely FlaMAS (Federated Learning Based on a SPADE MAS). This proposal has been experimented on in the SPADE platform with the well-known MNIST dataset.…”

Section: Introductionmentioning

confidence: 99%

Multi-Agent Systems and Machine Learning for Wind Turbine Power Prediction from an Educational Perspective

Soygazi

2023

Sustainability

View full text Add to dashboard Cite

Artificial intelligence (AI) is an umbrella term that encompasses different fields of study, and topics related to these fields are addressed separately or within the scope of AI. Multi-agent systems (MASs) and machine learning (ML) are the core concepts of AI that are taught during AI courses. The separate explanation of these core research areas is common, but the emergence of federated learning has triggered their combined usage. This paper describes a practical scenario in the energy domain where these technologies can be used together to provide a sustainable energy solution for predicting wind turbine active power production. The projects in the AI course were assigned prior to the step-by-step learning of MASs and ML. These concepts were applied using a wind turbine energy dataset collected in Turkey to predict the power production of wind turbines. The observed performance improvements, achieved by applying various agent architectures and data partitioning scenarios, indicate that boosting methods such as LightGBM yield better results even when the settings are modified. Additionally, a questionnaire about the assignments was filled out by the student groups to assess the impact of learning MASs and ML through project-based education. The application of MASs and ML in a hybrid way proves valuable for learning core concepts related to AI education, as evidenced by feedback from students.

show abstract

FLaMAS: Federated Learning Based on a SPADE MAS

Cited by 10 publications

References 25 publications

Experimenting with Agent-Based Model Simulation Tools

Experimenting with Agent-Based Model Simulation Tools

DACFL: Dynamic Average Consensus-Based Federated Learning in Decentralized Sensors Network

Multi-Agent Systems and Machine Learning for Wind Turbine Power Prediction from an Educational Perspective

Contact Info

Product

Resources

About