Pith. sign in

REVIEW 4 major objections 6 minor 34 references

Continuous GNN-based Anomaly Detection on Edge using Efficient Adaptive Knowledge Graph Learning

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read This paper claims that a deployed video anomaly detector can keep adapting to shifting anomaly types by rewriting only the token embeddings of its own knowledge graph—pruning and recreating nodes as concepts drift—without any cloud-based…

desk verdict A plausible continuous KG adaptation mechanism for edge VAD, but the experiments validate a supervised version of the loop, not the unsupervised deployment claim. read the letter →

arxiv 2411.09072 v2 pith:LLRFTSPJ submitted 2024-11-13 cs.LG

classification cs.LG
keywords videoanomalydetectionknowledgegraphneuralnetworkedgecomputingcontinuouslearningadaptivetokenembeddingupdateinterpretableretrieval
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 tries to establish that a lightweight graph-neural-network video anomaly detector can keep adapting to new anomaly types after deployment on an edge device, without cloud-based regeneration of its knowledge graph. The mechanism is to treat the most anomalous recent frames as pseudo-labels and update only the token embeddings of graph concept nodes; a node whose embedding drifts away from its concept is pruned and replaced. If the claim holds, deployed detectors can track evolving anomaly trends at a small fraction of the compute and memory cost of regenerating the graph with a large language model, at a small accuracy cost. The experiments on the UCF-Crime surveillance benchmark show test AUC dropping at an anomaly-target shift and then recovering, with weak shifts recovering faster than strong shifts.

What carries the argument

The central object is a mission-specific knowledge graph: a hierarchical directed acyclic graph in which each node is a short-text concept carrying a token embedding, and edges connect only nodes in adjacent levels. A hierarchical GNN propagates a frame embedding from a sensor node up to an embedding node, producing the anomaly decision. The adaptive loop is the load-bearing mechanism: after deployment, only the token embeddings of graph nodes are trainable; the top-$K$ anomalous recent frames provide the loss; and an $L_2$ divergence test on each node decides whether to prune it and create a replacement node with a random embedding and random connections at the same level. Interpretability comes from decoding each updated embedding back to its nearest tokens in the byte-pair-encoding vocabulary. The graph structure acts as the system's memory, and embedding drift is the signal that triggers structural change.

What would settle it

Run the adaptation loop on a shifted anomaly type using the paper's score-based top-$K$ selection rather than labeled samples, and track whether the retrieved concept words converge to the new anomaly and whether test AUC recovers. If the highest-scoring frames are mostly false positives or a different anomaly class, the pruning-and-creation drift will move the graph to the wrong concepts, showing that the self-adaptation claim fails under its own selection rule.

Watch

Extended reading notes

Core claim

The paper claims that an anomaly detector built as a lightweight GNN over a hierarchical knowledge graph can continue learning after deployment by rewriting the graph itself, entirely on an edge device. The adaptation signal is the recent anomaly-score distribution: the top-$K$ frames by score within the last $N$ frames are treated as anomalies, and backpropagating the decision loss on them updates only the token embeddings of knowledge-graph nodes. After each update, the method compares old and new embeddings with the $L_2$ distance; a node whose distance grows is considered to have drifted and is pruned together with its edges, and a new node with a random token embedding and random edges is created at the same level. On the UCF-Crime benchmark, with anomaly targets shifted between stealing, robbery, and explosion, the authors report that test AUC (area under the ROC curve) drops at the shift and then recovers, and that the adapted graph remains interpretable—for example, a node initially meaning 'Sneaky' converges toward 'Firearm' when the target shifts from stealing to robbery. Their cost comparison shows roughly $3\times 10^{10}$ FLOPs per month and about 5 joules per update on the device, versus $4\times 10^{15}$ FLOPs and 200 GB memory for cloud regeneration, at a slight AUC cost of about 0.91 versus 0.93.

Load-bearing premise

The method assumes that the $K$ most anomalous recent frames are genuine examples of the newly emerging anomaly type, so that updating only the graph's token embeddings will steer the concept nodes toward that type.

