The DOI matches its Crossref record; 17 of the 17 references it cites were found in a registry or an index; none of the 25 sources its claims cite is checked for support yet.
In-Edge AI: Intelligentizing Mobile Edge Computing, Caching and Communication by Federated Learning
Open accessAbstract
Recently, along with the rapid development of mobile communication technology, edge computing theory and techniques have been attracting more and more attention from global researchers and engineers, which can significantly bridge the capacity of cloud and requirement of devices by the network edges, and thus can accelerate content delivery and improve the quality of mobile services. In order to bring more intelligence to edge systems, compared to traditional optimization methodology, and driven by the current deep learning techniques, we propose to integrate the Deep Reinforcement Learning techniques and Federated Learning framework with mobile edge systems, for optimizing mobile edge computing, caching and communication. And thus, we design the “In-Edge AI” framework in order to intelligently utilize the collaboration among devices and edge nodes to exchange the learning parameters for a better training and inference of the models, and thus to carry out dynamic system-level optimization and application-level enhancement while reducing the unnecessary system communication load. “In-Edge AI” is evaluated and proved to have near-optimal performance but relatively low overhead of learning, while the system is cognitive and adaptive to mobile communication systems. Finally, we discuss several related challenges and opportunities for unveiling a promising upcoming future of “In-Edge AI.”
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
The paper proposes “In-Edge AI,” integrating Deep Reinforcement Learning and Federated Learning into mobile edge systems to optimize edge caching and computation offloading. It reports proof-of-concept simulations where DDQN with FL approaches centralized DDQN performance while reducing wireless transmission cost.
Key findings
- In edge caching, DDQN with FL achieved hit rates close to centralized DDQN for three edge nodes (E1, E2, E3) and outperformed LRU, LFU, and FIFO.
- In computation offloading, DDQN with FL achieved near centralized DDQN performance and was superior to mobile execution, edge node execution, and greedy execution for three UEs (m1, m2, m3).
- Centralized DDQN performed better at the beginning of training, but after several FL model merges, DDQN with FL became near to centralized DDQN.
- FL clients upload only model updates, reducing total wireless transmission compared with centralized DRL, where UEs must upload the whole training data.
Methods
- Two simulations were run: edge caching and computation offloading. Edge caching used Xender’s one-month trace (01/09/2016 to 30/09/2016) with 9,514 active mobile users, 188,447 content files, and 2,107,100 content requests, fitted by a Zipf distribution with ∂ = 1.58; 6 edge nodes cooperated, coordinated by a cloud server.
- Computation offloading assumed 5 MHz bandwidth divided into 10 wireless channels, 10 UEs as FL clients, channel gain quantized into 6 levels, and tasks generated by Bernoulli distribution with average arrival rate λ_T per time epoch.
- DRL used Double DQN with tanh activation and Adam optimizer; a single-layer fully-connected feedforward neural network with 200 neurons served as MainNet and TargetNet. Parameters: replay memory M = 5000, minibatch B = 200, discount γ = 0.9, exploration ε = 0.001, learning rate ζ = 0.005, target replacement period φ = 250. A centralized DRL baseline was assumed to receive all data.
Limitations
- The paper states FL cannot outperform centralized DL, only achieve near-best performance, and that FL trades advantages for heavier client computation load and more energy cost or burden on UEs and edge nodes; these remain open questions.
- The paper states privacy and security are not its focus, and that In-Edge AI is future-oriented: state-of-the-art edge chips support only elementary DL training, deploying a randomly initialized DRL model could paralyze the MEC system, and delay requirements remain challenging; transfer learning is proposed to alleviate training.
- The paper states the training-process comparison assumes an ideal wireless environment where massive centralized training data and lightweight FL updates can be uploaded within a discretized time epoch, whereas in practice massive training data cannot be uploaded without delay. It also assumes slowly changing content popularity, equal content sizes, and static popularity, user preference, and average request arrival rates over a relatively long period. Evident from the text: evaluation is simulation-based proof-of-concept, not a real-world deployment.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 of 25 cited sources checked · 0 supported · 0 flagged
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“On the perspective of edge caching, E 1 , E 2 and E 3 are chosen to cache contents, and hit rates of them are improving and finally maintain within a certain range along with the process of training DRL agents.”
Strength:plain· not yet calibratedCausal:correlational· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes? Load-bearing: unclear✓ Based: passesA result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Without FL framework, viz., using centralized DRL, UEs must upload the whole training data via wireless channels and thus consume more communication resources, which is demonstrated by Fig. 6(d).”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passes- Rests on an assumption: In the framework of FL, every client only needs to upload the update of its model.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We perform experiments on investigating the scenarios of edge caching and computation offloading in mobile edge system, and the 'In-Edge AI' is evaluated and proved to have the abilities to achieve near-optimal performance.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIn the paper
“We capture Xender's trace for one month (from 01/09/2016 to 30/09/2016), including 9,514 active mobile users, conveying 188,447 content files, and 2,107,100 content requests [13], of which the content popularity distribution can be well fitted by a Zipf distribution with ∂ = 1 . 58 .”
What the sources say
- Not matchedcites [13] [13] X. Li, X. Wang, P.-J. Wan, Z. Han, V. C.M. Leung, 'Hierarchical Edge Caching in Device-to-Device Aided Mobile Networks: Modeling, Optimization, and Design', IEEE Journal on Selected Areas in Communications , Special Issue on Caching for Communication Systems and Networks, Early Access, 2018. -
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Strength:boosted· not yet calibratedCausal:correlational· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes? Load-bearing: unclear✓ Based: passes- Not matched
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, once the model merging of FL has been processed several times, the performance of DDQN with FL becomes near to that of centralized DDQN.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesA result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“As illustrated in Fig. 6(c), the performance of centralized DDQN is better than that of DDQN with FL at the beginning of training.”
Strength:plain· not yet calibratedCausal:correlational· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes? Load-bearing: unclear✓ Based: passesA conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In a word, the problem existed in the resource allocation optimization of the MEC system is ' lack of intelligence '.”
Load-bearingStrength:plain· not yet calibratedTests✓ Asserted: passes? Checkable: unclear✓ Load-bearing: passes✓ Based: passesIn the paper
“2) reacting cognitively to the mobile communication environment and conditions of cellular networks”
What the sources say
- Not checked yetcites [11] [11] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, 'Communication-Efficient Learning of Deep Networks from Decentralized Data,' in Proceedings of the International Conference on Artificial Intelligence and Statistics , Apr. 2017. - source
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIn the paper
“3) adapting well with heterogeneous UEs in a practical cellular network”
What the sources say
- Not checked yetcites [11] [11] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, 'Communication-Efficient Learning of Deep Networks from Decentralized Data,' in Proceedings of the International Conference on Artificial Intelligence and Statistics , Apr. 2017. - source
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesNone of the sources cited for this claim have been checked yet.
In the paper
“2) Dynamic Conditions: dynamics of the integrated communication and computation system are not well addressed;”
What the sources say
- Not checked yetcites [3] [3] M. Chen and Y. Hao, 'Task Offloading for Mobile Edge Computing in Software Defined Ultra-Dense Network,' in IEEE J. Sel. Areas Commun. , vol. 36, no. 3, pp. 587-597, Mar. 2018 -
- Not checked yetcites [4] [4] Y. Mao, C. You, J. Zhang, K. Huang, K. B. Letaief, 'A Survey on Mobile Edge Computing: The Communication Perspective', in IEEE Commun. Surv. Tutorials , vol. 19, no.4, pp. 2322-2358, Aug. 2017. - source
Load-bearingStrength:plain· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passes
- It may state two separate claims as one. 25
- Probably not a substantial claim. 21
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“'In-Edge AI' is evaluated and proved to have near-optimal performance but relatively low overhead of learning, while the system is cognitive and adaptive to the mobile communication systems.”
