A Crowdsourcing Framework for On-Device Federated Learning
read more
Citations
Federated Machine Learning: Survey, Multi-Level Classification, Desirable Criteria and Future Directions in Communication and Networking Systems
Federated Learning Meets Blockchain in Edge Computing: Opportunities and Challenges
Federated Learning in Vehicular Edge Computing: A Selective Model Aggregation Approach
Federated Learning for Vehicular Internet of Things: Recent Advances and Open Issues
Federated Learning for Internet of Things: Recent Advances, Taxonomy, and Open Challenges
References
Distributed Optimization and Statistical Learning Via the Alternating Direction Method of Multipliers
Communication-Efficient Learning of Deep Networks from Decentralized Data
Communication-Efficient Learning of Deep Networks from Decentralized Data
Federated Learning: Strategies for Improving Communication Efficiency
Mobile crowdsensing: current state and future challenges
Related Papers (5)
Frequently Asked Questions (14)
Q2. What future works have the authors mentioned in the paper "A crowdsourcing framework for on-device federated learning" ?
For future work, the authors will focus on mobile crowdsourcing framework to enable the self-organizing FL that considers task offloading 14 strategies for the resource constraint devices. Another direction is to study the impact of discriminatory pricing scheme for participation. The authors also plan to further investigate on participating client ’ s behavior, in terms of incentive and communication efficiency, to incorporate cooperative data trading scenario for the proposed framework [ 48 ], [ 49 ]. The authors will consider the scenario where the central coordinating MEC server is replaced by one of the participating clients and devices can offload their training task to the edge computing infrastructure.
Q3. What is the value of the bound in (19)?
(20)Because the cost of communication is proportional to the speed and energy consumption in a distributed scenario [20], the bound defined in (19) explains the efficiency in terms of MEC server’s resource restriction for attaining ǫ accuracy.
Q4. What is the function that broadcasts the global parameters for the next communication round?
MEC ServerGlobal Modelin (8), and broadcasts the global parameters required for the participating clients to solve their local subproblems for the next communication round.
Q5. What is the motivation for the MEC server to build a centralized model?
the MEC server builds a high quality centralized model characterized by its utility function, with the data distributed over the participating clients by offering the reward.
Q6. How does the author show that the heuristic approach can achieve the maximum utility?
the authors show that their mechanism design can achieve the optimal solution while outperforming a heuristic approach for attaining the maximal utility with up to 22% of gain in the offered reward.
Q7. What is the MEC server’s ability to maintain the maximum local consensus accuracy?
Since the threshold accuracy θth can be adjusted by the MEC server for each round of solution, each participating client will maintain a response towards the maximum local consensus accuracy θth.
Q8. What is the main argument for the improvement in generalization performance of local SGD?
authors in their recent work [25] argue for the sufficient improvement in generalization performance with the variant of local SGD rather than the large mini-batch sizes, even in a non-convex setting.
Q9. What is the effect of the proposed framework on the local consensus accuracy?
Through a probabilistic model, the authors have designed and presented numerical results on an admission control strategy for the number of client’s participation to attain the corresponding local consensus accuracy.
Q10. What is the way to measure the utility of a MEC server?
To be more specific about this relation, the authors can observe that with the increased value of (1 − θ), i.e., lower relative accuracy (high local accuracy), the MEC server can attain better utility due to corresponding increment in value of x(ǫ).
Q11. What is the definition of accuracy for the local clients?
The authors consider the local θ accuracy for the participating clients is an i.i.d and uniformly distributed random variable over the range [θmin, θmax], then the PDF of the responses can be defined as fθ(θ) =1 θmax−θmin .
Q12. What is the case for the utility maximization problem?
for the measured θ ∗ from the participating clients at MEC server, the utility maximization problem can be formulated as follows:max r≥0,x(ǫ)U(x(ǫ), r|θ∗), (21)s.t. x(ǫ)1−maxk θ∗k(r) ≤ δ. (22)In constraint (22), maxk θ ∗ k(r) characterizes the worst case response for the server side utility maximization problem with the bound on permissible global iterations.
Q13. What is the learning setting for a strongly convex model?
The authors consider the learning setting for a strongly convex model such as logistic regression, as discussed in Section III, to characterize and demonstrate the efficacy of the proposed framework.
Q14. What is the significance of choosing a local th accuracy?
Their earlier discussion in Section IV and simulation results explain the significance of choosing a local θth accuracy to build a global model that maximizes the utility of the MEC server.