Editorial extensions

If this is right

  • After deployment, the detector can self-adjust to new anomaly trends without cloud connectivity, removing the need for monthly LLM-based knowledge-graph regeneration and its associated bandwidth and latency.
  • Only the knowledge-graph token embeddings are updated during adaptation; the GNN weights and the joint embedding model remain frozen.
  • A node whose token embedding moves away from its previous concept is pruned with its edges, and a new node with a random embedding and random edges is created at the same level.
  • The adapted knowledge graph stays human-readable: retrieving nearest vocabulary tokens shows concepts drifting, such as a 'Sneaky' node moving toward 'Firearm' when the anomaly target shifts from stealing to robbery.
  • The measured resource gap is large: roughly 30 billion FLOPs per month on the edge device and about 5 joules per update, compared with 4 quadrillion FLOPs and 200 GB memory for cloud regeneration, at a small AUC cost.

Reading between the lines

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

  • The same drift-and-prune loop could in principle adapt to shifts other than anomaly class, such as changes in normal behavior or sensor conditions, because the mechanism only watches embedding drift rather than a class label.
  • The approach presumes the joint embedding space stays fixed; if the underlying encoder were ever updated on-device, node pruning and creation would need to be coordinated with that update rather than run as an independent loop.
  • The nearest-token decoding gives operators an evolving audit trail of what concepts the detector currently believes it is looking for, which could support monitoring and trust decisions in deployed surveillance.
  • Because the evaluation feeds labeled samples of the shifted anomaly type into the adaptation loop, a natural next experiment is to run the loop with the paper's score-based top-$K$ selection and compare whether AUC recovery and concept convergence still hold.
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.

Referee Report

4 major / 6 minor

Summary. The paper proposes a continuous-learning extension of MissionGNN for video anomaly detection on edge devices. The method first generates a mission-specific knowledge graph with an LLM, trains a hierarchical GNN decision model, and then, at deployment, continues updating only the KG token embeddings. Frames are selected as pseudo-anomalies by taking the top-K highest anomaly scores within the last N frames; a three-phase prune/alter/create mechanism modifies the KG, and an interpretable retrieval step maps adapted embeddings to BPE vocabulary words. Experiments on UCF-Crime track test AUC across shifts from Stealing to Robbery (weak shift) and Stealing to Explosion (strong shift), and a computational table compares edge-based adaptation with cloud-based KG regeneration.

Significance. If validated, the framework would be a useful contribution: it removes the cloud-regeneration bottleneck of MissionGNN, keeps the encoder and GNN frozen, and produces human-readable KG updates. The interpretable retrieval mechanism is an appealing feature, and the computational-efficiency numbers (3e10 FLOPs/month versus 4e15 FLOPs/month) are striking. However, the experimental protocol does not exercise the unsupervised deployment loop that is the paper's central claim; the reported results support at most a supervised variant. No code or detailed numerical results are provided, so independent verification is not possible. The significance is therefore conditional on closing the gap between the claimed mechanism and the evaluation.

