Distributed Inference on Mobile Edge and Cloud: A Data-Cartography based Clustering Approach

Divya Jyoti Bajpai; Manjesh Kumar Hanawal

REVIEW 4 major objections 6 minor 32 references

Distributed Inference on Mobile Edge and Cloud: A Data-Cartography based Clustering Approach

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that routing text inputs to mobile, edge, or cloud models by their embedding distance to data-cartography pools cuts inference cost by more than 43% while keeping accuracy within 0.5% of cloud-only inference.

desk verdict Plausible new method, but the headline accuracy-drop claim doesn't survive contact with its own tables; still deserves refereeing. read the letter →

arxiv 2412.16616 v1 pith:GXSSWQE6 submitted 2024-12-21 cs.DC cs.AI

Divya Jyoti Bajpai , Manjesh Kumar Hanawal This is my paper · ORCID

classification cs.DCcs.AI

keywords distributedinferencedatacartographyearlyexitingcloudoffloadingBERTGLUEsamplecomplexityedgecomputing

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

This paper proposes a way to decide, before any heavy computation, whether a text input should be processed on a mobile device, an edge server, or the cloud. It uses data cartography: during training, each validation sample receives a confidence score and a variance score from the model's predictions across epochs, and samples are grouped into easy, medium, and hard pools according to thresholds. At inference time, the input's word embedding is compared with the average embedding of each pool, and the sample is sent to the device whose pool it is closest to. Across several GLUE tasks, the authors report that this routing lowers inference cost by more than 43%, and up to 75% for the larger BERT model, while losing under 0.5% accuracy compared with doing every inference on the cloud. The adaptive version updates the pool averages online to track distribution shift.

What carries the argument

The load-bearing object is the data-cartography pool. During training, for each validation sample the model records $\hat\mu_i$, the mean probability of the true label across epochs, and $\hat\sigma_i$, the variance of those probabilities. Thresholds $\alpha$ and $\beta$ split samples into easy ($\hat\mu_i \ge \alpha$, $\hat\sigma_i < \beta$), medium (high confidence, high variance), and hard (low confidence regardless of variance), and the sample's word embedding is appended to the corresponding pool. At inference the pools are reduced to their average embeddings $P^a_e$, $P^a_m$, and $P^a_h$, and a new sample is routed by minimum Euclidean distance to these centroids, in the manner of K-means assignment. The thresholds are chosen by maximizing an expected reward that subtracts processing and offloading costs from classifier confidence, which is how the accuracy-cost trade-off enters the construction.

What would settle it

Compare the pool label each validation sample gets from the $\alpha$-$\beta$ thresholds with the label it gets from nearest average embedding on the same validation set; if the agreement rate is low, especially between the medium and hard pools, then the distance rule is not reproducing the data-cartography classification and the cost savings are not explained by sample difficulty.

Watch

Extended reading notes

Core claim

The central claim is that sample difficulty, as defined by data cartography's training dynamics, can be read off in word-embedding space: a test sample can be assigned to mobile, edge, or cloud by the nearest average embedding of the easy, medium, and hard validation pools, without running the input through the small DNN first. The authors argue this clustering preserves the accuracy of the full backbone because no parameters are removed, and it lowers cost because only the embedding layer runs on mobile for routing. Their experiments on BERT-base and BERT-large over the GLUE benchmark show the routing cuts cost by more than 43% compared with cloud-only inference, keeps the accuracy drop under 0.5%, and sometimes improves on cloud-only accuracy because easy samples avoid the overthinking effect.

Load-bearing premise

The load-bearing assumption is that a new sample's distance to the average embedding of the easy, medium, and hard pools, built from validation-time confidence and variance scores, correctly recovers how hard that sample actually is for the smaller models; if embedding proximity does not track difficulty, the routing decisions and the reported accuracy-cost trade-off collapse.

Editorial extensions

If this is right

If the claim is correct, routing can be done with only the embedding layer on the mobile device, so no sample pays the full mobile processing cost before offloading.
On the reported GLUE tasks, the fixed and adaptive routers cut cost by more than 43% relative to cloud-only BERT, with savings up to 75% on BERT-large.
Because no parameters are pruned or quantized, the backbone's accuracy is preserved, and on QQP the method can even beat cloud-only inference by avoiding the overthinking effect.
The adaptive version updates the pool centroids online, giving a low-cost mechanism to follow distribution shift during test time.

