Pith. sign in

REVIEW 3 major objections 7 minor 40 references

Guardians of the Deep Fog: Failure-Resilient DNN Inference from Edge to Cloud

T0 review · 3 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Adding skip links that bypass physical nodes keeps a partitioned neural network accurate when nodes fail, raising accuracy by almost 16 percentage points in the harshest tested setting.

desk verdict A clearly written paper with a simple, plausible idea for failure-resilient distributed DNN inference, whose experiments are consistent with its model but leave the failure model assumptions and reproducibility untested. read the letter →

arxiv 1909.00995 v2 pith:MLCOEGNF submitted 2019-09-03 cs.NI cs.LG

Ashkan Yousefpour , Siddartha Devic , Brian Q. Nguyen , Aboudy Kreidieh , Alan Liao , Alexandre M. Bayen , Jason P. Jue This is my paper · ORCID
classification cs.NIcs.LG
keywords distributedDNNinferenceskiphyperconnectionsfailureresiliencyedge-fog-cloudreliabilitysettingaverageaccuracydeepFogGuardcomputing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Distributing a deep neural network across edge, fog, and cloud nodes can save bandwidth and latency, but when a physical node fails it takes the DNN layers hosted on that node with it, and inference can degrade to random guessing. This paper proposes deepFogGuard, a way to make such a previously partitioned distributed DNN failure-resilient without runtime reconfiguration: before deployment, add skip hyperconnections—direct links that carry activation vectors from one physical node to an ancestor node while skipping one or more nodes in between—and train the augmented network. The paper reports that in a health-activity sensing task the augmented network's average accuracy is almost 16 percentage points higher than the original network's in the Hazardous reliability setting, and that deepFogGuard outperforms the vanilla network under every reliability setting tested in both the health and multi-camera object classification experiments. If the claim holds, edge-to-cloud IoT inference can ride through node failures gracefully instead of guessing blindly until recovery.

What carries the argument

The load-bearing mechanism is the skip hyperconnection: a vector-carrying link from one physical node to an ancestor physical node that skips one or more intermediate nodes, the distributed-DNN analogue of a residual connection that skips layers. At each receiving node, an Add operation sums the vectors arriving over all incoming hyperconnections elementwise, with null vectors from failed sources ignored; an expansion layer with identity activation then hands the summed vector to the node's first ordinary layer. All hyperconnection weights are fixed to the all-ones vector, so the DNN's own weights must learn to use the redundant paths. Training is performed on the full augmented graph (in simulation, not over the live distributed system), which is what lets the network compensate for paths that may be absent at inference time. Together these parts create multiple routes from input to output, reducing the probability that a failure combination severs information flow entirely and forces random guessing.

What would settle it

Re-run the health and camera experiments with correlated failure combinations (for example, all fog nodes failing at once) or with failed nodes sending stale or partial outputs instead of the null vector $\Phi$, then compare deepFogGuard's average accuracy under Equation (3) with the paper's reported numbers; if the gap over vanilla shrinks or reverses, the resiliency claim is confined to the independent-null-failure regime.

Watch

Extended reading notes

Core claim

The paper's central claim is that skip hyperconnections make a distributed DNN failure-resilient: when physical nodes fail, activation can still reach the output layers through redundant paths, so the system gives partial or approximate classifications rather than random guessing. The paper formalizes this by modeling a failed physical node as emitting the null vector $\Phi$, by defining an Add operation at each receiving node that sums incoming hyperconnection vectors elementwise while ignoring $\Phi$, and by decreeing that a node whose incoming vectors are all $\Phi$ outputs $\Phi$; if $\Phi$ reaches the last layer, the system performs random guessing. Resiliency is then measured as average accuracy $A(G_V,R_V)$ over all $2^n$ node failure combinations, weighted by the Bernoulli probability $p(B_V|R_V)$ of each combination under a reliability setting. The experiments compare deepFogGuard to the vanilla distributed DNN under Normal, Poor, and Hazardous reliability settings for a vertically distributed health-activity network and a horizontally and vertically distributed multi-camera network; deepFogGuard has higher average accuracy in every comparison, with the largest gap in the Hazardous setting.

Load-bearing premise

The central claim rests on the assumption that physical nodes fail independently with known probabilities and that a failed node contributes a null vector, so a node receiving only nulls outputs null and the output layer guesses randomly; if failures are correlated, partial, or deliver stale values instead of nulls, the measured gains may not transfer to real deployments.

