Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Silent Neuron Theory and Plasticity Preservation for Deep Reinforcement Learning in Adaptive Video Streaming

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims plasticity loss in deep RL is a silent-neuron problem, and resetting those neurons recovers learning.

desk verdict The central theorem is false and the empirical support is thin, but the dual-criterion reset heuristic and the streaming application are worth a second look. read the letter →

arxiv 2505.01584 v4 pith:Q3UQ5ZSU submitted 2025-05-02 cs.LG cs.AI

Zhiqiang He , Zhi Liu This is my paper · ORCID
classification cs.LGcs.AI
keywords silentneurontheoryneuralplasticitylossdormantphenomenonresetadaptivevideostreamingnon-stationaryreinforcementlearningqualityofexperiencenetworkresourceadaptation
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

The paper's central claim is that plasticity loss in deep reinforcement learning is not captured by 'dormant neurons'—units with near-zero forward output—because such units can still participate in learning through nonzero backward gradients. It defines a doubly inactive 'silent neuron' as one whose expected output and expected gradient are simultaneously near zero, and proves a characterization (Theorem 4.5) tying a combined activity index to those two conditions. On that basis it builds Reset Silent Neuron (ReSiN), which periodically reinitializes silent units, and reports that in adaptive video streaming under shifting bandwidth it achieves up to 168% higher bitrate and 108% better QoE than existing solutions with comparable smoothness. The potential payoff is a mechanism-level cure for a common failure: networks that stop adapting when the environment changes, without needing to know the network statistics in advance.

What carries the argument

The load-bearing object is the activity index $\xi_{l,i} = \frac{\mathbb{E}_{x\in\mathcal{D}}|h_{l,i}(x)|\,\mathbb{E}_{x\in\mathcal{D}}|g_{l,i}(x)|}{\frac{1}{H_l}\sum_j \mathbb{E}_{x\in\mathcal{D}}|h_{l,j}(x)|}$, where $h_{l,i}$ is the neuron's forward output on input $x$, $g_{l,i}$ is its gradient from the aggregated network output, and $\mathcal{D}$ is the input distribution. Theorem 4.5 says that under boundedness and non-degeneracy assumptions, $\xi_{l,i}<\epsilon$ for arbitrarily small $\epsilon$ is equivalent to $\mathbb{E}|h_{l,i}|<\sqrt{\epsilon}$ and $\mathbb{E}|g_{l,i}|<\sqrt{\epsilon}$, which is the formal definition of a silent neuron. ReSiN uses that criterion as its reset signal: it reinitializes the incoming weights of units with $\xi_{l,i}<\epsilon$ and zeroes their outgoing weights, so the theorem is what justifies resetting based on both propagation directions rather than on output dormancy alone.

What would settle it

On the same non-stationary streaming traces, run matched ablation resets at equal count: output-only dormancy, gradient-only, the product index $\xi_{l,i}$, and random selection; if the product-based rule does not outperform the others on final QoE and bitrate, the silent-neuron characterization is not what carries the improvement.

Watch

Extended reading notes

Core claim

The central discovery is that output dormancy alone overstates plasticity loss: a neuron with $\mathbb{E}_x|h_{l,i}(x)|\approx 0$ can still have substantial gradient $\mathbb{E}_x|g_{l,i}(x)|$, so it remains a participant in learning. The paper defines the activity index $\xi_{l,i}$ as the product of forward and backward expected magnitudes divided by the layer's mean output magnitude, and proves in Theorem 4.5 that for arbitrarily small $\epsilon$, $\xi_{l,i}<\epsilon$ is equivalent, under the stated regularity assumptions, to both $\mathbb{E}_x|h_{l,i}(x)|<\sqrt{\epsilon}$ and $\mathbb{E}_x|g_{l,i}(x)|<\sqrt{\epsilon}$. Silent neurons—those satisfying both—are the units the paper identifies as truly disconnected from information flow and learning. ReSiN resets exactly these units, and the empirical claim is that this reset rule preserves plasticity and raises bitrate and QoE in non-stationary adaptive streaming while remaining competitive in stationary settings.