Reading between the lines

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

Editorial inference: The method in effect claims that nearest-centroid distance in embedding space is a faithful proxy for the $\alpha$-$\beta$ training-dynamics labels; the paper shows t-SNE plots but does not report a quantitative agreement rate, so a label-versus-centroid agreement study on the validation split is a natural test.
Editorial inference: Because the routing signal is only an embedding distance, the same scheme should transfer to other transformer backbones, languages, and even vision models using a feature layer instead of word embeddings, though the paper only demonstrates BERT on GLUE.
Editorial inference: The reported savings are computed under one fixed cost ratio ($\lambda_m=\frac32\lambda$, $o_1=\frac52\lambda$, $o_2=3\lambda$); deployments with different battery, bandwidth, and cloud-pricing structures would need to re-solve the reward maximization, though the ablation suggests accuracy is stable across cost changes.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Worth a look if you work on mobile/edge inference. The idea is simple and sensible: use data cartography (training dynamics) to label validation samples as easy/medium/hard, pool their embeddings, and at test time route each sample to mobile/edge/cloud based on distance to the pool centroid. That specific combination is new relative to the early-exit and split-computing work they cite, and it avoids the usual requirement of running the sample through the mobile tower first. The experiments show consistent cost reductions across GLUE, with accuracy close to the cloud baseline, and they compare against reasonable baselines like DeeBERT, AdaEE, and I-SplitEE. Credit where due: the method is clearly explained, the cost model is explicit, and they include an ablation on cost parameters.

The soft spots are real but correctable. The headline claim in the abstract—accuracy drop less than 0.5%—is contradicted by their own Table II: BERT-large drops 0.6 points on CoLA (59.5 to 58.9) and QNLI (93.1 to 92.5). The conclusion says <0.3%, which Table I also refutes (MRPC drops 0.5). That is a load-bearing number, so it either needs re-running with multiple seeds and error bars or a re-scoped claim that names where the bound holds. Second, Algorithm 1's medium-pool condition disagrees with the prose and with Equation 5: the prose says medium is high-confidence, high-variance, but the algorithm's line 9 routes µ<α, σ≥β to medium. That is an internal inconsistency a referee will catch immediately. Third, the paper never quantifies how well the embedding-distance assignment agrees with the threshold-based labels; the t-SNE pictures are suggestive but not evidence. That agreement is the load-bearing assumption of the whole routing scheme, so it needs a number.

I'm less worried about the threshold selection on validation, since accuracy and cost are still measured on held-out test data. The self-consistency of the cost model is mild, not fatal.

Bottom line: this is a plausible engineering contribution with a couple of fixable flaws. It deserves a serious referee. I'd send it to review with a request for re-scoped claims, an algorithm/prose cleanup, and a quantitative agreement analysis. Would I cite it? Yes, if the claims get corrected—it's a useful baseline for data-cartography-based routing.

Referee Report

4 major / 6 minor

Summary. The paper proposes DIMEC-DC, a distributed inference framework for NLP that routes each input sample to a mobile, edge, or cloud device according to its estimated difficulty. Sample difficulty is derived from data-cartography training dynamics: confidence and variance of the ground-truth label probability over training epochs are computed on a validation split, thresholds alpha and beta are chosen to maximize a reward function, and samples are grouped into easy, medium, and hard pools of word embeddings. At inference time, a test sample is embedded and assigned to the nearest pool average, with easy samples inferred on mobile, medium samples on edge, and hard samples on cloud. The authors evaluate the method on six GLUE tasks with BERT-base and BERT-large, compare against several baselines, and claim a cost reduction of more than 43% with an accuracy drop below 0.5% relative to all-cloud inference.