major comments (4)
  1. [Section IV-B] The adaptation evaluation does not exercise the deployed unsupervised mechanism. Section IV-B states that 'KG adaptive learning is performed using samples of the initially selected target anomaly type and corresponding non-anomalous samples from the training set' and that after the shift 'KG adaptation continues under the new anomaly type.' This provides the adaptation loop with labels for the shifted anomaly type. In contrast, Section III-D describes an unsupervised deployment loop that selects 'the top K data points with the highest anomaly scores within the most recent N data points' and considers them as anomalies without labels. The reported Figure 5 therefore validates a supervised version of the mechanism, not the edge-deployment mechanism promised in the abstract and in Section III-A. Please add an experiment that runs the deployed loop on unlabeled frames using top-K selection, or explicitly limit the claims to supervised adaptation.
  2. [Section III-D] The update objective for the top-K pseudo-anomalies is unspecified. After selecting the frames, the text says to 'compute loss functions' and perform backpropagation to update KG token embeddings, while all other weights are frozen. The decision model f_dec in Eq. (5) outputs n+1 probabilities, and the architecture has no output neuron for a shifted anomaly type before adaptation. If the loss is the original decision loss, a class target is required but is not available in an unlabeled stream; if the loss is only an anomaly-score loss, there is no mechanism to steer an embedding such as 'Sneaky' toward 'Firearm' as reported in Figure 6. Please specify the exact loss, the target variables, and how the new anomaly type's concept is identified during deployment.
  3. [Section III-D] The pseudo-anomaly selection rule is self-referential and under-tested. The paper sets K = |Δm| × N and says that t' and N 'should be tuned using a validation set,' so the pseudo-anomaly set is a fitted function of the model's own anomaly-score statistics. There is no experiment showing that the top-K frames isolate the new anomaly type rather than normal drift or unrelated false positives. In addition, the distance-convergence criterion in Figure 4 ('Distance is Converging?') is not formalized, and node creation uses a 'random token embedding' with 'random edge connections'; these degrees of freedom are never swept or ablated. Please provide a precise algorithm with thresholds and update schedules, and report a sensitivity analysis over K, the convergence criterion, and the random initialization.
  4. [Figure 5 and Table I] The quantitative evidence for the central performance claim is incomplete. Figure 5 plots 'Test AUC Score' on an axis from 0 to 1 but reports no numeric values, error bars, seed counts, or a definition of 'Category 1-4' on the x-axis. Table I reports only point averages of 0.91 (proposed) versus 0.93 (baseline) with no confidence intervals or number of runs. The claim that the method 'maintains high anomaly detection performance' is therefore not quantitatively established. Please report numerical AUC values, variance across runs, and statistical comparisons, or explicitly state that the differences are not statistically evaluated.
minor comments (6)
  1. [Abstract and Figure 1] The terminology is inconsistent: the abstract and Figure 1 use 'alternating' while Section III-D uses 'altering'; please choose one term and use it throughout.
  2. [Section II-B] There is a typo: 'wich' should be 'which.'
  3. [Section IV-A.1] The loss balance coefficients λspa and λsmt are set to 0.001, but no loss equation in the paper defines these coefficients; please add the corresponding loss terms.
  4. [Section III-D] The text says the token update uses 'the mechanism explained in Figure 3,' but Figure 3 describes LLM-based KG generation, not embedding updates; please correct the cross-reference.
  5. [Figure 5] Please clarify which row of Figure 5 corresponds to the weak shift and which to the strong shift, and define the four 'Category' steps on the x-axis.
  6. [Table I] The 'Edge Device Computational Cost per Adaptation (FLOPs/day)' entry is asserted without indicating whether it is measured or estimated; please clarify the source of this number.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the adaptive KG learning loop is a heuristic self-training mechanism, and the labeled-sample evaluation is a validity gap rather than a derivation that reduces to its inputs.

full rationale

The paper's central derivation is not circular. The continuous KG adaptation mechanism in Section III-D selects top-K frames by anomaly score, computes a loss, and backpropagates to KG token embeddings. This is an operational self-training heuristic; K = |Δm|×N is a heuristic formula and the top-K rule is a pseudo-labeling procedure, not an equation in which the predicted quantity is defined in terms of the fitted quantity. The experimental protocol in Section IV-B indeed uses labeled samples of the target anomaly type during adaptation, so it does not validate the unsupervised deployment loop; however, that is an external-validity limitation, not circularity. MissionGNN [16] is cited for the base KG generation and hierarchical GNN architecture; although the authors overlap, the citation supplies the foundation architecture and is not invoked as a uniqueness theorem or as the source of the new adaptation result. No fitted parameter is renamed as a prediction, and no conclusion is forced by the paper's own definitions. The skepticism about unmatched evaluation is a correctness concern and should be recorded separately from circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a chain of domain assumptions: the MissionGNN baseline works, ImageBind aligns modalities well enough for token-embedding gradients to transfer, the top-K score selection identifies the new anomaly class, and random node creation improves the graph. No external benchmark or ablation independently validates the adaptive components, and the evaluation protocol substitutes labeled data for the unsupervised top-K loop.