Load-bearing premise

The argument hinges on treating a neuron that is near-silent over the observed input distribution as genuinely inert everywhere: if the training domain $\mathcal{D}$ does not represent all inputs the network will later see, a neuron classified as silent could still carry knowledge that a reset would destroy.

Editorial extensions

If this is right

  • Output-only dormant-neuron reset will misclassify units that still carry gradients, so ReSiN should dominate ReDO-style reset on the same non-stationary traces.
  • Because the reset uses only internal network states and not environment statistics, ReSiN should adapt to bandwidth shifts it has never seen, without retraining or re-tuning.
  • The same reset rule works for both policy and value networks, since the activity index is computed from aggregated outputs rather than from one loss function.
  • Performance should stay high in stationary settings too, since plasticity loss also occurs under fixed distributions as value estimates become outdated during learning.

Reading between the lines

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

  • Beyond the streaming case, the dual-pass test could serve as a general plasticity monitor in any deep RL or continual-learning system, flagging units for reset before a distribution shift causes a collapse; the paper does not test that broader use.
  • The theorem suggests a direct ranking experiment: order hidden units by $\xi_{l,i}$ and ablate resets in that order; if dual-pass ordering beats output-only ordering at matched reset budgets, the theory's mechanism is confirmed, not just its end-task performance.
  • Because the index averages over $\mathcal{D}$, it is only as representative as the replay or state buffer; in open-ended environments, ReSiN might need periodic state-coverage sampling to keep the silent-neuron detector honest. This is a practical extension, not a claim in the paper.
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

4 major / 5 minor

Summary. The paper studies plasticity loss in deep reinforcement learning for adaptive video streaming under non-stationary network conditions. It introduces "Silent Neuron" theory, defining an activity index that combines forward output and backward gradient, and claims Theorem 4.5, which states that a small activity index is equivalent to both forward and backward expectations being below sqrt(epsilon). Based on this, the paper proposes ReSiN, which resets near-silent neurons during PPO training. The evaluation uses a video-streaming simulator, comparing PPO variants and reporting up to 168% higher bitrate and 108% better QoE. The abstract also claims a tighter performance bound for ReSiN under non-stationary conditions.

Significance. The problem is timely: plasticity loss in deep RL is a recognized obstacle to continual and non-stationary learning, and a simple reset mechanism guided by both forward and backward signals is an appealing idea. If Theorem 4.5 were correct, the paper would offer a principled criterion for identifying dead units and a practical intervention. The empirical diagnostics in Figures 2 and 5, particularly the persistence of dormant neurons, are useful observations. However, the formal anchor is invalid as stated, and the experimental evidence is not strong enough to carry the central claim on its own. The paper does not provide code, numerical tables, or error bars, and the theoretical and empirical gaps are load-bearing for the advertised contributions.