Significance. If the method performs as claimed, it is a worthwhile practical contribution: routing decisions require only the embedding layer on the mobile device, no parameters are removed from the backbone, and the adaptive version attempts to track distribution shift. The paper also provides a source-code link, cost ablations, and comparisons to several relevant baselines. However, the headline accuracy-drop claim is contradicted by the paper's own tables, the pool-creation algorithm is internally inconsistent, and the nearest-centroid assignment is not quantitatively validated against the threshold-based labels. These issues are fixable but currently prevent the central claim from being accepted as stated.

major comments (4)

[Abstract, Section VIII, Tables I-II] The central claim that DIMEC-DC maintains an accuracy drop of less than 0.5% is contradicted by the paper's own results. In Table II, the BERT-large CoLA accuracy drops from 59.5 to 58.9 (0.6 points) and QNLI from 93.1 to 92.5 (0.6 points); in Table I, the BERT-base MRPC accuracy drops from 89.2 to 88.7 (0.5 points). The conclusion's stronger claim of less than 0.3% is also false for MRPC. Because the quantified bound is the headline result, the experiments should be rerun with multiple seeds and reported with error bars, or the claim should be re-scoped to the settings where it actually holds.
[Section IV-B, Section IV-C, Algorithm 1, Eq. (5)] Algorithm 1's pool-creation conditions contradict both the prose and the reward function. The prose in Section IV-B and Eq. (5) classify high-confidence, high-variance samples as medium, but Algorithm 1 labels such samples as hard because its medium branch requires confidence below alpha and variance above beta. This inconsistency changes the composition of the medium and hard pools and therefore changes routing decisions. Please correct Algorithm 1 and state the intended mapping from (confidence, variance) to (easy, medium, hard) unambiguously.
[Section IV-D] The nearest-centroid assignment is load-bearing but is not quantitatively validated. The method reduces each pool to its mean embedding and assigns test samples to the closest mean, yet no experiment reports how often this assignment reproduces the alpha/beta threshold labels on the validation set. The t-SNE visualizations in Figures 4b and 4c are qualitative only. If embedding-space proximity does not track the training-dynamics difficulty labels, the cost-accuracy trade-off is not explained by data cartography. Please report a quantitative agreement measure, such as cluster purity or a confusion matrix between threshold-based and centroid-based labels.
[Section V.B, Table I caption, Section VI] The definition of the BERT baseline is inconsistent. Section V.B says the BERT baseline is 'deployed on the mobile device and only processing cost is incurred,' while Table I's caption and Section VI describe the baseline as the original BERT model with all samples processed on the cloud. Since the abstract's accuracy-drop and cost-reduction claims are relative to all-cloud inference, the baseline deployment must be stated consistently and the cost model applied to the baseline must be explicit.

minor comments (6)

[Eq. (5) and Section IV-C] The text after Eq. (5) says a medium-pool sample satisfies 'x ∈ Pe', which should be 'x ∈ Pm'. Also, the reward function's condition for the edge case uses 'σi > β' while Algorithm 1 uses 'σi ≥ β' for medium; boundary conventions should be clarified.
[Section IV-C and Section V-D] Cost notation is inconsistent: Section IV-C uses λm, λe, oe, oc, γ, while Section V-D introduces λ1, λ2, o1, o2, µ for the same quantities. Please unify the notation.
[Algorithm 2] Algorithm 2 does not specify how ties are broken when the test sample is equidistant from two pool averages; since the default routing differs by pool, this should be stated.
[Section V-A, Figure 3] Figure 3 includes an RTE panel, but RTE is not listed in the dataset description in Section V-A and no RTE results appear in Tables I or II. Please clarify whether RTE is part of the evaluation and what the figure shows.
[Section V-C and Tables I-II] The tables report single accuracy values without standard deviations or seeds. Given that the key claim concerns accuracy differences of half a percentage point, repeat experiments and report variance are necessary to assess whether observed differences are meaningful.
[Tables I-II] The BERT-large evaluation covers only CoLA, MRPC, and QNLI, while the BERT-base evaluation covers six GLUE tasks. Please state why the remaining tasks are omitted for BERT-large, or report them, to support the claim of evaluation on a variety of NLP tasks.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: thresholds are tuned on a validation split, and accuracy/cost are reported on a held-out test split; the paper's own tables contradict the headline accuracy-drop bound, but that is a correctness issue rather than a circularity.

