Pith. sign in

REVIEW 4 major objections 5 minor 33 references

Deep Reinforcement Learning for Network Slicing with Heterogeneous Resource Requirements and Time Varying Traffic Dynamics

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A neural policy trained by REINFORCE can split two coupled network resources across slices better than a fixed equal split.

desk verdict Useful formulation of a real slicing problem, but the undefined loss function and an inverted normalization step in the pseudo-code make the empirical claim unverifiable. read the letter →

arxiv 1908.03242 v1 pith:SGLWGOSJ submitted 2019-08-08 cs.NI

Jaehoon Koo , Veena B. Mendiratta , Muntasir Raihan Rahman , Anwar Walid This is my paper · ORCID
classification cs.NI
keywords networkslicingdeepreinforcementlearningpolicygradientREINFORCEresourceallocationMarkovdecisionprocess5Gbandwidthandcompute
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 deep reinforcement learning agent can allocate two coupled resources—bandwidth and virtual machines—across multiple network slices better than a fixed equal split of the budget. The authors formulate network slicing as a Markov decision process with two service modes, service upon arrival and batch service, and solve it with the REINFORCE policy gradient algorithm using separate neural network policies for each resource. They report that in nearly every budget scenario the learned allocations produce lower total loss than an equal-slicing baseline on both synthetic and real workload traces. If this holds, operators could train slicing policies offline on historical traffic and apply them in real time to meet 5G latency and reliability targets.

What carries the argument

The central mechanism is a REINFORCE policy-gradient algorithm with a deep neural network policy per resource, together with a proportional projection step that enforces the finite budget in every time step. Each network takes a feature vector consisting of received request amounts, buffer level, and last arrival or service time, and outputs positive real-valued allocations for the three slices; the leaky ReLU activation keeps the allocations positive. The REINFORCE update adjusts the policy parameters along an estimated gradient of the expected discounted loss, and the projection step rescales the raw network outputs whenever the summed allocation would exceed the bandwidth or compute budget.

What would settle it

Train the proposed agents on the first 90 percent of the combined SWIM and LTE traces, then inflate the arrival rate of one slice by a factor of two in the remaining 10 percent and compare total loss to the equal-slicing baseline; if the learned policy's loss rises to match or exceed equal slicing, the stationarity premise would be violated.

Watch

Extended reading notes

Core claim

The discovery is that a model-free policy-gradient learner can discover how to partition two finite resource pools among slices with heterogeneous request distributions, without being told the arrival process or the resource-requirement distributions in advance. In the proposed MDP, the state for each resource records received request amounts, buffer levels, and the time since the last arrival or service; the action is a vector of per-slice allocations; and the reward is the negative of a weighted sum of processing delays and resource-use costs. When a proposed allocation would exceed the bandwidth or compute budget, the output is rescaled proportionally. On the evaluated scenarios, the trained neural policies obtain lower total loss—equivalently higher reward—than equal slicing in almost every budget tier, and the real-data training curves show buffer levels falling as the policies learn.

Load-bearing premise

The load-bearing premise is that traffic dynamics are stationary, so a policy trained offline on historical traces remains close to optimal when deployed on new, time-varying traffic; the paper states this Markovian-stationarity assumption explicitly and has no built-in mechanism to adapt if the distribution shifts.

Editorial extensions

If this is right

  • At decision time, each resource allocation is a single forward pass through a trained network, so the approach avoids solving an optimization problem on every arrival or batch service.
  • The formulation covers two service models, immediate allocation upon arrival and periodic allocation of all queued requests, so the same training procedure applies to both operating modes.
  • Because the reward is a weighted sum of processing delay and resource-use cost, an operator can change the balance factor to emphasize latency over utilization without changing the network architecture.
  • The proportional projection guarantees that allocations never exceed the per-step bandwidth or compute budget, making the learned policy directly usable in capacity-constrained settings.