major comments (4)
  1. [§4.2 and Appendix E, Theorem 4.5] The claimed equivalence is false as stated. From ξ<ε the proof only obtains E|h|·E|g| < M_h ε; a small product does not imply E|h|<√ε and E|g|<√ε without additional constant bounds. The reverse direction yields ξ < ε/m, which implies ξ<ε only if m ≥ 1, while Assumption 3 only states m>0. A concrete counterexample satisfying every stated assumption is H_l=1, E|h|=0.9, E|g|=0.0001, denominator=0.9, M_h=M_g=1, m=0.5, ε=0.1; here ξ=0.0001<ε but E|h|=0.9>√ε. Because Algorithm 1 resets exactly when ξ<ε, the theoretical justification for the reset criterion is invalid.
  2. [Appendix B, Lemma 4.2] The proof applies the mean value theorem to h_{l,i}(x+δx) for arbitrary δx and uses the fact that h_{l,i}=0 on D to claim that the difference h_{l,i}(x+δx)-h_{l,i}(x) is zero. This is only valid if x+δx ∈ D, which is not assumed; D may be a strict subset of R^{k_l}. The conclusion ∇h=0 on D therefore does not follow. Since Theorem 4.4's proof invokes Lemma 4.2, the bidirectional dormancy characterization is also unsupported.
  3. [Abstract and §4] The abstract states that "we establish a tighter performance bound for ReSiN under non-stationary network conditions," but no such bound appears anywhere in the manuscript. The only formal results are characterization lemmas for dormant and silent neurons; there is no theorem bounding the return, regret, or convergence of ReSiN. A claimed theoretical contribution that is absent cannot support the central claim.
  4. [§5, Figures 6-8] The empirical evaluation compares only PPO variants; it does not include established adaptive-bitrate baselines (e.g., Pensieve, BOLA, MPC) despite the abstract's claim of improvements over "existing solutions." The reported learning curves have no error bars, confidence intervals, or seed-level tables, and the reset thresholds ε1, ε2 and frequency F are evaluated on the same streaming benchmark used for the headline results. Without held-out conditions or quantitative summaries, the claims of up to 168% higher bitrate and 108% better QoE cannot be verified.
minor comments (5)
  1. [§3.1 and §3.2] There are typos: "chuncks" should be "chunks," and "Object Function" should be "Objective Function."
  2. [Definition 4.2 and Eq. (4.5)] The index i is overloaded: the definition uses i for the neuron, while the gradient expression Σ_i f_theta(x_i) uses i for training samples. Please use separate indices to avoid ambiguity.
  3. [Algorithm 1 and Theorem 4.5] Algorithm 1 uses separate thresholds ε1 and ε2 on ξ^g and ξ^d, whereas Theorem 4.5 concerns a single product index ξ with a threshold ε. The connection between the theorem and the algorithm's two-sided threshold should be made explicit.
  4. [Appendix C, Eq. (C.2)] Eq. (C.2) contains a garbled formula ("sl,i = 0 1/HH ..."); it should be typeset cleanly.
  5. [Figure 7] The IQM plot lacks error bars or confidence intervals; the number of seeds should be stated and the underlying numerical values reported in a table.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the defective epsilon-equivalence is an invalid proof, not a definitional restatement; the reset criterion and experiments are not forced by the definition.

full rationale

The central theoretical result, Theorem 4.5, is mathematically flawed but not circular. Definition 4.2 defines the activity index xi as a product ratio, while Theorem 4.5 asserts an equivalence between xi < epsilon and the separate factor-wise bounds E|h| < sqrt(epsilon) and E|g| < sqrt(epsilon). These are different statements: the theorem does not reduce to the definition by construction. The proof fails because a small product does not imply each factor is small, and epsilon/m does not imply epsilon when m < 1, but this is a correctness defect rather than a circularity. Moreover, Algorithm 1 does not actually use the product threshold; it resets when xi_g <= epsilon1 and xi_d <= epsilon2, directly applying the dual forward/backward criterion, so the algorithm is not derived from the invalid theorem. The reset thresholds are tuned on the same task in Appendix G, which raises a standard hyperparameter-selection concern, but the reported gains are not literally the fitted values and are not statistically forced. The only self-citation by the authors, He et al. (2025), appears in related-work discussion of two-stage optimization and is not load-bearing for the Silent Neuron claim. The abstract also promises a tighter performance bound that does not appear in the body, and code and data are not yet released, but absence of support or proof is not circularity. No quoted equation or fitted parameter can be exhibited as a reduction of the paper's prediction to its own inputs, so the appropriate circularity score is 0.