- Also stated in: § I. INTRODUCTION · p. 2 · plain
“we provide proof-of-concept evaluation and verify that the proposed scheme has advantages on the balance of performance and cost.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
- Also stated in: § I. INTRODUCTION · p. 2 · plain
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“It can be easily seen that the specific performance of FL is near close to the results of centralized DDQN in terms of achieved hit rates of 3 clients (edge nodes), and outperform LRU, LFU, and FIFO.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In the simulation of computation offloading, the average utilities of three UEs m 1 , m 2 and m 3 are increasing with the decreasing of training loss, and also maintain within a certain range at last.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In Fig. 6(b), it can be observed that the DDQN with FL achieves the near performance of centralized DDQN and is superior to another three baseline policies;”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Certainly, due to the limitation of federated optimization, FL could not outperform the centralized DL trained model, but could still achieve the near best performance, which will demonstrated in Section IV.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
In the paper
“edge nodes, i.e., base stations equipped with computation/storage capability, could deal with the computation and content requests of UEs, and consequently this scheme improves the Quality-of-Service (QoS) of Mobile Network Operators (MNOs) and the Quality-ofExperience (QoE) of UEs and relieves the load of backbone networks, and pressure of clouds (data centers) [2].”
What the sources say
- Not checked yetcites [2] [2] X. Wang, M. Chen, T. Taleb, A. Ksentini, V. C. M. Leung, 'Cache in the Air: Exploiting Content Caching and Delivery Techniques for 5G Systems,' in IEEE Communications , vol. 52, no. 2, pp. 131-139, Feb. 2014. -
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
In the paper
“Federated Learning [11] is introduced as a framework for training DRL agents in a distributed manner while 1) largely reducing the amount of data which should be uploaded via the wireless uplink channel”
What the sources say
- Not checked yetcites [11] [11] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, 'Communication-Efficient Learning of Deep Networks from Decentralized Data,' in Proceedings of the International Conference on Artificial Intelligence and Statistics , Apr. 2017. - source
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“MEC circumvents the long propagation delays introduced by transmitting data from mobile devices to remote cloud computing infrastructures, and hence is able to support latency-critical mobile and Internet of Things (IoT) applications.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“3) Temporal Isolation: most of them do not consider the longterm effect of current decision on resource allocation except for Lyapunov optimization, viz., in a highly time-varying MEC system, most of proposed optimization algorithms is optimal or close-to-optimal only for a snapshot of the system.”
What the sources say
- Not checked yetcites [3] [3] M. Chen and Y. Hao, 'Task Offloading for Mobile Edge Computing in Software Defined Ultra-Dense Network,' in IEEE J. Sel. Areas Commun. , vol. 36, no. 3, pp. 587-597, Mar. 2018 -
- Not checked yetcites [4] [4] Y. Mao, C. You, J. Zhang, K. Huang, K. B. Letaief, 'A Survey on Mobile Edge Computing: The Communication Perspective', in IEEE Commun. Surv. Tutorials , vol. 19, no.4, pp. 2322-2358, Aug. 2017. - source
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
In the paper
“But [7] is based on Q-Learning, which is not feasible for the practical MEC system where the stateaction space is tremendous”
What the sources say
- Not checked yetcites [7] [7] A. Sadeghi, F. Sheikholeslami, and G. B. Giannakis, 'Optimal and Scalable Caching for 5G Using Reinforcement Learning of Space-Time Popularities,' in IEEE J. Sel. Top. Signal Process. , vol. 12, no. 1, pp. 180-190, Feb. 2018. - source
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
In the paper
“the emerging idea of Mobile Edge Computing (MEC) is proposed [1] as a novel paradigm for easing the burden of backbone networks by pushing the computation/storage resources to the proximity of the User Equipments (UEs).”
What the sources say
- Not matchedcites [1] [1] ETSI, 'Mobile Edge Computing - Introductory Technical White Paper', Sep. 2014. -
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesProbably not a substantial claim.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“none of these works has thought over 1) in what form the training data shall be gathered (whether in a distributed or a centralized way), 2) where the reinforcement learning agent should be placed and trained (whether in UEs, edge nodes or remote cloud infrastructures), 3) how the update process of reinforcement learning agents should be proceeded and collaborated, and 4) the privacy protection of”
Load-bearingStrength:plain· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes? Based: unclearIt may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“a recent new trend is to optimize wireless communication by Artificial Intelligence (AI) techniques [4] [5], include but not limited to the application of AI for Physical Layer (PHY), Data Link Layer, and traffic control [6].”
What the sources say
- Not checked yetcites [4] [4] Y. Mao, C. You, J. Zhang, K. Huang, K. B. Letaief, 'A Survey on Mobile Edge Computing: The Communication Perspective', in IEEE Commun. Surv. Tutorials , vol. 19, no.4, pp. 2322-2358, Aug. 2017. - source
- Not checked yetcites [5] [5] Q. Mao, F. Hu, and Q. Hao, 'Deep Learning for Intelligent Wireless Networks: A Comprehensive Survey,' in IEEE Commun. Surv. Tutorials , Early Access, 2018. - source
- Not checked yetcites [6] [6] F. Tang, B. Mao, Z. M. Fadlullah, N. Kato, O. Akashi, T. Inoue, and K. Mizutani, 'On Removing Routing Protocol from Future Wireless Networks: A Real-Time Deep Learning Approach for Intelligent Traffic Control,' in IEEE Wireless Communications , vol. 25, no. 1, pp. 154-160, Feb. 2018. - source
Strength:plain· not yet calibratedCausal:direct· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✓ Based: passesProbably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“the focus of [8] is on vehicular networks.”
What the sources say
- Not checked yetcites [8] [8] Y. He, N. Zhao, and H. Yin, 'Integrated Networking, Caching, and Computing for Connected Vehicles: A Deep Reinforcement Learning Approach,' IEEE Trans. Veh. Technol. , vol. 67, no. 1, pp. 44-55, Jan. 2018. -
Load-bearingStrength:plain· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesProbably not a substantial claim.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“1) the computation capability of a UE is relatively weak, and it will consume long time or is even impossible to train the DRL agent on large-scale data;”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
- Rests on an assumption: The computation capability of a UE is relatively weak.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This will cause more energy cost on UEs or burden the computation of both UEs and edge nodes, and these issues are still open questions.”
Strength:plain· not yet calibratedCausal:direct· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✓ Based: passesProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Interestingly, this is the reverse direction of the aforementioned Federated Learning, but also a kind of application of the 'In-Edge AI'.”
Strength:plain· not yet calibratedTests✓ Asserted: passes✗ Checkable: fails✗ Load-bearing: fails✗ Based: failsProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“3) if training data is transformed for privacy consideration, server-side proxy data is less relevant than on-device data.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. The paper may give no reason why it holds.
- Rests on an assumption: If training data is transformed for privacy, server-side proxy data is less relevant than on-device data.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Obviously, the DRL model should not be deployed directly while setting weights of neural networks at random. Otherwise, the MEC system will be paralyzed because the DRL model could only make random decisions at preliminary exploration.”
Strength:hedged· not yet calibratedCausal:direct· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✓ Based: passesProbably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“By incorporating this work, the requirement on computation capacity of UEs could be relaxed along with decreasing the inferring delay.”