Editorial extensions

If this is right

  • A distributed DNN can keep producing useful classifications through node failures without waiting for recovery or re-partitioning, since the redundant paths are already trained in.
  • The resiliency gain is largest when the original architecture is a vertical chain with no built-in redundancy; architectures that already receive partial data from multiple sources gain less.
  • Average accuracy over Bernoulli node-failure combinations gives a numerical score for comparing distributed-DNN designs before deployment, under the model's assumptions.
  • The added resilience costs resources in failure-free operation: skip hyperconnections use extra bandwidth and require keep-alive monitoring of neighbor nodes.
  • In the tested edge-fog-cloud hierarchies, skip hyperconnections that jump over a single physical node were sufficient; skipping more than one node gave no observed gain.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Training a vanilla network with random null-vector augmentation (simulated node dropout) could isolate how much of deepFogGuard's gain comes from the extra paths themselves versus from training-time augmentation that resembles dropout.
  • Correlated failures—such as an entire fog tier losing power at once—would likely shrink the reported gains, and re-weighting Equation (2) with correlated events would quantify that shrinkage.
  • The same Add-and-expansion machinery could be inserted into partitioned CNNs or RNNs, since the paper only tests fully connected layers; this is a natural but untested extension.
  • One could test whether reweighting surviving hyperconnections after a failure adds robustness beyond deepFogGuard's fixed all-ones weights; the paper notes this as future work.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. The paper proposes deepFogGuard, an architecture augmentation scheme for distributed DNNs that adds skip hyperconnections (connections bypassing one or more physical nodes) to previously partitioned models, with the goal of preserving inference accuracy when physical nodes fail. It introduces a formal model of resiliency: each physical node has an independent survival probability (Definition 2), failures are represented as binary combinations (Definition 3), the probability of a failure combination is given by a product over independent Bernoulli trials (Eq. 2), and the resiliency metric is the expected accuracy over all failure combinations (Eq. 3). Failed nodes are assumed to output a null vector Φ, which the Add operation (Eq. 1) ignores. The paper evaluates deepFogGuard on a health activity classification dataset and a multi-camera object classification dataset under several reliability settings, reporting that deepFogGuard outperforms a vanilla distributed DNN, most strikingly by almost 16 percentage points in the Hazardous health setting (Fig. 3).

Significance. If the reported results hold up, deepFogGuard offers a simple, training-time-only mechanism for making distributed DNN inference resilient to node failures, and the formal framework in Definitions 1–4 and Eq. (3) provides a clean, interpretable way to quantify resiliency. The experimental gains are consistent with the path-redundancy intuition: with all one-hop skip hyperconnections present, a single physical-node failure still leaves an information path to the cloud. The paper's strengths include clear definitions, a well-specified experimental setup, and outcome comparisons across several reliability settings. However, the significance is bounded by three factors: the failure model assumes independent Bernoulli failures and null-vector semantics, which may not reflect real correlated or partial failures; the experiments report point estimates without error bars, despite the small size of the camera dataset; and the overhead and configuration of skip hyperconnections are not quantified. No code or model release accompanies the paper.