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

The central claim depends on hand-chosen reset thresholds and QoE constants, and on a set of unstated mathematical assumptions about the data domain and the stability of mid-training resets. The invented Silent Neuron concept has no independent falsifiable handle beyond the paper's own definition.

free parameters (5)
  • gradient reset threshold epsilon_1 = not reported
    ReSiN resets neurons when gradient activity satisfies xi_g <= epsilon_1; Figure 10 ablates this threshold but no final value or selection rule is given.
  • dormancy reset threshold epsilon_2 = not reported
    ReSiN also requires output activity xi_d <= epsilon_2; Figure 9 ablates this threshold and the chosen value is not stated.
  • reset frequency F = not reported
    Algorithm 1 resets every F iterations; F is never specified, and the performance claims depend on it.
  • QoE quality function constants alpha and beta = not reported
    Equation (3.6) defines perceived quality as log(alpha+b) - beta/b; the values shape the reward and are not reported.
  • QoE weighting coefficients mu_1 and mu_2 = not reported
    Equation (3.5) weights quality variation and rebuffering penalties; these values strongly affect learned behavior and are not specified.
assumptions (5)
  • domain assumption Neuron outputs and gradients are continuously differentiable, bounded in expectation, and the mean activation denominator is strictly positive (Assumptions 1-3, Sections 4.1 and 4.2).
    Used in Lemmas 4.1-4.3 and Theorems 4.4-4.5; no check is provided that the trained PPO networks satisfy these conditions.
  • ad hoc to paper The data distribution domain D can be treated as the full input space so that mean value theorem arguments can reach points outside D.
    Appendix B needs h(x+dx)=0 for arbitrary dx while only x in D is known; this is not stated and is generally false for a sampled distribution.
  • ad hoc to paper A neuron with zero gradient on D is constant on D.
    Appendix D and Lemma 4.3 apply this without requiring D to be connected; if D has multiple components, outputs can differ between components.
  • domain assumption Plasticity loss is the primary cause of PPO's performance collapse under network switching.
    Inferred from correlations in Figures 1-5; the paper does not isolate plasticity loss from policy gradient variance or optimization noise.
  • ad hoc to paper Resetting neuron weights during PPO training preserves or improves policy improvement.
    Algorithm 1 modifies the network mid-training; no convergence or stability argument is provided for the reset operation.
invented entities (1)
  • Silent Neuron (activity index xi)
    purpose: Classify a neuron as plasticity-dead when both forward output and backward gradient are near zero, and trigger reset.
    The metric is defined by the paper's Eq. (4.6); no external measurement or falsifiable prediction outside the paper's own reward and task is offered.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Silent Neuron Theory and Plasticity Preservation for Deep Reinforcement Learning in Adaptive Video Streaming." pith.science (2026). https://pith.science/paper/Q3UQ5ZSU

@misc{pith2026250501584,
  author       = {Pith},
  title        = {Pith review of: Silent Neuron Theory and Plasticity Preservation for Deep Reinforcement Learning in Adaptive Video Streaming},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q3UQ5ZSU}},
  note         = {Machine review of arXiv:2505.01584}
}
read the original abstract

Adaptive video streaming optimizes Quality of Experience (QoE) metrics by selecting appropriate bitrates according to varying network bandwidth and user demands. In practice, however, real-world network bandwidth often exhibits heterogeneity relative to training environments. Current methods predominantly tackle this problem through learning-based approaches designed to improve generalization performance. While our systematic investigation reveals a critical limitation: neural networks suffer from plasticity loss, significantly impeding their ability to adapt to heterogeneous network conditions. Through theoretical analysis of neural propagation mechanisms, we demonstrate that existing dormant neuron metrics inadequately characterize neural plasticity loss. To address this limitation, we have developed the Silent Neuron theory, which provides a more comprehensive framework for understanding plasticity degradation. Based on these theoretical insights, we propose the Reset Silent Neuron (ReSiN), which preserves neural plasticity through strategic neuron resets guided by both forward and backward propagation states. Moreover, we establish a tighter performance bound for ReSiN under non-stationary network conditions. In our implementation of an adaptive video streaming system, ReSiN has shown significant improvements over existing solutions, achieving up to 168% higher bitrate and 108% better quality of experience (QoE) while maintaining comparable smoothness. Furthermore, ReSiN consistently outperforms in stationary environments, demonstrating its robust adaptability across different network conditions.