What the sources say
- Not checked yetcites [14] [14] E. Li, Z. Zhou, X. Chen, 'Edge Intelligence: On-Demand Deep Learning Model Co-Inference with Device-Edge Synergy,' in ACM SIGCOMM, MECOMM Workshop , 2018. -
Strength:hedged· not yet calibratedCausal:direct· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes? Load-bearing: unclear✓ Based: passesIt may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Nevertheless, using additional computation could decrease the consumption of communication rounds needed to train a model. In addition, FL only asks a part of clients, in one round, to upload their updates, which handles the circumstance where clients are often unpredictably off-line.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
- Rests on an assumption: UEs are often unpredictably off-line or allocated with poor communication resources.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“each edge node can support AI tasks on system level dynamically, not only for its own but for the global optimization and balancing of the whole MEC system.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. The paper may give no reason why it holds.
In the paper
“However, most distributed DRL architectures, described by Fig. 4(b), could not handle unbalanced and non-IID data and cope with the privacy issues [11].”
What the sources say
- Not checked yetcites [11] [11] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, 'Communication-Efficient Learning of Deep Networks from Decentralized Data,' in Proceedings of the International Conference on Artificial Intelligence and Statistics , Apr. 2017. - source
Load-bearingStrength:hedged· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“With the aid of FL framework, edge nodes equipped with AI computation in 'edge' could combine the cloud and massive UEs together and form a powerful AI entity with strong cognitive ability provided by massive UEs and edge nodes.”
Tests✓ Asserted: passes✗ Checkable: fails✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. It may be too vague to check against evidence. The paper may give no reason why it holds.
In the paper
“The information needs to be uploaded for FL is the minimal update necessary to improve the DRL agent. Further, techniques of secure aggregation and differential privacy could be applied naturally, which could avoid that privacy-sensitive data are contained in local updates.”
What the sources say
- Not checked yetcites [11] [11] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, 'Communication-Efficient Learning of Deep Networks from Decentralized Data,' in Proceedings of the International Conference on Artificial Intelligence and Statistics , Apr. 2017. - source
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✓ Based: passesIt may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“2) the training data which should be uploaded to edge nodes or cloud is privacy-sensitive, and it might cause potential privacy accidents;”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
- Rests on an assumption: Training data that should be uploaded to edge nodes or cloud is privacy-sensitive.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Based on the pre-established DRL, the requirement of training (not inferring) is alleviated, i.e., the DRL agent could reach the satisfied accuracy level after a small number of mini-batches or even directly inferring.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
- Rests on an assumption: Learning takes long time of training as well as inferring according to the required accuracy level.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“And when the wireless environment or the pattern of UEs' requests is changed, the UE could also take advantage of the established DRL to adjust neural networks in its DRL model.”
Strength:hedged· not yet calibratedCausal:conditional· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✓ Based: passesProbably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Specifically, powerful (both in computation and energy) UEs could decide to perform more mini-batches in training for further decreasing the communication cost and vice versa.”
Strength:hedged· not yet calibratedCausal:direct· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✓ Based: passesProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Rests on an assumption: In FL, additional computation could be used to decrease the number of communication rounds to train a model.
- Rests on an assumption: An effective way to add computation is increasing computation per UE, i.e., adding more local SGD updates per round.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Consequently, the performance of In-Edge AI shall not be easily affected by the unbalanced data and sometimes the poor communication environment.”
Load-bearingStrength:plain· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
- Rests on an assumption: FL inherently could address key issues about the availability of UEs and edge nodes and the unbalanced and non-IID data.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, if using transfer learning, there shall be less time or computation consumption on the side of training.”
Strength:plain· not yet calibratedCausal:conditional· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✗ Based: failsProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
- Rests on an assumption: Learning takes long time of training as well as inferring according to the required accuracy level.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Nonetheless, this may be solved if the DL model is not trained from scratch.”
Strength:hedged· not yet calibratedCausal:conditional· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✗ Based: failsProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In-Edge AI is a tight confederation among mobile operators, SPs/CPs, and mobile users, while capable entities may contribute more to the overall optimization of the edge caching and computing tasks, but a large number of UEs may rely on the AI of edge nodes and other UEs.”
Strength:hedged· not yet calibratedCausal:conditional· not yet calibratedTests✓ Asserted: passes✗ Checkable: fails✗ Load-bearing: fails✗ Based: failsProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“DRL can be treated as a type of specialized edge computing task at the service-level.”
Strength:plain· not yet calibratedTests✓ Asserted: passes✗ Checkable: fails✗ Load-bearing: fails✗ Based: failsProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Then, the established DRL model could be distributed to mobile UEs.”
Strength:hedged· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✓ Based: passesProbably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Further, its ability of handling non-IDD data allows massive UEs in different wireless environments to train their own DL model without concerning about the negative effect.”
Strength:plain· not yet calibratedCausal:direct· not yet calibratedTests✓ Asserted: passes✓ Checkable: passes✗ Load-bearing: fails✗ Based: failsProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
- Rests on an assumption: FL inherently could address key issues about the availability of UEs and edge nodes and the unbalanced and non-IID data.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Specifically, due to the fact that the coordinating server in FL only executes tasks of merging updates instead of taking over the whole training, the computation load of clients is inevitably heavier on account of the local training process.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes✗ Based: failsProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“though keeping and training a DRL agent in every UE and edge node can achieve the best performance, it is only practical to apply distributed DRL in the MEC system, because there shall not have enough time and data for training.”
Tests✓ Asserted: passes✓ Checkable: passes✓ Load-bearing: passes? Based: unclearIt may state two separate claims as one.
- Rests on an assumption: There is not enough time and data for training a DRL agent in every UE and edge node.
The paper relies on this without support.
A typical 5G node is expected to have 2000 or more configurable parameters.
“a typical 5G node is expected to have 2000 or more configurable parameters”
The authors state this as an assumption.
Content popularity changes slowly and all contents have the same size.
“we assume that the content popularity changes slowly and all contents have the same size.”
The authors state this as an assumption.
All content popularity, user preference and average arrival rate of requests are static during a relatively long period.
“We assume that all content popularity, user preference and average arrival rate of requests are static during a relatively long period.”
The authors state this as an assumption.
Edge node computing performance is much stronger than the UE's, so f_E is much greater than f_L.
“the formula f E ≫ f L holds (the computing performance of edge nodes is much stronger than the UE's).”
The paper relies on this without support.
Training data is large in quantity when considering massive UEs.
“the training data is large in quantity when considering massive UEs”
The paper relies on this without support.
Training data that should be uploaded to edge nodes or cloud is privacy-sensitive.
“the training data which should be uploaded to edge nodes or cloud is privacy-sensitive”
The paper relies on this without support.
If training data is transformed for privacy, server-side proxy data is less relevant than on-device data.
“if training data is transformed for privacy consideration, server-side proxy data is less relevant than on-device data.”
The paper relies on this without support.
The computation capability of a UE is relatively weak.
“the computation capability of a UE is relatively weak”
The paper relies on this without support.
The training process of a DRL agent may introduce extra energy consumption of a UE.
“the training process of a DRL agent may introduce extra energy consumption of a UE.”
The paper relies on this without support.
DRL techniques require intensive computation capacity for finding optimal solutions.
“DRL techniques require intensive computation capacity for finding the optimal solutions.”
The paper relies on this without support.
There is not enough time and data for training a DRL agent in every UE and edge node.
“there shall not have enough time and data for training.”
The paper relies on this without support.
UEs are often unpredictably off-line or allocated with poor communication resources.
“UEs are often unpredictably off-line or allocated with poor communication resources.”
The paper relies on this without support.
Some UEs may have more computation tasks and some may experience more states of mobile networks, resulting in varying amounts of training data among UEs.
“Some UEs may have more computation tasks to be handled and some may experience more states of mobile networks, resulting in varying amounts of training data among UEs.”