major comments (3)
  1. [Section 4.1, Eq. (2); Section 2 (null vector semantics)] The entire expected-accuracy metric and all reported gains rest on the assumptions that failures are independent Bernoulli events and that a failed node emits the null vector Φ, which the Add operation (Eq. 1) simply ignores. Real fog/edge failures are often correlated (shared power, cable cuts, rack failures) or partial (overloaded nodes returning stale or delayed values). Under correlated failures, the independence in Eq. (2) no longer holds, and a skip hyperconnection that bypasses one failed node can still be useless if the ancestor it connects to also fails. Under partial failures, a stale vector entering Eq. (1) corrupts the sum rather than being ignored. Because the Hazardous-setting gains in Fig. 3 are computed entirely from this model, the central claim of failure-resilient inference from edge to cloud needs either an explicit scoping argument as to why these failure modes are excluded, or additional experiments with correlated and stale failure models.
  2. [Section 4.4, Fig. 3] The paper states that results are averaged 'over 10 runs' but reports only point estimates, with no error bars, confidence intervals, or significance tests. The camera dataset has only roughly 1,400 data points (Section 4.2), so the reported Hazardous gap of 10.45 percentage points (91.17 vs. 80.72) and the health Hazardous gap of 16.40 points (79.37 vs. 62.97) could be within run-to-run or train/test variance. Please report the mean and standard deviation (or a confidence interval) across the 10 runs, ideally with multiple data splits, to demonstrate that the observed improvements are stable and not an artifact of a single seed or split.
  3. [Section 3 and Section 6 (limitations)] The paper acknowledges that 'determining the number of skip hyperconnections is a non-trivial task' and that redundant paths consume additional bandwidth, but it does not quantify the overhead or provide a procedure for selecting the topology. Since deepFogGuard sends data along skip hyperconnections even in the absence of failures, the practical value of the method depends on the cost of this overhead and on whether the reported gains are robust to the particular topology shown in Fig. 2. At minimum, the authors should state how the all-one-hop-skip topology was chosen and report the additional latency, bandwidth, and training cost incurred by the augmented architecture.
minor comments (7)
  1. [Section 1] There is a typographical error: 'thie current approach' should be 'the current approach'.
  2. [Section 3] The phrase 'the determining the number of skip hyperconnections is a non-trivial task' should be rephrased, e.g., 'determining the number of skip hyperconnections is a non-trivial task'.
  3. [Section 3] The statement 'We tend towards passive resiliency' is vague; please clarify what 'passive' means in this context and how it differs from other forms of resiliency.
  4. [Section 4.3] The paper says it 'only includes the skip hyperconnections that skip one physical node' but does not justify why all possible one-hop skips are included in the reported architecture; a brief explanation or a diagram enumerating the connections would help.
  5. [Figure 3] The table above the plot is not explained in the text, and the header 'Config Health + dFG Health + Vanilla' is confusing; consider removing the table or clearly labeling the columns as deepFogGuard and Vanilla.
  6. [Section 4.2] The camera dataset is very small (about 1,400 data points) and class-imbalanced; this limitation should be stated more prominently, as it affects the generalizability of the results.
  7. [Section 4.4 and Section 5] The experiments compare only against a vanilla distributed DNN; to strengthen the claim that the gains are due to skip hyperconnections specifically, the authors should consider a baseline that uses other fault-tolerance techniques (e.g., dropout or fault-injection training) while keeping the architecture fixed.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; the only self-citation is a background reference and the central claim rests on measured accuracy, not on fitted or self-referential terms.

full rationale

The paper's central derivation is self-contained and empirical. Skip hyperconnections are structural additions inspired by external residual-connection work (He et al. [13]), and the resiliency metric in Eq. (3) is a definitional expectation over independent Bernoulli failure combinations whose accuracy values A(GV ∅ BV) are measured on held-out test data from external datasets. No parameter is fitted to a subset of results and then renamed as a prediction; no uniqueness theorem is imported from the authors' prior work; and no ansatz is smuggled in via self-citation. The only self-citation, [34], is used for background definitions of fog/edge nodes and IoT application categories and does not carry the resiliency claim. The assumptions about independent failures, null-vector semantics for failed nodes, and exclusion of cloud/IoT failures are scope conditions of the experimental model, not circular inputs that force the reported accuracy gains.

Assumptions & free parameters 2 free parameters · 6 assumptions · 2 invented entities

The central claim rests on a simplified failure model (independent, complete node failures with null-vector output), a fixed hyperconnection weight choice, and a hand-selected skip topology. The paper contributes the skip hyperconnection idea and a clean expectation formula, but it draws the residual-connection inspiration and the standard DNN machinery from prior literature.