free parameters (4)
  • K (top-K anomaly count) = K = |Δm| × N; N and t' validation-tuned, values not reported
    The number of frames treated as anomalies each update is derived from the model's own mean anomaly-score change, and the window parameters are tuned on a validation set (Section III-D).
  • Distance convergence criterion = not specified
    The pruning decision compares L2 distance between old and updated token embeddings and prunes when distance increases, but no threshold, measurement schedule, or seeds are given (Section III-D, Figure 4).
  • Loss balance coefficients λspa and λsmt = 0.001
    Training hyperparameters inherited from the MissionGNN pipeline; they are part of the fitted system, though not central to the adaptation claim (Section IV-A1).
  • Number and placement of randomly created nodes and edges = not specified
    Node creation adds a new node with a random token embedding and random edge connections, but the amount and placement are unspecified and not ablated (Section III-D).
assumptions (5)
  • domain assumption MissionGNN's LLM-generated hierarchical KG and GNN reasoning are a valid, working baseline.
    The proposed method assumes the fixed KG and GNN from the self-cited MissionGNN paper [16] perform well; no independent reproduction is supplied.
  • domain assumption ImageBind's joint embedding space aligns video frames and BPE token embeddings well enough for KG token updates to transfer to visual anomaly detection.
    The adaptive loop updates token embeddings while freezing the image encoder, which assumes gradients through the frozen joint embedding model are informative (Sections III-C and III-D).
  • ad hoc to paper Top-K high anomaly-score frames within the last N frames are the newly emerging anomaly type, not normal drift.
    Section III-D defines K from a mean score change and treats those frames as anomalies, but this is an unverified heuristic and is not tested in the evaluation.
  • ad hoc to paper Randomly created nodes with random edges move the KG toward useful new concepts after embedding updates.
    Section III-D creates nodes with random token embeddings and random edges with no justification or ablation showing this helps adaptation.
  • domain assumption GPT-4 and ConceptNet generate mission-relevant KG structure.
    Section III-B relies on LLM generation and ConceptNet; if the initial KG is poor or missing relevant concepts, the adaptation mechanism may not recover useful semantics.
invented entities (1)
  • Randomly initialized KG nodes and edges
    purpose: Replace pruned nodes to give the KG new capacity to represent shifted anomaly concepts
    Created with random token embeddings and random edge connections (Section III-D); no independent evidence that they encode meaningful concepts beyond the single qualitative retrieval example in Figure 6.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Continuous GNN-based Anomaly Detection on Edge using Efficient Adaptive Knowledge Graph Learning." pith.science (2026). https://pith.science/paper/LLRFTSPJ

@misc{pith2026241109072,
  author       = {Pith},
  title        = {Pith review of: Continuous GNN-based Anomaly Detection on Edge using Efficient Adaptive Knowledge Graph Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LLRFTSPJ}},
  note         = {Machine review of arXiv:2411.09072}
}
read the original abstract

The increasing demand for robust security solutions across various industries has made Video Anomaly Detection (VAD) a critical task in applications such as intelligent surveillance, evidence investigation, and violence detection. Traditional approaches to VAD often rely on finetuning large pre-trained models, which can be computationally expensive and impractical for real-time or resource-constrained environments. To address this, MissionGNN introduced a more efficient method by training a graph neural network (GNN) using a fixed knowledge graph (KG) derived from large language models (LLMs) like GPT-4. While this approach demonstrated significant efficiency in computational power and memory, it faces limitations in dynamic environments where frequent updates to the KG are necessary due to evolving behavior trends and shifting data patterns. These updates typically require cloud-based computation, posing challenges for edge computing applications. In this paper, we propose a novel framework that facilitates continuous KG adaptation directly on edge devices, overcoming the limitations of cloud dependency. Our method dynamically modifies the KG through a three-phase process: pruning, alternating, and creating nodes, enabling real-time adaptation to changing data trends. This continuous learning approach enhances the robustness of anomaly detection models, making them more suitable for deployment in dynamic and resource-constrained environments.