The paper relies on this without support.
FL distributes the quality of knowledge across a large number of devices without necessarily centralizing the training data.
“The core benefit of FL lies in distributing the quality of knowledge across a large number of devices without necessarily centralizing the training data.”
The paper relies on this without support.
Server-side proxy data is less relevant than on-device data.
“Server-side proxy data is less relevant than on-device data.”
The paper relies on this without support.
In the MEC system, massive UEs and edge nodes could act as perceptrons and acquire various, abundant and personalized data for updating the global DL model.
“In the MEC system, massive UEs and edge nodes could act as perceptrons and acquire various, abundant and personalized data for updating the global DL model.”
The paper relies on this without support.
FL inherently could address key issues about the availability of UEs and edge nodes and the unbalanced and non-IID data.
“FL inherently could address key issues about the availability of UEs and edge nodes and the unbalanced and non-IID data.”
The paper relies on this without support.
In FL, additional computation could be used to decrease the number of communication rounds to train a model.
“In FL, additional computation could be used to decrease the number of communication rounds to train a model.”
The paper relies on this without support.
An effective way to add computation is increasing computation per UE, i.e., adding more local SGD updates per round.
“An effective way to add computation is increasing computation per UE, i.e., adding more local SGD updates per round.”
The paper relies on this without support.
Learning takes long time of training as well as inferring according to the required accuracy level.
“Learning takes long time of training as well as inferring according to the required accuracy level.”
The paper relies on this without support.
In practice, massive training data are actually unable to be uploaded without any delay.
“In practice, massive training data are actually unable to be uploaded without any delay.”
The paper relies on this without support.
In the framework of FL, every client only needs to upload the update of its model.
“In the framework of FL, every client only needs to upload the update of its model.”
The paper relies on this without support.
In 5G, URLLC scenarios strongly desire very small delays and high reliability of the mobile system.
“In 5G, the defined Ultra-Reliable Low-Latency Communications (URLLC) scenarios strongly desire very small delays and high reliability of the mobile system.”
The paper relies on this without support.
General deep learning-based optimization and prediction schemes take quite long running time of recursions for converging to the results, which is inappropriate for mobile edge systems.
“However, general deep learning-based optimization and prediction schemes take quite long running time of recursions for converging to the results, which is inappropriate for mobile edge systems; particularly the system-level edge computing tasks desiring rapid responses in the scale of millisecond.”
The authors state this as an assumption.
The experiment assumes an ideal wireless environment.
“Nonetheless, this experiment assumes an ideal wireless environment.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Again, due to FL’s properties, imbalanced and non-IID information can be handled appropriately, ensuring that the AI system’s performance is not limited and that a large device in multiple contexts may be trained with their local data and local model without difficulty [211].”
From Federated learning-based AI approaches in smart healthcare: Concepts, taxonomies, challenges and open issues (2023) · cites this paper as [211]
In the citing paper
“In [9] an adaptive control scheme was proposed to adapt the global FL aggregation frequency.”
From Federated Learning for Edge Networks: Resource Optimization and Incentive Mechanism (2020) · cites this paper as [9]
In the citing paper
“This adaptive control scheme offers a desirable tradeoff between global model aggregation and local model update to minimize the loss function with resource budget constraint.”
From Federated Learning for Edge Networks: Resource Optimization and Incentive Mechanism (2020) · cites this paper as [9]
Citations
Cites
Human-level control through deep reinforcement learning
Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., Graves, A., Riedmiller, M., Fidjeland, A. K., Ostrovski, G., et al. (2015) · Nature · doi:10.1038/nature14236 · Cited by 30,357 · cited 4×
“By taking advantage of Deep Q-Learning and its improved version, e.g., Double DQN [12], the control policy Φ = (Φc(Υ),Φe(Υ)) could be trained and achieves increasing the utility of UEs for the long-term performance optimization.”
BackgroundInfluentialReinforcement Learning: An Introduction
Sutton, R. S., & Barto, A. (2005) · IEEE Transactions on Neural Networks · doi:10.1109/tnn.2004.842673 · Cited by 25,268
Communication-Efficient Learning of Deep Networks from Decentralized Data
McMahan, H. B., Moore, E., Ramage, D., Hampson, S., & Arcas, B. A. Y. (2016) · arXiv · doi:10.48550/arxiv.1602.05629 · Cited by 14,215 · cited 14×
“Nonetheless, the privacy and security is not our focus in this work, more information about these issues could be found in references of [11] A. Integration of Federated Learning within Edge for In-Edge AI Between AI in cloud and AI in UEs, there is an “edge”.”
BackgroundMethodInfluentialReinforcement Learning: An Introduction
Johnson, J. D., Li, J., & Chen, Z. (2000) · Neurocomputing · doi:10.1016/s0925-2312(00)00324-6 · Cited by 8,825
A Survey on Mobile Edge Computing: The Communication Perspective
Mao, Y., You, C., Zhang, J., Huang, K., & Letaief, K. B. (2017) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2017.2745201 · Cited by 5,607 · cited 2×
“Several pioneer works have been proposed and realize quite good results in their assuming settings based on convex optimization, game theory and so on [3] [4] .”
BackgroundMethodDeep Reinforcement Learning with Double Q-Learning
Van Hasselt, H., Guez, A., & Silver, D. (2016) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v30i1.10295 · Cited by 3,318 · cited 13×
“However, once the model merging of FL has been processed several times, the performance of DDQN with FL becomes near to that of centralized DDQN.”
MethodInfluentialTask Offloading for Mobile Edge Computing in Software Defined Ultra-Dense Network
Chen, M., & Hao, Y. (2018) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2018.2815360 · Cited by 1,085 · cited 1×
“Several pioneer works have been proposed and realize quite good results in their assuming settings based on convex optimization, game theory and so on [3] [4] .”
BackgroundSelf-citationCache in the air: exploiting content caching and delivery techniques for 5G systems
Wang, X., Chen, M., Taleb, T., Ksentini, A., & Leung, V. (2014) · IEEE Communications Magazine · doi:10.1109/mcom.2014.6736753 · Cited by 1,063 · cited 1×
“, base stations equipped with computation/storage capability, could deal with the computation and content requests of UEs, and consequently this scheme improves the Quality-of-Service (QoS) of Mobile Network Operators (MNOs) and the Quality-ofExperience (QoE) of UEs and relieves the load of backbone networks, and pres…”
BackgroundSelf-citationDeep Learning for Intelligent Wireless Networks: A Comprehensive Survey
Mao, Q., Hu, F., & Hao, Q. (2018) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2018.2846401 · Cited by 789 · cited 1×
“, a typical 5G node is expected to have 2000 or more configurable parameters, a recent new trend is to optimize wireless communication by Artificial Intelligence (AI) techniques [4] [5], include but not limited to the application of AI for Physical Layer (PHY), Data Link Layer, and traffic control [6].”
MethodWhen Mobile Blockchain Meets Edge Computing
Xiong, Z., Zhang, Y., Niyato, D. T., Wang, P., & Han, Z. (2018) · IEEE Communications Magazine · doi:10.1109/mcom.2018.1701095 · Cited by 570
Flagged first, then by citations
Research works printed but not linked: 1
- 2014, Mobile Edge Computing – Introductory Technical White Paper
Cited by
Sorted by citations
Items 301–400 of 1,000
Wu, Y. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.3018691 · Cited by 38
Yuan, M., Zhang, L., He, F., Tong, X., & Li, X. (2022) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3495243.3517016 · Cited by 40
Artificial Intelligence as a Service (AIaaS) for Cloud, Fog and the Edge: State-of-the-Art Practices
Syed, N., Anwar, A., Baig, Z. A., & Zeadally, S. (2025) · ACM Computing Surveys · doi:10.1145/3712016 · Cited by 40
Towards Mobile Edge Computing: Taxonomy, Challenges, Applications and Future Realms
Qadir, J., de Abajo, B. S., Khan, A., García-Zapirain, B., Díez, I. D. L. T., & Mahmood, H. (2020) · IEEE Access · doi:10.1109/access.2020.3026938 · Cited by 39
“Wang and colleagues proposed an intelligentizing mobile edge computing (In-Edge AI) framework to utilize the collaboration of the devices and edge nodes intelligently [120].”