full rationale

DIMEC-DC's derivation chain is: (1) compute confidence/variance on the validation split using saved epoch weights (Eqs. 1-2/4); (2) choose thresholds alpha and beta by maximizing the reward in Eq. 5 on that same validation split; (3) form easy/medium/hard embedding pools and reduce them to centroids; (4) route test samples by nearest centroid and measure accuracy and cost on the test split. The only place where the input and output metrics share a formula is the cost model: the same user-defined costs (lambda, o, gamma) appear in the reward used for threshold selection and in the reported cost savings. This is a mild self-consistency, but it is not a circular identity because Table I/II accuracy and cost are computed on a held-out test split after the thresholds are fixed, and test routing does not use ground-truth labels or validation confidence. The load-bearing assumption that embedding distance to pool centroids recovers training-dynamics complexity is empirically testable and is not presupposed by the equations. The absence of a quantitative agreement analysis between threshold labels and centroid labels is an empirical validation gap, not a circular reduction. The abstract's '<0.5% accuracy drop' claim is contradicted by the paper's own BERT-large results (CoLA and QNLI drop 0.6) and by BERT-base MRPC (drop 0.5), and the conclusion's '<0.3%' is likewise contradicted by Table I; however, an internal numerical inconsistency is a correctness problem, not a circularity. Self-citations ([11],[29],[30]) appear only as baseline comparisons and are not load-bearing. Data cartography itself is imported from external reference [15]. No step reduces to its own input by construction, so the circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the two thresholds, the hand-chosen cost parameters, the layer split, and two strong domain assumptions about embedding geometry and training-dynamics transfer. No new physical entities are introduced; the pools and centroids are algorithmic objects computed from data.

free parameters (4)

threshold alpha = chosen from {0.55, 0.6, 0.65, 0.7, 0.8} per dataset
Confidence threshold for easy/medium/hard pools, selected by maximizing the reward in Equation 5 on the validation split.
threshold beta = chosen from {0.05, 0.08, 0.11, 0.14, 0.17} per dataset
Variance threshold for pools, co-selected with alpha on validation.
cost coefficients lambda_m, lambda_e, o_e, o_c, gamma = lambda_e = lambda, lambda_m = 1.5*lambda, o_e = 2.5*lambda, o_c = 3*lambda; gamma not specified in the main table
User-defined cost model used both to select thresholds and to report cost reductions; results are contingent on these ratios.
layer split m, n = m = 4, n = 6 for BERT-base; m = 6, n = 12 for BERT-large
Number of layers deployed on mobile and edge; chosen by user based on device resources and cost structure.

assumptions (4)

domain assumption Validation-set training dynamics at the m-th layer classifier are a reliable proxy for sample complexity at inference time.
Introduced in Section IV-B, where confidence and variance are computed on the validation split and used to build pools.
domain assumption Word embeddings of samples with similar complexity cluster in Euclidean space, so nearest-centroid assignment recovers the complexity class.
Stated in Section IV-D, where pool averages are used as K-means centroids; no quantitative validation is provided.
domain assumption The cost model is linear, additive, and accurately represents mobile, edge, and cloud processing and offloading costs.
Defined in Section IV-C and used in the reward function; the reported cost savings are relative to this model.
domain assumption Adaptively updating pool averages with test samples preserves cluster validity under distribution shift.
Algorithm 2 updates centroids with each arriving sample; the paper asserts this captures shift without convergence or contamination analysis.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Distributed Inference on Mobile Edge and Cloud: A Data-Cartography based Clustering Approach." pith.science (2026). https://pith.science/paper/GXSSWQE6

@misc{pith2026241216616,
  author       = {Pith},
  title        = {Pith review of: Distributed Inference on Mobile Edge and Cloud: A Data-Cartography based Clustering Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GXSSWQE6}},
  note         = {Machine review of arXiv:2412.16616}
}

read the original abstract

The large size of DNNs poses a significant challenge for deployment on devices with limited resources, such as mobile, edge, and IoT platforms. To address this issue, a distributed inference framework can be utilized. In this framework, a small-scale DNN (initial layers) is deployed on mobile devices, a larger version on edge devices, and the full DNN on the cloud. Samples with low complexity (easy) can be processed on mobile, those with moderate complexity (medium) on edge devices, and high complexity (hard) samples on the cloud. Given that the complexity of each sample is unknown in advance, the crucial question in distributed inference is determining the sample complexity for appropriate DNN processing. We introduce a novel method named \our{}, which leverages the Data Cartography approach initially proposed for enhancing DNN generalization. By employing data cartography, we assess sample complexity. \our{} aims to boost accuracy while considering the offloading costs from mobile to edge/cloud. Our experimental results on GLUE datasets, covering a variety of NLP tasks, indicate that our approach significantly lowers inference costs by more than 43\% while maintaining a minimal accuracy drop of less than 0.5\% compared to performing all inferences on the cloud. The source code is available at https://anonymous.4open.science/r/DIMEC-1B04.