Reading between the lines

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

  • A stricter deployment test would train on one traffic regime and evaluate after a regime shift, such as doubling one slice's arrival rate; the paper's own stationarity assumption implies the policy would then need retraining, and the paper does not report that experiment.
  • The same proportional-projection budget constraint could be extended to a third resource type such as storage; the main question the paper leaves open is whether the policy-gradient updates remain stable as the allocation vector grows.
  • Comparing against a water-filling or priority baseline rather than only equal slicing would isolate how much of the gain comes from learning request distributions versus simply reacting to rising buffer levels; the paper's results do not separate those effects.
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 proposes a policy-gradient deep reinforcement learning approach to network slicing, in which a controller allocates two heterogeneous resources (bandwidth and virtual machines) to multiple slices under finite resource budgets. Two service modes are considered: service upon arrival and batch service. The authors formulate the allocation problem as an MDP, solve it with the REINFORCE algorithm using separate neural-network policy agents for each resource, and compare the learned policies against an equal-slicing baseline on synthetic Poisson/uniform data and on combined Facebook compute traces and 4G LTE bandwidth traces. The central empirical claim is that the learned policies achieve lower total loss than equal slicing in most budget scenarios, as reported in Tables III and IV.

Significance. Network slicing with joint allocation of heterogeneous resources under dynamic traffic is an important problem, and a demonstrated learning-based policy would be practically valuable. The paper has strengths: it formulates both service-upon-arrival and batch-service problems, uses trace-driven evaluation with real workloads, and compares against an external equal-slicing baseline, so there is no circularity in the evaluation protocol. However, the significance cannot be assessed in the current manuscript because the objective function that defines both training and evaluation is left unspecified, the pseudo-code contains an apparent inversion of the budget-projection condition, and the abstract's claims about utilization, latency, and demands satisfied are not directly reported. These issues make the reported results non-reproducible and the central claim unverifiable as written.

major comments (4)
  1. [Section IV-A, Eq. (4), Tables III-IV] The optimization objective L = L_QoS(β, δ) + w L_Res(b, v) is never fully defined. The text states only that L_QoS 'integrates delays in processing bandwidth and VM requests measuring buffer levels' and that L_Res is 'a cost for bandwidth and computing resources'; no equations, functional forms, units, or scaling constants are provided. Because this same L is used as the training objective J in Fig. 3 and as the evaluation metric in Tables III and IV, all reported improvements over equal slicing are unverifiable. An independent implementation cannot reproduce the results, and a reader cannot determine whether a different specification of L_QoS or L_Res would preserve the reported ranking. Please give explicit formulas for L_QoS and L_Res, including how the buffer levels β and δ enter and how the resource costs are computed.
  2. [Fig. 3, Section IV-B] The pseudo-code in Fig. 3 normalizes the allocations when the sum is less than or equal to the budget: lines 4-6 apply the projection when sum b_i <= B_t, and lines 8-10 do the same for v_i. The text in Section IV-B states the opposite, namely that allocations are projected when the sum of allocations exceeds the budgets. This inversion is not a cosmetic issue: if the experiments follow the code, the budget constraints may not have been enforced as described. Please correct the condition (presumably to 'if sum > B_t') and clarify whether the reported results were produced by the corrected or the inverted version.
  3. [Section III-A] The MDP formulation states: 'We make the markovian assumption that the future state (s_{t+1}) only depend on the current state (s_t), and also that the dynamics are stationary and do not change over time.' This stationarity assumption directly conflicts with the paper's stated motivation of 'time varying traffic dynamics' and with the abstract's claim that the method adapts to dynamic traffic. Because the policies are trained offline on historical traces, it is not clear that they would remain effective if the underlying arrival processes or resource-request distributions change after deployment. Please either relax the stationarity assumption, provide evidence that the considered traces are stationary, or evaluate the trained policies under explicitly non-stationary traffic.
  4. [Tables III and IV] The abstract claims improvements in 'resource utilization, latency performance, and demands satisfied,' but Tables III and IV report only a single scalar loss value, whose components are undefined (see first major comment). No direct measurements of utilization, latency, or satisfied demand are presented. In addition, the tables give only point estimates with no confidence intervals, standard deviations, or number of seeds; REINFORCE is a high-variance estimator, and some total differences are small (e.g., Table IV, batch service, large budget, VM total: NN 6.612E+03 vs ES 6.642E+03). Please report the individual loss components that map to the abstract's claims and include statistical measures across episodes and training runs.
minor comments (5)
  1. [Section III-C] The text cites 'REINFORCE [10]' but reference [10] is the AlphaGo paper; the correct citation for REINFORCE is [14] (Williams, 1992).
  2. [Section V, Section VI-C] The merging of the Facebook CPU trace and the 4G LTE bandwidth trace is described as taking the maximum bandwidth value over each inter-arrival interval, but no justification is given for this choice, and the trace values are said to have been 'appropriately scaled down' without specifying the scaling factors. Please document these preprocessing decisions for reproducibility.
  3. [Section VI-B] The hyperparameter selection is described only as 'based on the results,' with learning rates, layer counts, and unit counts enumerated; the validation results that motivated the chosen settings are not shown. Please include them or state how the choices were made.
  4. [Table I and Section IV-A] The reward descriptions in Table I are qualitative ('-Delays in processing requests and resource use costs'); please connect them explicitly to the formal loss L_QoS and L_Res introduced in Eq. (4).
  5. [General] There are several typos and stylistic issues, including 'a arrival process' (Section IV-A), 'startegy' (Section VI-C), and 'Were we also use the Adam optimizer' (Section VI-C).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the neural-network policies are trained and evaluated on the same reward function, which is standard RL, and the equal-slicing baseline is an external comparison; the undefined loss terms are a reproducibility concern, not a circular reduction.