Figures

Figures reproduced from arXiv: 2505.01584 by the authors.

Figure 1
Figure 1. Performance comparison of bitrate re￾wards in adaptive video streaming achieved by PPO under different bandwidth conditions. To demonstrate the impact of non-stationarity on optimization al￾gorithms, we conducted a systematic study using adaptive video streaming as our test scenario. This choice was made because adap￾tive video streaming presents the key challenges in network resource adaptation, such as real-time d… view at source ↗
Figure 2
Figure 2. Evolution of neural network plasticity loss. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison of PPO under different network conditions in terms of QoE metrics. The figure shows [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: System metrics under different plasticity maintenance strategies in non-stationary environments are compared in [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: The activity patterns of neurons across network layers and components, such as the ratio of dormant neurons [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Performance comparison of PPO variants across different network conditions and plasticity maintenance [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Comparison of QoE Methods using the IQM (Interquartile Mean) with a focus on the 25th-75th percentile returns. [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Dormancy patterns in policy and value networks, specifically in layers 0-1. It explores how plasticity maintenance [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Performance comparison under different dormant Reset threshold. [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]
Figure 10
Figure 10. Figure 10: Performance comparison under different gradient Reset threshold. [PITH_FULL_IMAGE:figures/full_fig_p020_10.png]
Figure 11
Figure 11. Figure 11: Performance comparison under different learning rates. [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. PRIME: Plasticity Recovery in Multi-Agent Environments for UAV-Assisted Emergency Communication Networks

    cs.MA 2026-07 conditional novelty 6.0 of 10

    PRIME recovers plasticity in shared-parameter multi-agent RL by resetting only neurons that are simultaneously forward-dormant and gradient-silent, improving IQM return by 24.9% over MAPPO under phase changes.

Reference graph

Works this paper leans on

46 extracted references · 40 canonical work pages · cited by 1 Pith paper

  1. [1]

    Alomar, P

    A. Alomar, P. Hamadanian, A. Nasr-Esfahany, A. Agarwal, M. Alizadeh, and D. Shah. Causalsim: A causal framework for unbiased trace-driven simulation. In M. Balakrishnan and M. Ghobadi, editors, NSDI, pages 1115--1147. USENIX Association, 2023

    work page 2023

  2. [2]

    Bentaleb, M

    A. Bentaleb, M. Lim, M. N. Akcay, A. C. Begen, and R. Zimmermann. Bitrate adaptation and guidance with meta reinforcement learning. IEEE Transactions on Mobile Computing, 2024

    work page 2024

  3. [3]

    S. Bi, H. Chen, X. Li, S. Wang, Y. Wu, and L. Qian. A two-stage deep reinforcement learning framework for mec-enabled adaptive 360-degree video streaming. IEEE Transactions on Mobile Computing, 2024

    work page 2024

  4. [4]

    J. Cao, W. S. Cleveland, D. Lin, and D. X. Sun. On the nonstationarity of internet traffic. In M. K. Vernon, editor, SIGMETRICS/Performance, pages 102--112. ACM, 2001. ISBN 1-58113-334-0. URL http://dblp.uni-trier.de/db/conf/sigmetrics/sigmetrics2001.html#CaoCLS01

    work page 2001

  5. [5]

    S. Deng, R. Netravali, A. Sivaraman, and H. Balakrishnan. Wifi, lte, or both? measuring multi-homed wireless internet performance. In Proceedings of the 2014 Conference on Internet Measurement Conference, pages 181--194, 2014

    work page 2014

  6. [6]