MethodZhou, S., Jadoon, W., & Khan, I. A. (2023) · Electronics · doi:10.3390/electronics12112452 · Cited by 37
Narula, M., Meena, J., & Vishwakarma, D. K. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108128 · Cited by 38
Security, privacy, and robustness for trustworthy AI systems: A review
Saeed, M. M., & Alsharidah, M. (2024) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2024.109643 · Cited by 45
True-data testbed for 5G/B5G intelligent network
Huang, Y., Liu, S., Zhang, C., You, X., & Wu, H. (2021) · Intelligent and Converged Networks · doi:10.23919/icn.2021.0002 · Cited by 36
Enabling AI in future wireless networks: a data life cycle perspective
Nguyen, D. C., Cheng, P., Ding, M., López‐Pérez, D., Pathirana, P. N., Li, J., Seneviratne, A. P., Li, Y., & Poor, H. V. (2024) · Figshare · Cited by 36
Zhang, L., Du, Q., Lu, L., & Zhang, S. (2023) · Electronics · doi:10.3390/electronics12173651 · Cited by 35
Storage-Aware Joint User Scheduling and Bandwidth Allocation for Federated Edge Learning
Liu, S., Shen, Y., Yuan, J., Wu, C., & Yin, R. (2024) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2024.3451711 · Cited by 40
“To utilize these data in an efficient manner, a new computing paradigm, i.e., Edge Learning (EL), has been proposed by migrating the training procedure from central cloud servers to the edge of wireless networks [9], [10], [11].”
On the Performance of Federated Learning Algorithms for IoT
Tahir, M., & Ali, M. I. (2022) · IoT · doi:10.3390/iot3020016 · Cited by 35
“For example, [17] presented an adaptive averaging strategy, and authors in [18] presented an In-Edge AI framework to tackle the communication bottleneck in federated networks.”
FedSup: A communication-efficient federated learning fatigue driving behaviors supervision approach
Zhao, C., Gao, Z., Wang, Q., Xiao, K., Mo, Z., & Deen, M. J. (2022) · Future Generation Computer Systems · doi:10.1016/j.future.2022.08.009 · Cited by 39
Customized Federated Learning for accelerated edge computing with heterogeneous task targets
Jiang, H., Liu, M., Yang, B., Liu, Q., Li, J., & Guo, X. (2020) · Computer Networks · doi:10.1016/j.comnet.2020.107569 · Cited by 35
Mutalemwa, L. C., & Shin, S. (2020) · IEEE Access · doi:10.1109/access.2020.3037357 · Cited by 34
“Traditional DRL frameworks usually employ centralized learning schemes with a single caching agent and they are more suited for small scale network systems [30], [85], [86].”
BackgroundMethodInfluentialZhou, C., Tian, H., Zhang, H., Zhang, J., Dong, M., & Jia, J. (2021) · ACM International Conference on Computing Frontiers (CF) · doi:10.1145/3457388.3458655 · Cited by 34
“At the same time, due to the similar concerns, machine learning is also sinking from the cloud to the edge and is closely integrated with edge computing, such as keyboard prediction[3], video surveillance[15], edge intelligence[16] and so on.”
BackgroundWhen machine learning meets Network Management and Orchestration in Edge-based networking paradigms
Shahraki, A., Ohlenforst, T., & Kreyß, F. (2022) · Journal of Network and Computer Applications · doi:10.1016/j.jnca.2022.103558 · Cited by 35
Distributed Intelligence in Wireless Networks
Liu, X., Yu, J., Liu, Y., Gao, Y., Mahmoodi, T., Lambotharan, S., & Tsang, D. H. K. (2023) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2023.3265425 · Cited by 41
“In[140], the authors first applied the FL framework to train the DRL agents for intelligent joint resource management of communication and computation in MEC systems.”
MethodSun, Y., Wei, T., Li, H., Zhang, Y., & Wu, W. (2020) · IEEE Access · doi:10.1109/access.2020.2973359 · Cited by 33
“First, we present the system stability in Fig.3 with the control parameter V = [10 11 , 5 ∗ 10 11 , 10 12 , 10 13 ] and task average arrival rate ¯ λ = 2Mbps.”
MethodAuxiliary Diagnosis of COVID-19 Based on 5G-Enabled Federated Learning
Wang, R., Xu, J., Ma, Y., Talha, M., Al-Rakhami, M. S., & Ghoneim, A. (2021) · IEEE Network · doi:10.1109/mnet.011.2000704 · Cited by 33
“This article is based on AI technology to train the model with multi-source heterogeneous data used for the implementation and deployment of federated learning at the same time [13].”
MethodEICache: A learning-based intelligent caching strategy in mobile edge computing
Tang, B., & Kang, L. (2021) · Peer-to-Peer Networking and Applications · doi:10.1007/s12083-021-01266-4 · Cited by 33
“Wang et al. [64] applied the ideas of deep reinforcement learning and joint learning to the joint optimization of computing, caching, and communication resources in mobile edge networks.”
BackgroundSmooth and Low Latency Video Streaming for Autonomous Cars During Handover
Marai, O. E., & Taleb, T. (2020) · IEEE Network · doi:10.1109/mnet.011.2000258 · Cited by 33
“consuming traffic due to the high size of the video frames even after compression [5].”
BackgroundOptimized Multi-Service Tasks Offloading for Federated Learning in Edge Virtualization
Tam, P., Math, S., & Kim, S. (2022) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2022.3200057 · Cited by 33
“To secure local data privacy and support distributed on-device ML/DL model builders, federated learning (FL) framework is applicable by modifying the conventional raw data uploading into collaborative local-global model updates [5], [6], [7].”
MethodIncentive-Aware Blockchain-Assisted Intelligent Edge Caching and Computation Offloading for IoT
Wang, Q., Chen, S., & Wu, M. C. (2023) · Engineering · doi:10.1016/j.eng.2022.10.014 · Cited by 33
Incentive Mechanism Design for Resource Sharing in Collaborative Edge LearningPreprint
Lim, W. Y. B., Ng, J. S., Xiong, Z., Niyato, D. T., Leung, C., Miao, C., & Yang, Q. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.00511 · Cited by 36
“However, given the growing complexity and capacity of future networks, the dynamic conditions in the network cannot be completely modeled or solved with standard approaches [12].”
A Blockchain-Driven IIoT Traffic Classification Service for Edge Computing
Qi, H., Wang, J., Li, W., Wang, Y., & Qiu, T. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.3035431 · Cited by 32
Majidi, F., Khayyambashi, M. R., & Barekatain, B. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3086623 · Cited by 32
Albaseer, A. M., Abdallah, M. M., Al‐Fuqaha, A., Erbad, A., & Dobre, O. A. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3194833 · Cited by 32
Distributed Resource Management for Blockchain in Fog-Enabled IoT Networks
Yang, L., Li, M., Zhang, H., Ji, H., Xiao, M., & Li, X. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.3028071 · Cited by 31
“Here, the system is composed of a large amount of IoT devices, fog nodes (FNs) with edge computing capabilities, and one primary FN (PFN) with storage capability [11], [12].”