full rationale

The paper's derivation chain is self-contained in the relevant sense. The objective in Eq. (4) is defined as a loss L = L_QoS + w L_Res and the REINFORCE agents are trained to minimize the expected discounted version of that same loss; using the same scalar as training objective and evaluation metric is the normal RL protocol, not a fitted input masquerading as a prediction. The comparison against equal slicing is an external baseline policy, and the reported tables give losses for each class and resource, so the NN advantage is not obtained by construction from the training objective alone. There are no load-bearing self-citations: the cited REINFORCE algorithm, Adam optimizer, and related-work papers are all external. No uniqueness theorem or ansatz is imported from the authors' prior work. The undefined forms of L_QoS and L_Res, the stationarity assumption versus time-varying traffic, and the apparent inversion of the budget-normalization inequality in Fig. 3 are validity/reproducibility concerns, but none of them exhibits an equation or fitted parameter that reduces the claimed prediction to its own input. Accordingly, the appropriate circularity score is 0.

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

The central claim rests on a standard RL MDP with several domain assumptions: stationarity, infinite buffers, shared arrival processes, and an ad hoc trace merging rule. There are no newly invented physical or algorithmic entities, but the hand-chosen reward weight, unstated discount factor, and unreported trace scaling are free parameters that the results depend on.

free parameters (4)
  • Balance factor w = w=1
    The loss is L = L_QoS + w L_Res, and the paper states 'We put the same weights on delays and resource use costs, w=1'. The central results depend on this hand-chosen weighting.
  • Discount factor gamma = not stated
    Gamma appears in Eq. (4) and in the cumulative reward, but its value is never reported. All reported loss values depend on it.
  • Real trace scaling factors = not reported
    Section VI-C states the traces were 'appropriately scaled down', but the exact factors are omitted, so absolute loss magnitudes and the comparison cannot be reproduced.
  • Leaky ReLU slope a = not reported
    The activation is defined as h(x)=x if x>0 and h(x)=ax otherwise, with 'a small constant a', but the value of a is never given.
assumptions (4)
  • domain assumption The state transition is Markovian and the dynamics are stationary.
    Section III-A states this explicitly. It underlies the MDP formulation and REINFORCE, but conflicts with the paper's motivation of time varying traffic dynamics.
  • domain assumption Infinite buffers for both bandwidth and VM requests.
    Section IV-A states 'We assume that we have infinite buffers for both resources'. Real systems have finite buffers, and this assumption affects the delay and loss calculations.
  • domain assumption For each class, bandwidth and VM requests share the same arrival process.
    Section IV-A states 'requests for bandwidth and VMs have the same arrival process for each class'. This couples the two resources and shapes the merged workload.
  • domain assumption The merged real workload uses the maximum bandwidth value in each inter-arrival interval.
    Section V states this rule for combining the separate CPU and bandwidth traces. The resulting bandwidth request sequence is an artificial construction that directly determines the experimental inputs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Reinforcement Learning for Network Slicing with Heterogeneous Resource Requirements and Time Varying Traffic Dynamics." pith.science (2026). https://pith.science/paper/SGLWGOSJ