free parameters (2)
  • Hyperconnection weight vectors w_ij = 1 (all elements)
    The paper sets all hyperconnection weights to 1 and claims this is without loss of generality because training compensates; this simplification is not varied or learned.
  • Skip hyperconnection topology = All hyperconnections that skip one physical node
    The set of skip hyperconnections is chosen by hand; the paper notes the right set is task-dependent, so the reported gains are tied to this specific topology.
assumptions (6)
  • domain assumption Physical node failures are independent Bernoulli events with known reliability probabilities (Eq. 2).
    The resiliency metric A(GV,RV) sums over failure combinations with p(BV|RV) as a product of independent probabilities; correlated or unknown failure rates would change the expected accuracy.
  • domain assumption A failed physical node emits the null vector Φ; a node with all-null inputs emits Φ; Φ at the output triggers random guessing (Section 2).
    This defines the failure semantics used in the experiments and in Eq. (3).
  • domain assumption Hyperconnection weight vectors are set to 1 without loss of generality (Section 2 Remark).
    The paper asserts training can compensate, but this is not demonstrated across different weight values.
  • domain assumption The cloud node never fails and IoT or sensor nodes never fail (Section 4.3).
    The experiments exclude exactly the failure modes that would produce null or missing input data, narrowing the claim to intermediate node failures.
  • domain assumption Zero-padding matches hyperconnection vector dimensions (Section 2).
    The paper assumes dimension mismatch can be resolved by zero-padding without affecting learned behavior.
  • domain assumption Training can be performed in a simulated distributed environment and the resulting weights remain valid for physical deployment (Section 2, Training Process).
    The paper does not deploy or test on real physical nodes; it simulates failures during inference.
invented entities (2)
  • Skip hyperconnection
    purpose: Adds a direct vector pathway from a descendant physical node to an ancestor node so that inference can bypass failed intermediate nodes.
    The concept is introduced by this paper; its benefit is shown only in the paper's own experiments and no external validation is provided.
  • Expansion layer (l#_i)
    purpose: Receives the summed vector from incoming hyperconnections and expands it to the units of the first layer on a physical node.
    An architectural component introduced to interface hyperconnections with standard DNN layers; it behaves as an identity layer and adds no independent evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Guardians of the Deep Fog: Failure-Resilient DNN Inference from Edge to Cloud." pith.science (2026). https://pith.science/paper/MLCOEGNF

@misc{pith2026190900995,
  author       = {Pith},
  title        = {Pith review of: Guardians of the Deep Fog: Failure-Resilient DNN Inference from Edge to Cloud},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MLCOEGNF}},
  note         = {Machine review of arXiv:1909.00995}
}
read the original abstract

Partitioning and distributing deep neural networks (DNNs) over physical nodes such as edge, fog, or cloud nodes, could enhance sensor fusion, and reduce bandwidth and inference latency. However, when a DNN is distributed over physical nodes, failure of the physical nodes causes the failure of the DNN units that are placed on these nodes. The performance of the inference task will be unpredictable, and most likely, poor, if the distributed DNN is not specifically designed and properly trained for failures. Motivated by this, we introduce deepFogGuard, a DNN architecture augmentation scheme for making the distributed DNN inference task failure-resilient. To articulate deepFogGuard, we introduce the elements and a model for the resiliency of distributed DNN inference. Inspired by the concept of residual connections in DNNs, we introduce skip hyperconnections in distributed DNNs, which are the basis of deepFogGuard's design to provide resiliency. Next, our extensive experiments using two existing datasets for the sensing and vision applications confirm the ability of deepFogGuard to provide resiliency for distributed DNNs in edge-cloud networks.

Figures

Figures reproduced from arXiv: 1909.00995 by the authors.

Figure 1
Figure 1. Partitioning a DNN and distributing it across phys [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Distributed DNN architecture in deepFogGuard for: (a) health activity classification, (b) multi-camera object clas￾sification experiments. The expansion layers are not shown. (Physical node numbers are unique within each experiment) p(BV |RV ) = Ö |V | i=1 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Average accuracy (%) vs. reliability setting increasing the resiliency of the distributed DNNs, drastically out￾performing Vanilla. In the health activity classification experiment (Fig. 3a), in the Hazardous reliability setting, deepFogGuard increases average ac￾curacy by almost 16%, relative to the baseline Vanilla. Similarly, in the Poor reliability setting, the average accuracy of deepFogGuard is around 6% highe… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 34 canonical work pages

  1. [1]

    Martín Abadi, Paul Barham, Jianmin Chen, Zhifeng Chen, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Geoffrey Irving, Michael Isard, et al

  2. [2]

    Muhammad Ali, Ashiq Anjum, M Usman Yaseen, A Reza Zamani, Daniel Balouek- Thomert, Omer Rana, and Manish Parashar. 2018. Edge enhanced deep learning system for large-scale video stream analytics. InIEEE 2nd International Conference on Fog and Edge Computing (ICFEC) . IEEE, 1–10

    work page 2018

  3. [3]

    Oresti Banos, Claudia Villalonga, Rafael Garcia, Alejandro Saez, Miguel Damas, Juan A Holgado-Terriza, Sungyong Lee, Hector Pomares, and Ignacio Rojas. 2015. Design, implementation and validation of a novel open framework for agile development of mobile health applications. Biomedical engineering online 14, 2 (2015)

    work page 2015

  4. [4]

    George Ravuama Bolt. 1992. Fault Tolerance in Artificial Neural Networks . Ph.D. Dissertation. University of York

    work page 1992

  5. [5]