Figures

Figures reproduced from arXiv: 2411.09072 by the authors.

Figure 1
Figure 1. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall framework for our proposed KG adaptive learning framework utilizing the novel concept of hierarchical graph neural network. Duplicated Concepts – nodes that represent concepts already present in previous levels – and Invalid Edges – connections that violate the rule that edges only connect nodes from level i to level i + 1. If no errors are detected, the process proceeds to generate the next layer. If er… view at source ↗
Figure 3
Figure 3. Detailed description of our mission-specific knowledge graph generation framework. embedding is located in the embedding node. Each GNN layer Gl(.) consists of five sub-layers: a dense layer, a hierarchical message passing layer, a hierarchical aggregate layer, a batch normalization layer, and an activation layer. First, the dense layer (ϕl : R |V |×Dl−1 → R |V |×Dl ), which refines the embedding space to more accur… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Pipeline of our proposed knowledge graph modification mechanism. ⃗st,i. Here, pA(Ft) = 1−pN (Ft) represents the probability that the frame is abnormal, and pi|A(Ft) = ⃗st,i 1−pN (Ft) represents the conditional probability of anomaly type i given that the frame is abnor…
Figure 5
Figure 5. Figure 5: Test AUC scores measured across shifts in the anomaly target, comparing different types of anomalies. TABLE I DETAILED COMPUTATIONAL AND PERFORMANCE COMPARISON BETWEEN BASELINE AND PROPOSED METHODS Metric Baseline Method Proposed Method (Cloud-based KG Updates) (Edge-b…
Figure 6
Figure 6. Figure 6: Qualitative evaluation of knowledge updates within the proposed knowledge graph update loops, with each number representing the iteration count. We focus on key metrics such as computational cost, memory usage, and anomaly detection performance. Table I presents a comp…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 21 canonical work pages

  1. [1]

    Hierarchical scene normality-binding modeling for anomaly detection in surveillance videos,

    Q. Bao, F. Liu, Y . Liu, L. Jiao, X. Liu, and L. Li, “Hierarchical scene normality-binding modeling for anomaly detection in surveillance videos,” in Proceedings of the 30th ACM International Conference on Multimedia, pp. 6103–6112, 2022

  2. [2]

    Convolutional transformer based dual discriminator generative adversarial networks for video anomaly detection,

    X. Feng, D. Song, Y . Chen, Z. Chen, J. Ni, and H. Chen, “Convolutional transformer based dual discriminator generative adversarial networks for video anomaly detection,” in Proceedings of the 29th ACM International Conference on Multimedia , pp. 5546–5554, 2021

  3. [3]

    Soft computing techniques-based digital video forensics for fraud medical anomaly detec- tion,

    S. K. Nanda, D. Ghai, P. Ingole, and S. Pande, “Soft computing techniques-based digital video forensics for fraud medical anomaly detec- tion,” Computer Assisted Methods in Engineering and Science , vol. 30, no. 2, pp. 111–130, 2022

  4. [4]

    A real time crime scene intelligent video surveillance systems in violence detection framework using deep learning techniques,

    K. B. Sahay, B. Balachander, B. Jagadeesh, G. A. Kumar, R. Kumar, and L. R. Parvathy, “A real time crime scene intelligent video surveillance systems in violence detection framework using deep learning techniques,” Computers and Electrical Engineering , vol. 103, p. 108319, 2022

  5. [5]

    An iot enable anomaly detection system for smart city surveillance,

    M. Islam, A. S. Dukyil, S. Alyahya, and S. Habib, “An iot enable anomaly detection system for smart city surveillance,” Sensors, vol. 23, no. 4, p. 2358, 2023

  6. [6]

    A survey on deep learning techniques for video anomaly detection,

    J. J. P. Suarez and P. C. Naval Jr, “A survey on deep learning techniques for video anomaly detection,” arXiv preprint arXiv:2009.14146 , 2020

  7. [7]