@misc{pith2026190803242,
  author       = {Pith},
  title        = {Pith review of: Deep Reinforcement Learning for Network Slicing with Heterogeneous Resource Requirements and Time Varying Traffic Dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SGLWGOSJ}},
  note         = {Machine review of arXiv:1908.03242}
}
read the original abstract

Efficient network slicing is vital to deal with the highly variable and dynamic characteristics of network traffic generated by a varied range of applications. The problem is made more challenging with the advent of new technologies such as 5G and new architectures such as SDN and NFV. Network slicing addresses a challenging dynamic network resource allocation problem where a single network infrastructure is divided into (virtual) multiple slices to meet the demands of different users with varying requirements, the main challenges being --- the traffic arrival characteristics and the job resource requirements (e.g., compute, memory and bandwidth resources) for each slice can be highly dynamic. Traditional model-based optimization or queueing theoretic modeling becomes intractable with the high reliability, and stringent bandwidth and latency requirements imposed by 5G technologies. In addition these approaches lack adaptivity in dynamic environments. We propose a deep reinforcement learning approach to address this dynamic coupled resource allocation problem. Model evaluation using both synthetic simulation data and real workload driven traces demonstrates that our deep reinforcement learning solution improves overall resource utilization, latency performance, and demands satisfied as compared to a baseline equal-slicing strategy.

Figures

Figures reproduced from arXiv: 1908.03242 by the authors.

Figure 1
Figure 1. Example of 5G network slicing III. BACKGROUND ON RL AND DEEP RL In this section, we give a brief introduction of RL, specifi￾cally the use of neural networks for RL function approxima￾tion, and the policy gradient learning algorithm [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Deep Reinforcement Learning [13] B. Policy The agent chooses actions based on a learned policy, which is a probability distribution over actions for any state, π : π(s, a) → [0, 1], where π(s, a) denotes the probability of taking action a in state s, and P a π(s, a) = 1, ∀s. In most practical scenarios, especially network environments, there are exponentially many possible state,action (s,a) pairs (see section IV). … view at source ↗
Figure 3
Figure 3. REINFORCE [14] Algorithm for Network Slicing [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Service upon arrival are applied to this formulation as for the service upon arrival formulation [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Bandwidth Traces [17] V. DATA FOR MODELS In this section we briefly discuss the workload traces for CPU and bandwidth requests that are used for the evaluation of the algorithms (Section VI). Facebook Workload Trace. SWIM [16] is a workload generator for map-reduce clu…
Figure 7
Figure 7. Figure 7: Real data: Mean resource allocations, and buffer levels of three classes (large budget, service upon arrival, VM) [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 29 canonical work pages

  1. [31]

    Deep reinforcement learning for network slicing,

    Z. Zhao, R. Li, Q. Sun, Chi-Lin, C. Yang, X. Chen, M. Zhao, and H. Zhang, “Deep reinforcement learning for network slicing,” ArXiv, 2018

    work page 2018

  2. [1]

    Network function virtualization: State-of-the-art and re- search challenges,

    R. Mijumbi, J. Serrat, J.-L. Gorricho, N. Bouten, F. De Turck, and R. Boutaba, “Network function virtualization: State-of-the-art and re- search challenges,” IEEE Communications Surveys & Tutorials , vol. 18, no. 1, pp. 236–262, 2016

    work page 2016

  3. [2]

    Software-defined networking,

    N. McKeown, “Software-defined networking,” INFOCOM keynote talk , vol. 17, no. 2, pp. 30–32, 2009

    2009

  4. [3]

    Massive mimo for 5g,

    E. G. Larsson and L. Van der Perre, “Massive mimo for 5g,” IEEE 5G Tech F ocus, vol. 1, no. 1, pp. 1–4, 2017

    work page 2017

  5. [4]