Rahbari, D., Alam, M. M., Moullec, Y. L., & Jenihhin, M. (2021) · IEEE Access · doi:10.1109/access.2021.3104117 · Cited by 31
“In the scientific literature, FL has been combined with DRL; the resulting federated deep reinforcement learning (FDRL) is a distributed version of DRL that can be used in MEC [22].”
MethodChen, S., Wang, X., Zhou, P., Wu, W., Lin, W., & Wang, Z. (2022) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2022.3146871 · Cited by 31
“For example, [28] uses FL to train deep reinforcement learning agents in a distributed manner.”
MethodSelf-citationJoint Gradient Sparsification and Device Scheduling for Federated Learning
Lin, X., Liu, Y., Chen, F.-J., Ge, X., & Huang, Y. (2023) · IEEE Transactions on Green Communications and Networking · doi:10.1109/tgcn.2023.3239373 · Cited by 30
Data-Driven Participant Selection and Bandwidth Allocation for Heterogeneous Federated Edge Learning
Albaseer, A., Abdallah, M., Al-Fuqaha, A. I., & Erbad, A. (2023) · IEEE Transactions on Systems, Man & Cybernetics. Systems · doi:10.1109/tsmc.2023.3276329 · Cited by 23
Gui, J., Lin, L., Deng, X., & Cai, L. (2024) · IEEE/ACM Transactions on Networking · doi:10.1109/tnet.2024.3364161 · Cited by 30
“Although the distributed trained local model parameters need to be transferred to a centralized server to update the global foundation model parameters, the communication overhead is negligible when compared to aggregating the decentralized raw datasets to the centralized server [29].”
BackgroundISFL: Federated Learning for Non-i.i.d. Data With Local Importance Sampling
Zhu, Z., Shi, Y., Fan, P., Peng, C., & Letaief, K. B. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3398398 · Cited by 30
“In particular, due to the efficient sharing mechanism of FL, its combinations with the reinforcement learning (RL) or multiagent RL (MARL) have been considered, which is referred to as federated RL (FRL) [13].”
BackgroundLifecycle-Aware Online Video Caching
Li, T., Braud, T., Li, Y., & Hui, P. (2020) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2020.2984364 · Cited by 30
Self-aware distributed deep learning framework for heterogeneous IoT edge devices
Jin, Y., Cai, J., Xu, J., Huan, Y., Yan, Y., Huang, B., Guo, Y., Zheng, L., & Zou, Z. (2021) · Future Generation Computer Systems · doi:10.1016/j.future.2021.07.010 · Cited by 30
Somesula, M. K., Rout, R. R., & Somayajulu, D. V. L. N. (2022) · Computer Networks · doi:10.1016/j.comnet.2022.108876 · Cited by 32
Liu, B., Wang, L., & Liu, M. (2019) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros40897.2019.8967908 · Cited by 30
Asynchronous Wireless Federated Learning With Probabilistic Client Selection
Yang, J., Liu, Y., Chen, F., Chen, W., & Li, C. (2023) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2023.3337773 · Cited by 29
“Smartphones, wearable devices, smart homes, and various distributed wireless sensors generate massive amounts of data in real time [1], [2], [3], which can be used by machine learning techniques to bring diverse artificial intelligence (AI) services to people [4], [5], [6].”
BackgroundIntelligent dynamic service pricing strategy for multi-user vehicle-aided MEC networks
Liao, Y., Qiao, X., Yu, Q. G., & Liu, Q. (2020) · Future Generation Computer Systems · doi:10.1016/j.future.2020.07.051 · Cited by 29
Rjoub, G., Wahab, O. A., Bentahar, J., & Bataineh, A. S. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-65310-1_23 · Cited by 29
Energy-efficient client selection in federated learning with heterogeneous data on edge
Zhao, J., Feng, Y., Chang, X., & Liu, C. H. (2022) · Peer-to-Peer Networking and Applications · doi:10.1007/s12083-021-01254-8 · Cited by 29
“Due to these reasons, it has been gaining increasing popularity in various research and application fields, such as vehicle network [4], edge cache optimising and computation offloading [5], health care [6], mobile keyboard prediction [7], etc.”
BackgroundLiu, T., Zhou, H., Li, J., Shu, F., & Han, Z. (2022) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2022.3207916 · Cited by 29
“On one hand, to meet the need on privacy preserving, federated learning (FL), an emerging distributed learning paradigm, is increasingly attractive because of its superior characteristics on privacy preserving [10], [11], [12].”
BackgroundLi, Z., Chen, Z., Wei, X., Gao, S. B., Ren, C., & Quek, T. Q. S. (2022) · IEEE International Conference on Sensing, Communication, and Networking (SECON) · doi:10.1109/secon55815.2022.9918588 · Cited by 30
“For the traditional edge-cloud two-tier structure [6] communication costs and unable to train suitable personalized models.”
BackgroundMethodJazaeri, S. S., Asghari, P., Jabbehdari, S., & Javadi, H. H. S. (2023) · Multimedia Tools and Applications · doi:10.1007/s11042-023-15657-7 · Cited by 29
CB-DSL: Communication-Efficient and Byzantine-Robust Distributed Swarm Learning on Non-i.i.d. Data
Fan, X., Wang, Y., Huo, Y., & Tian, Z. (2023) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2023.3312345 · Cited by 30
“The valuable data directly collected by devices together with the resurgence of machine learning (ML) stimulate the latest trend of artificial intelligence (AI) at the edge of B5G networks, termed as edge learning or edge AI [2], [3].”
BackgroundDistributed artificial intelligence: Taxonomy, review, framework, and reference architecture
Janbi, N. F., Katib, I. A., & Mehmood, R. (2023) · Intelligent Systems with Applications · doi:10.1016/j.iswa.2023.200231 · Cited by 30
ParallelSFL: A Novel Split Federated Learning Framework Tackling Heterogeneity Issues
Liao, Y., Xu, Y., Xu, H., Yao, Z., Huang, L., & Qiao, C. (2024) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3636534.3690665 · Cited by 33
Enhancing the cognitive load theory and multimedia learning framework with AI insight
Twabu, K. Y. (2025) · Discover Education · doi:10.1007/s44217-025-00592-6 · Cited by 28
“These steps help to ensure that AI systems receive only the most relevant information, which improves processing efficiency and reduces the computational burden [35].”
Liu, C., Zhang, H., Ji, H., & Li, X. (2021) · China Communications · doi:10.23919/jcc.2021.02.013 · Cited by 28
Deployment of Artificial Intelligence Models on Edge Devices: A Tutorial Brief
Żyliński, M., Nassibi, A., Rakhmatulin, I., Malik, A., Papavassiliou, C., & Mandic, D. P. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3336831 · Cited by 27
“communications and data compression [2], enhancement of device performance [3], advanced on-chip process control.”
BackgroundMukherjee, M., Guo, M., Lloret, J., & Zhang, Q. (2020) · IEEE Network · doi:10.1109/mnet.001.2000004 · Cited by 25
Deep-Reinforcement-Learning-Based Latency Minimization in Edge Intelligence Over Vehicular Networks
Zhao, N., Wu, H., Yu, F. R., Wang, L., Zhang, W., & Leung, V. C. M. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3078480 · Cited by 27
Edge Network Optimization Based on AI Techniques: A Survey
Pooyandeh, M., & Sohn, I. (2021) · Electronics · doi:10.3390/electronics10222830 · Cited by 33
“In [34], some optimization methods were used to face unsure input, dynamic status, and temporal isolation.”