    Lingjiao Chen, Hongyi Wang, Zachary Charles, and Dimitris Papailiopoulos

    work page

  6. [6]

    Yitao Chen, Kaiqi Zhao, Baoxin Li, and Ming Zhao. 2019. Exploring the Use of Synthetic Gradients for Distributed Deep Learning across Cloud and Edge Resources. In 2nd{USENIX} Workshop on Hot Topics in Edge Computing (HotEdge 19)

    work page 2019

  7. [7]

    Trishul M Chilimbi, Yutaka Suzue, Johnson Apacible, and Karthik Kalyanaraman

    work page

  8. [8]

    Georgios Damaskinos, El Mahdi El Mhamdi, Rachid Guerraoui, Arsany Hany Ab- delmessih Guirguis, and SÃľbastien Louis Alexandre Rouault. 2019. AGGRE- GATHOR: Byzantine Machine Learning via Robust Gradient Aggregation. (2019). Conference on Systems and Machine Learning (SysML) 2019, Stanford, CA, USA

    work page 2019

Show all 40 references
  1. [9]

    Swarnava Dey, Jayeeta Mondal, and Arijit Mukherjee. 2019. Offloaded Execu- tion of Deep Learning Inference at Edge: Challenges and Insights. In 2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops). IEEE, 855–861

    2019

  2. [10]

    EM El Mhamdi, R Guerraoui, and S Rouault. 2017. On the robustness of a neural network. In 2017 IEEE 36th Symposium on Reliable Distributed Systems (SRDS) . 84–93

    2017

  3. [11]

    Tarek Elgamal, Atul Sandur, Phuong Nguyen, Klara Nahrstedt, and Gul Agha

  4. [12]

    Haibo He and Edwardo A Garcia. 2008. Learning from imbalanced data. IEEE Transactions on Knowledge & Data Engineering 9 (2008), 1263–1284

    2008

  5. [13]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . 770–778

    2016

  6. [14]

    Chuang Hu, Wei Bao, Dan Wang, and Fengming Liu. 2019. Dynamic Adaptive DNN Surgery for Inference Acceleration on the Edge. InIEEE INFOCOM 2019-IEEE Conference on Computer Communications . IEEE, 1423–1431

    2019

  7. [15]

    In 2018 IEEE 11th International Conference on Cloud Computing (CLOUD)

    DROPLET: Distributed Operator Placement for IoT Applications Spanning Edge and Cloud Resources. In 2018 IEEE 11th International Conference on Cloud Computing (CLOUD). IEEE, 1–8

    2018

  8. [16]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  9. [17]

    En Li, Zhi Zhou, and Xu Chen. 2018. Edge Intelligence: On-Demand Deep Learning Model Co-Inference with Device-Edge Synergy. In Proceedings of the 2018 Workshop on Mobile Edge Communications (MECOMM’18) . ACM, 31–36

    2018

  10. [18]

    Peng Liu, Bozhao Qi, and Suman Banerjee. 2018. EdgeEye: An Edge Service Framework for Real-time Intelligent Video Analytics. In Proceedings of the 1st International Workshop on Edge Systems, Analytics and Networking . ACM, 1–6

    2018

  11. [19]

    Yiping Kang, Johann Hauswald, Cao Gao, Austin Rovinski, Trevor Mudge, Jason Mars, and Lingjia Tang. 2017. Neurosurgeon: Collaborative intelligence between the cloud and mobile edge. InACM SIGARCH Computer Architecture News, Vol. 45. ACM, 615–629

    2017

  12. [20]

    Ahsan Morshed, Prem Prakash Jayaraman, Timos Sellis, Dimitrios Georgakopou- los, Massimo Villari, and Rajiv Ranjan. 2017. Deep osmosis: Holistic distributed deep learning in osmotic computing. IEEE Cloud Computing 4, 6 (2017), 22–32

    2017

  13. [21]

    Jihong Park, Sumudu Samarakoon, Mehdi Bennis, and Mérouane Debbah. 2018. Wireless network intelligence at the edge. arXiv preprint arXiv:1812.02858 (2018)

    2018 arXiv

  14. [22]

    Adam Paszke, Sam Gross, Soumith Chintala, and Gregory Chanan. 2017. Pytorch: Tensors and dynamic neural networks in python with strong gpu acceleration. (2017)

    2017

  15. [23]