Figures

Figures reproduced from arXiv: 2412.16616 by the authors.

**Figure 2.** In this figure, three types of reviews are input to the mobile device. It passes through the embedding layer on the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** The figure shows the clustering of samples based on the confidence and variance on the validation split of the datasets, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: The figure shows the accuracy of the individual devices i.e., mobile, edge and cloud. Figure on centre and right: The [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

**Figure 5.** Figure 5: The changes in accuracy and percentage change in cost values when one of the costs is varied while keeping others at [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 14 canonical work pages

[1]

A survey on vision transformer,

K. Han, Y . Wang, H. Chen, X. Chen, J. Guo, Z. Liu, Y . Tang, A. Xiao, C. Xu, Y . Xuet al., “A survey on vision transformer,” IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 1, pp. 87–110, 2022

2022
[2]

GLUE: A multi-task benchmark and analysis platform for natural language understanding,

A. Wang, A. Singh, J. Michael, F. Hill, O. Levy, and S. R. Bowman, “GLUE: A multi-task benchmark and analysis platform for natural language understanding,” 2019, in the Proceedings of ICLR

2019
[3]

To prune, or not to prune: exploring the efficacy of pruning for model compression,

M. Zhu and S. Gupta, “To prune, or not to prune: exploring the efficacy of pruning for model compression,” arXiv preprint arXiv:1710.01878 , 2017

arXiv 2017
[4]

Are sixteen heads really better than one?

P. Michel, O. Levy, and G. Neubig, “Are sixteen heads really better than one?” Advances in neural information processing systems , vol. 32, 2019

work page 2019
[5]

Ternarybert: Distillation-aware ultra-low bit bert,

W. Zhang, L. Hou, Y . Yin, L. Shang, X. Chen, X. Jiang, and Q. Liu, “Ternarybert: Distillation-aware ultra-low bit bert,” arXiv preprint arXiv:2009.12812, 2020

arXiv 2009
[6]

I-bert: Integer-only bert quantization,

S. Kim, A. Gholami, Z. Yao, M. W. Mahoney, and K. Keutzer, “I-bert: Integer-only bert quantization,” in International conference on machine learning. PMLR, 2021, pp. 5506–5518

2021
[7]

Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter,

V . Sanh, L. Debut, J. Chaumond, and T. Wolf, “Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter,” arXiv preprint arXiv:1910.01108, 2019

arXiv 1910
[8]

Tinybert: Distilling bert for natural language understanding,

X. Jiao, Y . Yin, L. Shang, X. Jiang, X. Chen, L. Li, F. Wang, and Q. Liu, “Tinybert: Distilling bert for natural language understanding,” arXiv preprint arXiv:1909.10351 , 2019

arXiv 1909

Show all 32 references

[9]

Deebert: Dynamic early ex- iting for accelerating bert inference,

J. Xin, R. Tang, J. Lee, Y . Yu, and J. Lin, “Deebert: Dynamic early ex- iting for accelerating bert inference,” arXiv preprint arXiv:2004.12993 , 2020

2004 arXiv
[10]

Bert loses patience: Fast and robust inference with early exit,

W. Zhou, C. Xu, T. Ge, J. McAuley, K. Xu, and F. Wei, “Bert loses patience: Fast and robust inference with early exit,” Advances in Neural Information Processing Systems , vol. 33, pp. 18 330–18 341, 2020

2020
[11]

Ceebert: Cross-domain inference in early exit bert,

D. J. Bajpai and M. K. Hanawal, “Ceebert: Cross-domain inference in early exit bert,” in To appear in proceedings of the 62nd conference of the Association for computational linguistics: Findings Volume , 2024

2024
[12]

Split computing and early exiting for deep learning applications: Survey and research challenges,

Y . Matsubara, M. Levorato, and F. Restuccia, “Split computing and early exiting for deep learning applications: Survey and research challenges,” ACM Computing Surveys , vol. 55, no. 5, pp. 1–30, 2022

2022
[13]