    Dohare, J

    S. Dohare, J. F. Hernandez-Garcia, Q. Lan, P. Rahman, A. R. Mahmood, and R. S. Sutton. Loss of plasticity in deep continual learning. Nature, 632 0 (8026): 0 768--774, 2024

    work page 2024

  7. [7]

    Elsayed, Q

    M. Elsayed, Q. Lan, C. Lyle, and A. R. Mahmood. Weight clipping for deep continual and reinforcement learning. Reinforcement Learning Journal, 5: 0 2198--2217, 2024

    work page 2024

  8. [8]

    Gulcehre, S

    C. Gulcehre, S. Srinivasan, J. Sygnowski, G. Ostrovski, M. Farajtabar, M. Hoffman, R. Pascanu, and A. Doucet. An empirical study of implicit regularization in deep offline RL . Transactions on Machine Learning Research, 2022. ISSN 2835-8856. URL https://openreview.net/forum?id=HFfJWx60IT

    work page 2022

Show all 46 references
  1. [9]

    Z. He, W. Qiu, W. Zhao, X. Shao, and Z. Liu. Understanding world models through multi-step pruning policy via reinforcement learning. Information Sciences, 686: 0 121361, 2025

    2025

  2. [10]

    P. Hu, Y. Chen, L. Pan, Z. Fang, F. Xiao, and L. Huang. Multi-user delay-constrained scheduling with deep recurrent reinforcement learning. IEEE/ACM Transactions on Networking, 2024

    2024

  3. [11]

    Huang, R.-X

    T. Huang, R.-X. Zhang, and L. Sun. Zwei: A self-play reinforcement learning framework for video transmission services. IEEE Transactions on Multimedia, 24: 0 1350--1365, 2021

    2021

  4. [12]

    Y. Jia, C. Zhang, Y. Huang, and W. Zhang. Lyapunov optimization based mobile edge computing for internet of vehicles systems. IEEE Transactions on Communications, 70 0 (11): 0 7418--7433, 2022

    2022

  5. [13]

    Juliani and J

    A. Juliani and J. T. Ash. A study of plasticity loss in on-policy deep reinforcement learning. arXiv preprint arXiv:2405.19153, 2024

    2024 arXiv

  6. [14]

    N. Kan, Y. Jiang, C. Li, W. Dai, J. Zou, and H. Xiong. Improving generalization for neural adaptive video streaming via meta reinforcement learning. In Proceedings of the 30th ACM International Conference on Multimedia, pages 3006--3016, 2022

    2022

  7. [15]

    Kumar, H

    S. Kumar, H. Marklund, and B. Van Roy. Maintaining plasticity in continual learning via regenerative regularization. 2023

    2023

  8. [16]

    Lebreton and K

    P. Lebreton and K. Yamagishi. Adaptive bitrate control mechanism based on long-term evaluation. In 2022 IEEE International Conference on Multimedia and Expo (ICME), pages 01--06. IEEE, 2022

    2022

  9. [17]

    Lebreton and K

    P. Lebreton and K. Yamagishi. Long-term adaptive bitrate control mechanism. IEICE Transactions on Communications, 107 0 (11): 0 817--830, 2024

    2024

  10. [18]

    H. Lee, H. Cho, H. Kim, D. Kim, D. Min, J. Choo, and C. Lyle. Slow and steady wins the race: Maintaining plasticity with hare and tortoise networks. In R. Salakhutdinov, Z. Kolter, K. Heller, A. Weller, N. Oliver, J. Scarlett, and F. Berkenkamp, editors, Proceedings of the 41s...