    Industry 4.0 — Wikipedia, the free encyclope- dia,

    Wikipedia contributors, “Industry 4.0 — Wikipedia, the free encyclope- dia,” 2019. [Online; accessed 22-April-2019]

    work page 2019

  6. [5]

    An introduction to network slicing,

    “An introduction to network slicing,” GSM Association , 2017

    work page 2017

  7. [6]

    M. J. Neely, Stochastic Network Optimization with Application to Com- munication and Queueing Systems . Morgan and Claypool Publishers, 2010

    work page 2010

  8. [7]

    Network optimization and control,

    S. Shakkottai and R. Srikant, “Network optimization and control,” F ound. Trends Netw., vol. 2, pp. 271–379, Jan. 2007

    work page 2007

Show all 33 references
  1. [8]

    Srikant and L

    R. Srikant and L. Ying, Communication Networks: An Optimization, Control and Stochastic Networks Perspective . New York, NY , USA: Cambridge University Press, 2014

    2014

  2. [9]

    R. S. Sutton and A. G. Barto, Reinforcement learning: An introduction . MIT press, 2018

    2018

  3. [10]

    Mastering the game of go without human knowledge,

    D. Silver, J. Schrittwieser, K. Simonyan, I. Antonoglou, A. Huang, A. Guez, T. Hubert, L. Baker, M. Lai, A. Bolton, Y . Chen, T. Lillicrap, F. Hui, L. Sifre, G. van den Driessche, T. Graepel, and D. Hassabis, “Mastering the game of go without human knowledge,”Nature, vol. 550,...

    2017

  4. [11]

    Network slicing for 5g: Challenges and opportunities,

    X. Li, M. Samaka, H. A. Chan, D. Bhamare, L. Gupta, C. Guo, and R. Jain, “Network slicing for 5g: Challenges and opportunities,” IEEE Internet Computing , 2017

    2017

  5. [12]

    Network slicing to enable scalability and flexibility in 5g mobile networks,

    P. Rost, C. Mannweiler, D. S. Michalopoulos, C. Sartori, V . Sciancale- pore, N. Sastry, O. Holland, S. Tayade, B. Han, D. Bega, D. Aziz, and H. Bakker, “Network slicing to enable scalability and flexibility in 5g mobile networks,” IEEE Communications Magazine , vol. 55, pp. 72...

    2017

  6. [13]

    Resource man- agement with deep reinforcement learning,

    H. Mao, M. Alizadeh, I. Menachey, and S. Kandulay, “Resource man- agement with deep reinforcement learning,” Association for Computing Machinery (ACM), 2016

    2016

  7. [14]

    Simple statistical gradient-following algorithms for connectionist reinforcement learning,

    R. J. Williams, “Simple statistical gradient-following algorithms for connectionist reinforcement learning,” Mach. Learn. , vol. 8, pp. 229– 256, May 1992

    1992

  8. [15]

    Rectifier nonlinearities improve neural network acoustic models,

    A. L. Maas, A. Y . Hannun, and A. Y . Ng, “Rectifier nonlinearities improve neural network acoustic models,” International Conference on Machine Learning , 2013

    2013

  9. [16]

    Statistical workload injector for mapreduce (swim)

    “Statistical workload injector for mapreduce (swim).” https://github.com/ SWIMProjectUCB/SWIM/wiki. Accessed: 2018-01-14

    2018

  10. [17]

    Http/2-based adaptive streaming of hevc video over 4g/lte networks,

    J. van der Hooft, S. Petrangeli, T. Wauters, R. Huysegems, P. R. Alface, T. Bostoen, and F. D. Turck, “Http/2-based adaptive streaming of hevc video over 4g/lte networks,” IEEE Communications Letters , vol. 20, no. 11, pp. 2177–2180, 2016

    2016

  11. [18]

    Adam: A method for stochastic optimization,

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

    2014

  12. [19]

    A new approach for allocating buffers and bandwidth to heterogeneous, regulated traffic in an atm node,

    A. Elwalid, D. Mitra, and R. H. Wentworth, “A new approach for allocating buffers and bandwidth to heterogeneous, regulated traffic in an atm node,” IEEE Journal on selected Areas in Communications , 1995

    1995

  13. [20]