Bert: Pre-training of deep bidirectional transformers for language understanding,

J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” arXiv preprint arXiv:1810.04805, 2018

2018 arXiv
[14]

Revolutionizing mobile interaction: Enabling a 3 billion parameter gpt llm on mobile,

S. Carreira, T. Marques, J. Ribeiro, and C. Grilo, “Revolutionizing mobile interaction: Enabling a 3 billion parameter gpt llm on mobile,” arXiv preprint arXiv:2310.01434 , 2023

2023 arXiv
[15]

Dataset cartography: Mapping and diagnosing datasets with training dynamics,

S. Swayamdipta, R. Schwartz, N. Lourie, Y . Wang, H. Hajishirzi, N. A. Smith, and Y . Choi, “Dataset cartography: Mapping and diagnosing datasets with training dynamics,” arXiv preprint arXiv:2009.10795 , 2020

2009 arXiv
[16]

Neurosurgeon: Collaborative intelligence between the cloud and mobile edge,

Y . Kang, J. Hauswald, C. Gao, A. Rovinski et al. , “Neurosurgeon: Collaborative intelligence between the cloud and mobile edge,” in ACM Computer Architecture News, vol. 45, 2017, pp. 615–629

2017
[17]

Bottlenet: A deep learning architecture for intelligent mobile cloud computing services,

A. E. Eshratifar, A. Esmaili, and M. Pedram, “Bottlenet: A deep learning architecture for intelligent mobile cloud computing services,” in 2019 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED). IEEE, 2019, pp. 1–6

2019
[18]

Bottlefit: Learning compressed representations in deep neural networks for effective and efficient split computing,

Y . Matsubara, D. Callegaro, S. Singh, M. Levorato, and F. Restuccia, “Bottlefit: Learning compressed representations in deep neural networks for effective and efficient split computing,” in 2022 IEEE 23rd Inter- national Symposium on a World of Wireless, Mobile and Multimedia...

2022
[19]

Bottlenet++: An end-to-end approach for feature compression in device-edge co-inference systems,

J. Shao and J. Zhang, “Bottlenet++: An end-to-end approach for feature compression in device-edge co-inference systems,” in 2020 IEEE Inter- national Conference on Communications Workshops (ICC Workshops) . IEEE, 2020, pp. 1–6

2020
[20]

Fast and accurate streaming cnn infer- ence via communication compression on the edge,

D. Hu and B. Krishnamachari, “Fast and accurate streaming cnn infer- ence via communication compression on the edge,” in 2020 IEEE/ACM Fifth International Conference on Internet-of-Things Design and Imple- mentation (IoTDI). IEEE, 2020, pp. 157–163

2020
[21]

Dis- tilled split deep neural networks for edge-assisted real-time systems,

Y . Matsubara, S. Baidya, D. Callegaro, M. Levorato, and S. Singh, “Dis- tilled split deep neural networks for edge-assisted real-time systems,” in Proceedings of the 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges , 2019, pp. 21–26

2019
[22]

Cut, distil and encode (cde): Split cloud-edge deep inference,

M. Sbai, M. R. U. Saputra, N. Trigoni, and A. Markham, “Cut, distil and encode (cde): Split cloud-edge deep inference,” in 2021 18th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON). IEEE, 2021, pp. 1–9

2021
[23]

Deep compressive offloading: Speeding up neural network inference by trading edge computation for network latency,

S. Yao, J. Li, D. Liu, T. Wang, S. Liu, H. Shao, and T. Abdelzaher, “Deep compressive offloading: Speeding up neural network inference by trading edge computation for network latency,” in Proceedings of the 18th conference on embedded networked sensor systems , 2020, pp. 476–488

2020
[24]

Neural compression and filtering for edge-assisted real-time object detection in challenged networks,

Y . Matsubara and M. Levorato, “Neural compression and filtering for edge-assisted real-time object detection in challenged networks,” in 2020 25th International Conference on Pattern Recognition (ICPR) . IEEE, 2021, pp. 2272–2279

2020
[25]

Early-exit deep neural networks for distorted images: providing an efficient edge offloading,

R. G. Pacheco, F. D. V . R. Oliveira, and R. S. Couto, “Early-exit deep neural networks for distorted images: providing an efficient edge offloading,” in IEEE Global Communications Conf. (GLOBECOM) , 2021, pp. 1–6