    2024

  11. [19]

    Lewandowski, H

    A. Lewandowski, H. Tanaka, D. Schuurmans, and M. C. Machado. Directions of curvature as an explanation for loss of plasticity, 2024. URL https://arxiv.org/abs/2312.00246

    2024 arXiv

  12. [20]

    W. Li, X. Li, Y. Xu, Y. Yang, and S. Lu. Metaabr: A meta-learning approach on adaptative bitrate selection for video streaming. IEEE Transactions on Mobile Computing, 23 0 (3): 0 2422--2437, 2023

    2023

  13. [21]

    W. Li, J. Huang, Y. Liang, Q. Su, J. Liu, W. Lyu, and J. Wang. Optimizing video streaming in dynamic networks: An intelligent adaptive bitrate solution considering scene intricacy and data budget. IEEE Transactions on Mobile Computing, 2024 a

    2024

  14. [22]

    W. Li, J. Huang, Q. Su, W. Jiang, and J. Wang. A learning-based approach for video streaming over fluctuating networks with limited playback buffers. Computer Communications, 214: 0 113--122, 2024 b

    2024

  15. [23]

    Lin and S

    J. Lin and S. Wang. Adaptive video streaming in integrated satellite-terrestrial networks: A low-complexity model predictive control approach. IEEE Wireless Communications Letters, 2024

    2024

  16. [24]

    J. Liu, J. Obando-Ceron, A. Courville, and L. Pan. Neuroplastic expansion in deep reinforcement learning. arXiv preprint arXiv:2410.07994, 2024

    2024 arXiv

  17. [25]

    Z. Liu, Q. Li, X. Chen, C. Wu, S. Ishihara, J. Li, and Y. Ji. Point cloud video streaming: Challenges and solutions. IEEE Network, 35 0 (5): 0 202--209, 2021

    2021

  18. [26]

    C. Lyle, Z. Zheng, E. Nikishin, B. A. Pires, R. Pascanu, and W. Dabney. Understanding plasticity in neural networks. In ICML, volume 202 of Proceedings of Machine Learning Research, pages 23190--23211. PMLR, 2023. URL http://dblp.uni-trier.de/db/conf/icml/icml2023.html#LyleZNPPD23

    2023

  19. [27]

    C. Lyle, Z. Zheng, K. Khetarpal, J. Martens, H. van Hasselt, R. Pascanu, and W. Dabney. Normalization and effective learning rates in reinforcement learning. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024 a . URL https://openreview.net/fo...

    2024

  20. [28]

    C. Lyle, Z. Zheng, K. Khetarpal, H. van Hasselt, R. Pascanu, J. Martens, and W. Dabney. Disentangling the causes of plasticity loss in neural networks, 2024 b . URL https://arxiv.org/abs/2402.18762

    2024 arXiv

  21. [29]

    H. Mao, R. Netravali, and M. Alizadeh. Neural adaptive video streaming with pensieve. In Proceedings of the conference of the ACM special interest group on data communication, pages 197--210, 2017

    2017

  22. [30]

    Z. Meng, Y. Guo, Y. Shen, J. Chen, C. Zhou, M. Wang, J. Zhang, M. Xu, C. Sun, and H. Hu. Practically deploying heavyweight adaptive bitrate algorithms with teacher-student learning. IEEE/ACM Transactions on Networking, 29 0 (2): 0 723--736, 2021

    2021

  23. [31]

    Nikishin, J

    E. Nikishin, J. Oh, G. Ostrovski, C. Lyle, R. Pascanu, W. Dabney, and A. Barreto. Deep reinforcement learning with plasticity injection. Advances in Neural Information Processing Systems, 36, 2024

    2024

  24. [32]

    Padakandla

    S. Padakandla. A survey of reinforcement learning algorithms for dynamically varying environments. ACM Comput. Surv., 54 0 (6), July 2021. ISSN 0360-0300. doi:10.1145/3459991. URL https://doi.org/10.1145/3459991

    2021 doi

  25. [33]

    Patel, J

    S. Patel, J. Zhang, N. Narodystka, and S. A. Jyothi. Practically high performant neural adaptive video streaming. Proc. ACM Netw., 2 0 (CoNEXT4), Nov. 2024. doi:10.1145/3696401. URL https://doi.org/10.1145/3696401

    2024 doi

  26. [34]