MethodYuan, Z., Azzino, T., Yu, H., Lyu, Y., Pei, H., Boldini, A., Mezzavilla, M., Beheshti, M., Porfiri, M., Hudson, T. E., et al. (2022) · IEEE Access · doi:10.1109/access.2022.3157876 · Cited by 29
Digital Twin-enabled AI Enhancement in Smart Critical Infrastructures for 5G
Gai, K., Xiao, Q., Qiu, M., Zhang, G., Chen, J., Wei, Y., & Zhang, Y. (2022) · ACM Transactions on Sensor Networks · doi:10.1145/3526195 · Cited by 28
“, production coordination/cooperation, full lifecycle management, parallel composition, and intelligent manufacturing [36, 40].”
BackgroundStability-Based Analysis and Defense against Backdoor Attacks on Edge Computing Services
Zhao, Y., Xu, K., Wang, H., Li, B., & Jia, R. (2021) · IEEE Network · doi:10.1109/mnet.011.2000265 · Cited by 26
“To fully and concisely analyze the backdoor attacks on intelligent edge computing services, we utilize FL [6, 12] to achieve collaborative learning under the edge computing scenario.”
BackgroundMethodInfluentialSun, X., Wang, X., Cai, D., Li, Z., Gao, Y., & Wang, X. (2020) · IEEE Access · doi:10.1109/access.2020.2981528 · Cited by 26
“Through deployment of cloud-like infrastructure near data sources, data can be preferentially handled at the edge, largely relaxing the whole network [12].”
BackgroundSelf-citationSmartphone-based diabetic macula edema screening with an offline artificial intelligence
Hwang, D., Yu, W., Lin, T.-C., Chou, S., Yarmishyn, A. A., Kao, Z.-K., Kao, C.-L., Yang, Y.-P., Chen, S., Hsu, C., et al. (2020) · Journal of the Chinese Medical Association · doi:10.1097/jcma.0000000000000355 · Cited by 25
Spectrum Sharing for Secrecy Performance Enhancement in D2D-Enabled UAV Networks
Yang, B., Taleb, T., Wu, Z., & Ma, L. (2020) · IEEE Network · doi:10.1109/mnet.011.2000093 · Cited by 26
“Thus, it can provide a strong security guarantee for transmitters to prevent the attacks of adversaries for supporting many applications, like covert military communications, location tracking and intercommu-nication of Internet of Things [16]–[19].”
BackgroundAdaptive DNN Partition in Edge Computing Environments
Miao, W., Zeng, Z., Wei, L., Li, S., Jiang, C., & Zhang, Z. (2020) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads51040.2020.00097 · Cited by 29
“As a combination of AI and edge computing, edge intelligence emerges [4].”
BackgroundFederated Learning at the Network Edge: When Not All Nodes Are Created Equal
Malandrino, F., & Chiasserini, C. F. (2021) · IEEE Communications Magazine · doi:10.1109/mcom.001.2001016 · Cited by 26
Ma, X., Zhao, J.-H., & Gong, Y. (2021) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2021.3107465 · Cited by 22
“Fortunately, with the enhancement of computing capability of mobile devices, it is envisioned that most devices are endowed with the ability of inferring and training the AI model in the near future [8].”
BackgroundHao, Q., & Li, T. (2021) · Future Generation Computer Systems · doi:10.1016/j.future.2021.09.031 · Cited by 25
A Codesigned Integrated Photonic Electronic Neuron
Marinis, L. D., Catania, A., Castoldi, P., Contestabile, G., Bruschi, P., Piotto, M., & Andriolli, N. (2022) · IEEE Journal of Quantum Electronics · doi:10.1109/jqe.2022.3177793 · Cited by 27
EIDLS: An Edge-Intelligence-Based Distributed Learning System Over Internet of Things
Wang, T., Sun, B., Wang, L., Zheng, X., & Jia, W. (2023) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2023.3240992 · Cited by 26
“In order to minimize EC, training time, and communication costs, the server needs to determine the appropriate amount of data and energy that the end devices use for training [18].”
BackgroundGonzález, C. C., Pupo, E. F., Iradier, E., Angueira, P., Murroni, M., & Montalbán, J. (2024) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2024.3373638 · Cited by 26
“Mathematically, the local agent’s reward is defined as [38].”
BackgroundCommunication-Efficient and Distributed Learning Over Wireless Networks: Principles and Applications
Park, J., Samarakoon, S., Elgabli, A., Kim, J., Bennis, M., Kim, S.-L., & Debbah, M. (2021) · Proceedings of the IEEE · doi:10.1109/jproc.2021.3055679 · Cited by 25
“Realizing the aforementioned pressing concern has recently sparked huge attention to the introduction of machine learning (ML) based approaches into communication system designs [11]–[14].”
BackgroundAn Intelligent Computation Demand Response Framework for IIoT-MEC Interactive Networks
Liao, Y., Shou, L., Yu, Q. G., Ai, Q., & Liu, Q. (2020) · IEEE Networking Letters · doi:10.1109/lnet.2020.3001178 · Cited by 25
“…techniques that can predict the uncertainty of future computation resource demand of MDs, learning-based strategies such as “edge-AI” proposed in [11] can be investigated for further consideration because this is a capable way to satisfy MDs’ customized resource preference at affordable service…”
BackgroundLuo, P., Yu, F. R., Chen, J., Li, J., & Leung, V. C. M. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3051611 · Cited by 26
A Secure and Privacy-Preserving Machine Learning Model Sharing Scheme for Edge-Enabled IoT
Zhou, X., Xu, K., Wang, N., Jiao, J., Dong, N., Han, M., & Xu, H. (2021) · IEEE Access · doi:10.1109/access.2021.3051945 · Cited by 25
“Driven by this trend, computing power is already shifting from the centralized cloud to edge side, or data source side [3], [4].”
BackgroundJatain, D., Sıngh, V., & Dahiya, N. (2021) · Journal of King Saud University - Computer and Information Sciences · doi:10.1016/j.jksuci.2021.05.016 · Cited by 25
“Other major works include addressing network delay issues in MATCHA (Wang et al., 2019a, 2019b), handling peer-to-peer network (Lalitha et al., 2019) and learning the embeddings with multi-modal representations in secure and efficient manner (Holzinger et al., 2021).”
Collaborative task offloading and resource scheduling framework for heterogeneous edge computing
Ren, J., Hou, T., Wang, H., Tian, H., Wei, H., Zheng, H., & Zhang, X. (2021) · Wireless Networks · doi:10.1007/s11276-021-02768-y · Cited by 26
Chen, Z., Cui, H., Wu, E., & Yu, X. (2022) · Sensors · doi:10.3390/s22020684 · Cited by 25
“Each hidden layer has a certain number of neurons, which have multiple inputs and a single output [11,12].”
BackgroundFederated learning based QoS-aware caching decisions in fog-enabled internet of things networks
Huang, X., Chen, Z., Chen, Q., & Zhang, J. (2022) · Digital Communications and Networks · doi:10.1016/j.dcan.2022.04.022 · Cited by 25
Mishra, K. N., Bhattacharjee, V., Saket, S., & Mishra, S. P. (2022) · Cluster Computing · doi:10.1007/s10586-022-03813-x · Cited by 25
“12 [87, 88].”
Personalized Federated Learning via Domain Adaptation with an Application to Distributed 3D Printing
Shi, N., & Kontar, R. A. (2022) · Technometrics · doi:10.1080/00401706.2022.2157882 · Cited by 25
“This approach has seen success in large scale mobile applications (Yang et al. 2018; Wang et al. 2019b; Lim et al. 2020).”