    Real-world anomaly detection in surveillance videos,

    W. Sultani, C. Chen, and M. Shah, “Real-world anomaly detection in surveillance videos,” in The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

  8. [8]

    Not only look, but also listen: Learning multimodal violence detection under weak supervision,

    P. Wu, j. Liu, Y . Shi, Y . Sun, F. Shao, Z. Wu, and Z. Yang, “Not only look, but also listen: Learning multimodal violence detection under weak supervision,” in European Conference on Computer Vision (ECCV), 2020

Show all 34 references
  1. [9]

    Future frame prediction for anomaly detection – a new baseline,

    W. Liu, D. L. W. Luo, and S. Gao, “Future frame prediction for anomaly detection – a new baseline,” in 2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2018

  2. [10]

    Scale-aware spatio-temporal relation learning for video anomaly detection,

    G. Li, G. Cai, X. Zeng, and R. Zhao, “Scale-aware spatio-temporal relation learning for video anomaly detection,” in European Conference on Computer Vision , pp. 333–350, Springer, 2022

  3. [11]

    Real-world anomaly detection in surveillance videos,

    W. Sultani, C. Chen, and M. Shah, “Real-world anomaly detection in surveillance videos,” in Proceedings of the IEEE conference on computer vision and pattern recognition , pp. 6479–6488, 2018

  4. [12]

    Weakly-supervised video anomaly detection with robust temporal feature magnitude learning,

    Y . Tian, G. Pang, Y . Chen, R. Singh, J. W. Verjans, and G. Carneiro, “Weakly-supervised video anomaly detection with robust temporal feature magnitude learning,” in Proceedings of the IEEE/CVF international conference on computer vision , pp. 4975–4986, 2021

  5. [13]

    Learning causal temporal relation and feature discrim- ination for anomaly detection,

    P. Wu and J. Liu, “Learning causal temporal relation and feature discrim- ination for anomaly detection,” IEEE Transactions on Image Processing , vol. 30, pp. 3513–3527, 2021

  6. [14]

    Delving into clip latent space for video anomaly recognition,

    L. Zanella, B. Liberatori, W. Menapace, F. Poiesi, Y . Wang, and E. Ricci, “Delving into clip latent space for video anomaly recognition,” arXiv preprint arXiv:2310.02835, 2023

  7. [15]

    Batchnorm- based weakly supervised video anomaly detection,

    Y . Zhou, Y . Qu, X. Xu, F. Shen, J. Song, and H. Shen, “Batchnorm- based weakly supervised video anomaly detection,” arXiv preprint arXiv:2311.15367, 2023

  8. [16]

    Missiongnn: Hierarchi- cal multimodal gnn-based weakly supervised video anomaly recogni- tion with mission-specific knowledge graph generation,

    S. Yun, R. Masukawa, M. Na, and M. Imani, “Missiongnn: Hierarchi- cal multimodal gnn-based weakly supervised video anomaly recogni- tion with mission-specific knowledge graph generation,” arXiv preprint arXiv:2406.18815, 2024

  9. [17]

    Pv-vtt: A privacy- centric dataset for mission-specific anomaly detection and natural lan- guage interpretation,

    R. Masukawa, S. Yun, Y . Yamaguchi, and M. Imani, “Pv-vtt: A privacy- centric dataset for mission-specific anomaly detection and natural lan- guage interpretation,” arXiv preprint arXiv:2410.22623 , 2024

  10. [18]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

  11. [19]

    Autolabel: Clip-based framework for open-set video domain adaptation,

    G. Zara, S. Roy, P. Rota, and E. Ricci, “Autolabel: Clip-based framework for open-set video domain adaptation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. 11504– 11513, 2023

  12. [20]

    Taskclip: Extend large vision-language model for task oriented object detection,

    H. Chen, W. Huang, Y . Ni, S. Yun, Y . Liu, F. Wen, A. Velasquez, H. Latapie, and M. Imani, “Taskclip: Extend large vision-language model for task oriented object detection,” arXiv preprint arXiv:2403.08108 , 2024

  13. [21]