    Kishan Mehrotra, Chilukuri K Mohan, Sanjay Ranka, and Ching-tai Chiu. 1994. Fault tolerance of neural networks . Technical Report. Syracuse University. Tech. Rep. RL-TR-94-93. Syracuse University

    1994

  16. [24]

    G. Roig, X. Boix, H. Ben Shitrit, and P. Fua. 2011. Conditional Random Fields for multi-camera object detection. In 2011 International Conference on Computer Vision. 563–570

    2011

  17. [25]

    Carlo H Sequin and RD Clay. 1990. Fault tolerance in artificial neural networks. In 1990 IJCNN international joint conference on neural networks . IEEE, 703–708

    1990

  18. [26]

    Alexander Sergeev and Mike Del Balso. 2018. Horovod: fast and easy distributed deep learning in TensorFlow. arXiv preprint arXiv:1802.05799 (2018)

    2018 arXiv

  19. [27]

    Dhananjay S Phatak and Israel Koren. 1995. Complete and partial fault tolerance of feedforward neural nets. IEEE Transactions on Neural Networks 6, 2 (1995), 446–456

    1995

  20. [28]

    Zeyi Tao and Qun Li. 2018. eSGD: Communication Efficient Distributed Deep Learning on the Edge. InUSENIX Workshop on Hot Topics in Edge Computing (Hot- Edge 18). USENIX Association, Boston, MA. https://www.usenix.org/conference/ hotedge18/presentation/tao

    2018

  21. [29]

    Surat Teerapittayanon, Bradley McDanel, and HT Kung. 2017. Distributed deep neural networks over the cloud, the edge and end devices. InDistributed Comput- ing Systems (ICDCS), 2017 IEEE 37th International Conference on . IEEE, 328–339

    2017

  22. [30]

    Cesar Torres-Huitzil and Bernard Girau. 2017. Fault and error tolerance in neural networks: A review. IEEE Access 5 (2017), 17322–17341

    2017

  23. [31]

    Rupesh K Srivastava, Klaus Greff, and Jürgen Schmidhuber. 2015. Training very deep networks. In Advances in neural information processing systems (NeurIPS) . 2377–2385

    2015

  24. [32]

    Shiqiang Wang, Murtaza Zafer, and Kin K Leung. 2017. Online placement of multi-component applications in edge computing environments. IEEE Access 5 (2017), 2514–2533

    2017

  25. [33]

    Shuochao Yao, Yiran Zhao, Aston Zhang, Lu Su, and Tarek Abdelzaher. 2017. Deepiot: Compressing deep neural network structures for sensing systems with a compressor-critic framework. In Proceedings of the 15th ACM Conference on Embedded Network Sensor Systems . ACM, 4

    2017

  26. [34]

    Ashkan Yousefpour, Caleb Fung, Tam Nguyen, Krishna Kadiyala, Fatemeh Jalali, Amirreza Niakanlahiji, Jian Kong, and Jason P Jue. 2019. All one needs to know about fog computing and related edge computing paradigms: A complete survey. Journal of Systems Architecture 98 (2019), 289 – 330

    2019

  27. [35]

    Shiqiang Wang, Tiffany Tuor, Theodoros Salonidis, Kin K Leung, Christian Makaya, Ting He, and Kevin Chan. 2019. Adaptive federated learning in re- source constrained edge computing systems. IEEE Journal on Selected Areas in Communications 37, 6 (2019), 1205–1221

    2019

  28. [36]

    Zhi-Hua Zhou and Shi-Fu Chen. 2003. Evolving fault-tolerant neural networks. Neural Computing & Applications 11, 3-4 (2003), 156–160

    2003

  29. [39]

    Li Zhou, Hao Wen, Radu Teodorescu, and David HC Du. 2019. Distributing Deep Neural Networks with Containerized Partitions at the Edge. In 2nd{USENIX} Workshop on Hot Topics in Edge Computing (HotEdge 19)

    2019

  30. [2014]

    In OSDI, Vol

    Project Adam: Building an Efficient and Scalable Deep Learning Training System.. In OSDI, Vol. 14. 571–582

  31. [2016]

    In OSDI, Vol

    Tensorflow: a system for large-scale machine learning.. In OSDI, Vol. 16. 265–283

  32. [2018]

    In Proceedings of the 35th International Conference on Machine Learning , Vol

    DRACO: Byzantine-resilient Distributed Training via Redundant Gradients. In Proceedings of the 35th International Conference on Machine Learning , Vol. 80. PMLR, 903–912. http://proceedings.mlr.press/v80/chen18l.html

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.