2021
[26]

Learning early exit for deep neural network inference on mobile devices through multi-armed bandits,

W. Ju, W. Bao, D. Yuan, L. Ge, and B. B. Zhou, “Learning early exit for deep neural network inference on mobile devices through multi-armed bandits,” in 2021 IEEE/ACM 21st International Symposium on Cluster, Cloud and Internet Computing (CCGrid) . IEEE, 2021, pp. 11–20

2021
[27]

Dynamic early exit scheduling for deep neural network inference through contextual bandits,

W. Ju, W. Bao, L. Ge, and D. Yuan, “Dynamic early exit scheduling for deep neural network inference through contextual bandits,” in Proceedings of the 30th ACM International Conference on Information & Knowledge Management , 2021, pp. 823–832

2021
[28]

Unsupervised early exit in dnns with multiple exits,

H. N. N U, M. K. Hanawal, and A. Bhardwaj, “Unsupervised early exit in dnns with multiple exits,” in Proceedings of the Second International Conference on AI-ML Systems, ser. AIMLSystems ’22. New York, NY , USA: Association for Computing Machinery, 2023

2023
[29]

Splitee: Early exit in deep neural networks with split computing,

D. J. Bajpai, V . K. Trivedi, S. L. Yadav, and M. K. Hanawal, “Splitee: Early exit in deep neural networks with split computing,” arXiv preprint arXiv:2309.09195, 2023

2023 arXiv
[30]

I-splitee: Image clas- sification in split computing dnns with early exits,

D. J. Bajpai, A. Jaiswal, and M. K. Hanawal, “I-splitee: Image clas- sification in split computing dnns with early exits,” arXiv preprint arXiv:2401.10541, 2024

2024 arXiv
[31]

Finite-time analysis of the multiarmed bandit problem,

P. Auer et al., “Finite-time analysis of the multiarmed bandit problem,” Machine Learning, vol. 47, pp. 235–256, 2002

2002
[32]

Adam: A method for stochastic optimization,

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

2014 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

A survey on vision transformer,

K. Han, Y . Wang, H. Chen, X. Chen, J. Guo, Z. Liu, Y . Tang, A. Xiao, C. Xu, Y . Xuet al., “A survey on vision transformer,” IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 1, pp. 87–110, 2022

2022

[2] [2]

GLUE: A multi-task benchmark and analysis platform for natural language understanding,

A. Wang, A. Singh, J. Michael, F. Hill, O. Levy, and S. R. Bowman, “GLUE: A multi-task benchmark and analysis platform for natural language understanding,” 2019, in the Proceedings of ICLR

2019

[3] [3]

To prune, or not to prune: exploring the efficacy of pruning for model compression,

M. Zhu and S. Gupta, “To prune, or not to prune: exploring the efficacy of pruning for model compression,” arXiv preprint arXiv:1710.01878 , 2017

arXiv 2017

[4] [4]

Are sixteen heads really better than one?

P. Michel, O. Levy, and G. Neubig, “Are sixteen heads really better than one?” Advances in neural information processing systems , vol. 32, 2019

work page 2019

[5] [5]

Ternarybert: Distillation-aware ultra-low bit bert,

W. Zhang, L. Hou, Y . Yin, L. Shang, X. Chen, X. Jiang, and Q. Liu, “Ternarybert: Distillation-aware ultra-low bit bert,” arXiv preprint arXiv:2009.12812, 2020

arXiv 2009

[6] [6]

I-bert: Integer-only bert quantization,

S. Kim, A. Gholami, Z. Yao, M. W. Mahoney, and K. Keutzer, “I-bert: Integer-only bert quantization,” in International conference on machine learning. PMLR, 2021, pp. 5506–5518

2021

[7] [7]

Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter,

V . Sanh, L. Debut, J. Chaumond, and T. Wolf, “Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter,” arXiv preprint arXiv:1910.01108, 2019

arXiv 1910

[8] [8]

Tinybert: Distilling bert for natural language understanding,

X. Jiao, Y . Yin, L. Shang, X. Jiang, X. Chen, L. Li, F. Wang, and Q. Liu, “Tinybert: Distilling bert for natural language understanding,” arXiv preprint arXiv:1909.10351 , 2019

arXiv 1909