    Expanding event modality applications through a robust clip-based encoder,

    S. Jeong, H. Chen, S. Yun, S. Cho, W. Huang, X. Liu, and M. Imani, “Expanding event modality applications through a robust clip-based encoder,” arXiv preprint arXiv:2412.03093 , 2024

  14. [22]

    The emergence of edge computing,

    M. Satyanarayanan, “The emergence of edge computing,” Computer, vol. 50, no. 1, pp. 30–39, 2017

  15. [23]

    Edge computing: Vision and challenges,

    W. Shi, J. Cao, Q. Zhang, Y . Li, and L. Xu, “Edge computing: Vision and challenges,” IEEE internet of things journal , vol. 3, no. 5, pp. 637–646, 2016

  16. [24]

    Hypersense: Hyperdi- mensional intelligent sensing for energy-efficient sparse data processing,

    S. Yun, H. Chen, R. Masukawa, H. Errahmouni Barkam, A. Ding, W. Huang, A. Rezvani, S. Angizi, and M. Imani, “Hypersense: Hyperdi- mensional intelligent sensing for energy-efficient sparse data processing,” Advanced Intelligent Systems , p. 2400228, 2024

  17. [25]

    Ex- ploiting boosting in hyperdimensional computing for enhanced reliability in healthcare,

    S. Jeong, H. E. Barkam, S. Yun, Y . Kim, S. Angizi, and M. Imani, “Ex- ploiting boosting in hyperdimensional computing for enhanced reliability in healthcare,” arXiv preprint arXiv:2411.14612 , 2024

  18. [26]

    Edge intelligence for mission cognitive wireless emergency networks,

    L. Wang, J. Zhang, J. Chuan, R. Ma, and A. Fei, “Edge intelligence for mission cognitive wireless emergency networks,” IEEE Wireless Communications, vol. 27, no. 4, pp. 103–109, 2020

  19. [27]

    Smart video surveillance system based on edge computing,

    A. C. Cob-Parro, C. Losada-Guti ´errez, M. Marr ´on-Romera, A. Gardel- Vicente, and I. Bravo-Mu˜noz, “Smart video surveillance system based on edge computing,” Sensors, vol. 21, no. 9, p. 2958, 2021

  20. [28]

    Hdreason: Algorithm-hardware code- sign for hyperdimensional knowledge graph reasoning,

    H. Chen, Y . Ni, A. Zakeri, Z. Zou, S. Yun, F. Wen, B. Khaleghi, N. Srini- vasa, H. Latapie, and M. Imani, “Hdreason: Algorithm-hardware code- sign for hyperdimensional knowledge graph reasoning,” arXiv preprint arXiv:2403.05763, 2024

  21. [29]

    Reliable hyperdimensional reasoning on unreliable emerging technologies,

    H. E. Barkam, S. Yun, H. Chen, P. Gensler, A. Mema, A. Ding, G. Mich- elogiannakis, H. Amrouch, and M. Imani, “Reliable hyperdimensional reasoning on unreliable emerging technologies,” in 2023 IEEE/ACM International Conference on Computer Aided Design (ICCAD) , pp. 1– 9, IEEE, 2023

  22. [30]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,” International Journal of Computer Vision , vol. 130, no. 9, pp. 2337–2348, 2022

  23. [31]

    Imagebind: One embedding space to bind them all,

    R. Girdhar, A. El-Nouby, Z. Liu, M. Singh, K. V . Alwala, A. Joulin, and I. Misra, “Imagebind: One embedding space to bind them all,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 15180–15190, 2023

  24. [32]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al., “Learning transferable visual models from natural language supervision,” in International con- ference on machine learning , pp. 8748–8763, PMLR, 2021

  25. [33]

    Neural machine translation of rare words with subword units,

    R. Sennrich, B. Haddow, and A. Birch, “Neural machine translation of rare words with subword units,” in Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (K. Erk and N. A. Smith, eds.), (Berlin, Germany), pp. 1715–1...

  26. [34]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” arXiv preprint arXiv:1711.05101 , 2017

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.