    Adaptable bandwidth planning using reinforcement learn- ing,

    D. Hetzer, “Adaptable bandwidth planning using reinforcement learn- ing,” Journal of Systemics, Cybernetics and Informatics , 2006

    2006

  14. [21]

    Adaptive call admission control under quality of service constraints: a reinforcement learning solution,

    H. Tong and T. X. Brown, “Adaptive call admission control under quality of service constraints: a reinforcement learning solution,” IEEE Journal on Selected Areas in Communications , 2000

    2000

  15. [22]

    Adaptive provisioning of differentiated services networks based on reinforcement learning,

    T. C.-K. Hui and C.-K. Tham, “Adaptive provisioning of differentiated services networks based on reinforcement learning,” IEEE Transactions on Systems, Man, and Cybernetics-Part C: Applications and Reviews , 2003

    2003

  16. [23]

    A reinforcement learning scheme for adaptive link allocation in atm networks,

    J. C. Ernst Nordstr ¨om, “A reinforcement learning scheme for adaptive link allocation in atm networks,” Rural Growth Linkages in the Eastern Cape Province of South Africa , 1995

    1995

  17. [24]

    Self- learning cloud controllers: Fuzzy q-learning for knowledge evolution,

    P. Jamshidi, A. Shariflooz, C. Pahly, A. Metzgerz, and G. Estrada, “Self- learning cloud controllers: Fuzzy q-learning for knowledge evolution,” International Conference on Cloud and Autonomic Computing , 2015

    2015

  18. [25]

    Rlpas: Reinforcement learning-based proactive auto-scaler for resource provisioning in cloud environment,

    J. V . B. Benifa and D. Dejey, “Rlpas: Reinforcement learning-based proactive auto-scaler for resource provisioning in cloud environment,” Mobile Networks and Applications , 2018

    2018

  19. [26]

    Optimising 5g infrastructure markets: The business of network slicing,

    D. Bega, M. Gramaglia, A. Banchs, V . Sciancaleporey, K. Samdanisz, and X. Costa-Perez, “Optimising 5g infrastructure markets: The business of network slicing,” IEEE Conference on Computer Communications , 2017

    2017

  20. [27]

    Slice as an evolutionary service: Genetic optimization for inter-slice resource management in 5g networks,

    B. Han, J. Lianghai, and H. D. Schotten, “Slice as an evolutionary service: Genetic optimization for inter-slice resource management in 5g networks,” IEEE Access , 2018

    2018

  21. [28]

    Software-defined networks with mobile edge computing and caching for smart cities: A big data deep reinforcement learning approach,

    Y . He, F. R. Yu, N. Zhao, V . C. Leung, and H. Yin, “Software-defined networks with mobile edge computing and caching for smart cities: A big data deep reinforcement learning approach,” IEEE Communications Magazine, 2017

    2017

  22. [29]

    A deep reinforcement learning based framework for power-efficient resource allocation in cloud rans,

    Z. Xu, Y . Wang, J. Tang, J. Wang, and M. C. Gursoy, “A deep reinforcement learning based framework for power-efficient resource allocation in cloud rans,” IEEE ICC 2017 Next Generation Networking and Internet Symposium , 2017

    2017

  23. [30]

    Deep rein- forcement learning (drl)-based resource management in software-defined and virtualized vehicular ad hoc networks,

    Y . He, F. R. Yu, N. Zhao, H. Yin, and A. Boukerche, “Deep rein- forcement learning (drl)-based resource management in software-defined and virtualized vehicular ad hoc networks,” Association for Computing Machinery (ACM), 2017

    2017

  24. [32]

    Konda, Actor-critic Algorithms

    V . Konda, Actor-critic Algorithms . PhD thesis, Massachusetts Institute of Technology, Cambridge, MA, USA, 2002. AAI0804543

    2002

  25. [33]

    Playing atari with deep reinforcement learn- ing,

    V . Mnih, K. Kavukcuoglu, D. Silver, A. Graves, I. Antonoglou, D. Wier- stra, and M. Riedmiller, “Playing atari with deep reinforcement learn- ing,” in NIPS Deep Learning Workshop , 2013

    2013

Pith tools

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