BackgroundSpace–Air–Ground–Sea Integrated Network with Federated Learning
Zhao, H., Ji, F., Wang, Y., Yao, K., & Chen, F. (2024) · Remote Sensing · doi:10.3390/rs16091640 · Cited by 24
“…deep reinforcement learning and FL were combined together and an “In-Edge AI” framework was proposed to intelligently utilize the collaboration between center and edge nodes, and a dynamic system-level optimization was presented to reduce the unnecessary communication consumption in [46].”
Federated learning for Healthcare 5.0: a comprehensive survey, taxonomy, challenges, and solutions
Amin, M., Ahmad, S., & Loh, W.-K. (2025) · Soft Computing · doi:10.1007/s00500-025-10508-z · Cited by 29
SecureFedPROM: A Zero-Trust Federated Learning Approach With Multi-Criteria Client Selection
Tahir, M., Mawla, T., Awaysheh, F. M., Alawadi, S., Gupta, M., & Ali, M. I. (2025) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2025.3560008 · Cited by 31
“In this evolving landscape, Federated Learning (FL) [3] has become a significant focus, as it enables model training on edge devices and aims for efficiency and quality in globally shared statistical models [4], [5].”
Towards Named AI Networking: Unveiling the Potential of NDN for Edge AI
Campolo, C., Lia, G., Amadeo, M., Ruggeri, G., Iera, A., & Molinaro, A. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-61746-2_2 · Cited by 24
Optimal Accuracy-Time Trade-off for Deep Learning Services in Edge Computing Systems
Hosseinzadeh, M., Wachal, A., Khamfroush, H., & Lucani, D. E. (2021) · International Conference on Communications · doi:10.1109/icc42927.2021.9500744 · Cited by 25
“More recently, EC has been shown to reduce response time for deployed DL services [7]–[9].”
BackgroundJin, W., Xu, R., Lim, S., Park, D.-H., Park, C., & Kim, D. (2021) · Sensors · doi:10.3390/s21020630 · Cited by 28
Smart and Sentient Retail High Streets
Torrens, P. M. (2022) · Smart Cities · doi:10.3390/smartcities5040085 · Cited by 26
Somesula, M. K., Rout, R. R., & Somayajulu, D. V. L. N. (2022) · Ad Hoc Networks · doi:10.1016/j.adhoc.2022.103051 · Cited by 24
Thamilselvan, R., Ekambaram, G., Palanisamy, N., Rajalaxmi, R. R., Muthusamy, S., Loganathan, A. K., Panchal, H. N., Thangaraj, K., & Ashokkumar, R. (2023) · Wireless Personal Communications · doi:10.1007/s11277-023-10450-z · Cited by 24
“The intelligence in the edges of the cloud environment enhance the performance improvement in terms of throughput and delay [5].”
BackgroundGan, D., Ge, X., & Li, Q. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3292368 · Cited by 26
“To solve the above problems, FL and DRL [8] are to build an in-edge AI framework.”
BackgroundMethodInfluentialA Survey on Privacy-Preserving Caching at Network Edge: Classification, Solutions, and Challenges
Zhang, X., Zhou, Y., Wu, D., Sheng, Q. Z., Riaz, S., Hu, M., & Xiao, L. (2024) · ACM Computing Surveys · doi:10.1145/3706630 · Cited by 23
“Kumaret al. [36] Cache Obf.”
Jeon, J., Park, S., Choi, M., Kim, J., Kwon, Y., & Cho, S. (2020) · Electronics · doi:10.3390/electronics9091359 · Cited by 23
“[33] proposed the cost-efficient method of federated learning across edge nodes.”
MethodGreen MEC Networks Design Under UAV Attack: A Deep Reinforcement Learning Approach
Zhao, R., Xia, J., Zhao, Z., Lai, S., Fan, L., & Li, D. (2021) · IEEE Transactions on Green Communications and Networking · doi:10.1109/tgcn.2021.3073939 · Cited by 23
Federated Learning With Heterogeneity-Aware Probabilistic Synchronous Parallel on Edge
Zhao, J., Han, R., Yang, Y., Catterall, B., Liu, C. H., Chen, L. Y., Mortier, R., Crowcroft, J., & Wang, L. (2021) · IEEE Transactions on Services Computing · doi:10.1109/tsc.2021.3109910 · Cited by 24
Multi-round auction-based resource allocation for edge computing: Maximizing social welfare
Xiao, J., Gao, Q., Yang, Z., Cao, Y., Wang, H., & Feng, Z. (2022) · Future Generation Computer Systems · doi:10.1016/j.future.2022.10.028 · Cited by 23
Zeng, X., Yu, Z., Zhang, W., Wang, X., Lu, Q., Wang, T., Gu, M., Tian, Y., & Wang, F. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3228792 · Cited by 23
Self-citationFederated Learning for Autoencoder-based Condition Monitoring in the Industrial Internet of Things
Becker, S., Styp-Rekowski, K., Stoll, O. V. L., & Kao, O. (2022) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata55660.2022.10020836 · Cited by 23
“developed two generalized frameworks, In-Edge AI [23] and EdgeFed [24], for improving FL at the edge.”
MethodA Comprehensive Survey on Software as a Service (SaaS) Transformation for the Automotive Systems
Blanco, D. F., Mouël, F. L., Lin, T., & Escudié, M.-P. (2023) · IEEE Access · doi:10.1109/access.2023.3294256 · Cited by 26
Hasan, T., Ahmad, F., Rizwan, M., Alshammari, N. O., Alanazi, S., Hussain, I., & Naseem, S. (2022) · Computational Intelligence and Neuroscience · doi:10.1155/2022/6138434 · Cited by 23
“*e federated learning (FL) framework and deep reinforcement learning (DRL) techniques were integrated into MEC in [16] to optimize EC, computing, and communication.”
MethodSun, X., Wang, Y., Duan, P., Zia, Q., Wang, W., & Cai, Z. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3358799 · Cited by 22
“Wang et al. [41] proposed the In-Eager Al framework, which aims to integrate federated learning with mobile edge systems, aiming to enhance caching, communication efficiency, and edge computing.”
BackgroundPark, S., & Choi, W. (2023) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2023.3242722 · Cited by 21
“Owing to this inherent analogy, the applicability of FL to wireless communication systems gains increasing traction and leads to active research [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20] in this regard.”
BackgroundAutomated and intelligent systems for next-generation-based smart applications
Deekshetha, H. R., & Tyagi, A. K. (2023) · Elsevier eBooks · doi:10.1016/b978-0-323-98352-5.00019-7 · Cited by 22
Metadata
- Authors
- Wang, X., Han, Y., Wang, C., Zhao, Q., Chen, X., & Chen, M.
- Affiliations
- Tianjin University (CN)
- Huawei Technologies (China) (CN)
- Arizona State University (US)
- Huazhong University of Science and Technology (CN)
- Venue
- IEEE Network
- Published
- 2019-09
- Volume
- 33
- Issue
- 5
- Pages
- 156-165
- ISSN
- 0890-8044 (ISSN-L)
- 1558-156X
- OpenAlex
- W2962804345
- Semantic Scholar
- 6f89a632ceb8fcb81eac3d7b52e937099659cc6a
- Cited by
- 1,030
- References
- 17
- Citations by year
- 2026: 44 · 2025: 95 · 2024: 127 · 2023: 165 · 2022: 215 · 2021: 196 · 2020: 160 · 2019: 26
- FWCI
- 81.72
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Network (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Physical Sciences › Computer Science › Computer Networks and Communications › Caching and Content Delivery
- Keywords
- mobile edge computing, federated learning, deep reinforcement learning, edge caching, content distribution, communication burden reduction, edge nodes, device collaboration, dynamic optimization
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