    Puderbaugh and P

    M. Puderbaugh and P. D. Emmady. Neuroplasticity. In StatPearls [Internet]. StatPearls Publishing, 2023

    2023

  27. [35]

    Sokar, R

    G. Sokar, R. Agarwal, P. S. Castro, and U. Evci. The dormant neuron phenomenon in deep reinforcement learning. In International Conference on Machine Learning, pages 32145--32168. PMLR, PMLR, 2023

    2023

  28. [36]

    Spiteri, R

    K. Spiteri, R. Urgaonkar, and R. K. Sitaraman. Bola: Near-optimal bitrate adaptation for online videos. IEEE/ACM transactions on networking, 28 0 (4): 0 1698--1711, 2020

    2020

  29. [37]

    Y. Sun, X. Yin, J. Jiang, V. Sekar, F. Lin, N. Wang, T. Liu, and B. Sinopoli. Cs2p: Improving video bitrate selection and adaptation with data-driven throughput prediction. In Proceedings of the 2016 ACM SIGCOMM Conference, pages 272--285, 2016

    2016

  30. [38]

    B. Wang, M. Su, W. Wang, K. Chen, B. Liu, F. Ren, M. Xu, J. Liu, and J. Wu. Enhancing low latency adaptive live streaming through precise bandwidth prediction. IEEE/ACM Transactions on Networking, 2024 a

    2024

  31. [39]

    S. Wang, J. Lin, and Y. Dai. Mmvs: Enabling robust adaptive video streaming for wildly fluctuating and heterogeneous networks. IEEE Transactions on Multimedia, 2024 b

    2024

  32. [40]

    Wu, C.-X

    S. Wu, C.-X. Wang, M. M. Alwakeel, X. You, et al. A general 3-d non-stationary 5g wireless channel model. IEEE Transactions on Communications, 66 0 (7): 0 3065--3078, 2017

    2017

  33. [41]

    G. Xu, R. Zheng, Y. Liang, X. Wang, Z. Yuan, T. Ji, Y. Luo, X. Liu, J. Yuan, P. Hua, S. Li, Y. Ze, H. D. III, F. Huang, and H. Xu. Drm: Mastering visual reinforcement learning through dormant ratio minimization. In The Twelfth International Conference on Learning Representatio...

    2024

  34. [42]

    J. Yin, H. Chen, Y. Xu, Z. Ma, and X. Xu. Learning accurate network dynamics for enhanced adaptive video streaming. IEEE Transactions on Broadcasting, 70 0 (3): 0 808--821, 2024

    2024

  35. [43]

    X. Yin, A. Jindal, V. Sekar, and B. Sinopoli. A control-theoretic approach for dynamic adaptive video streaming over http. In Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication, pages 325--338. ACM, 2015

    2015

  36. [44]

    L. Zhao, H. Li, E. Zhang, A. Hawbani, M. Lin, S. Wan, and M. Guizani. Intelligent caching for vehicular dew computing in poor network connectivity environments. ACM Trans. Embed. Comput. Syst., 23 0 (2), Mar. 2024. ISSN 1539-9087. doi:10.1145/3643038. URL https://doi.org/10.11...

    2024 doi

  37. [45]

    Zinner, O

    T. Zinner, O. Hohlfeld, O. Abboud, and T. Ho feld. Impact of frame rate and resolution on objective qoe metrics. In 2010 second international workshop on quality of multimedia experience (QoMEX), pages 29--34, Trondheim, 2010. IEEE

    2010

  38. [46]

    styles/rldmsubmit.sty0000664000000000000000000002162215006272430014026 0ustar rootroot eso-pic 0.95 0.05 = phvb at 8pt 1.5cm 0pt 0pt -2 -1in - - - -2 -1in =10000 =10000 empty #1#2#3 footnote \@makefnmark to 0pt ^ \@thefnmark \@makefntext##1 1em to1.8em @th ^ \@thefnmark ##1 \@...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.