Pith. sign in

REVIEW 5 major objections 6 minor 38 references

A separate Belief head can replace deep search in Go by filtering overconfident policy errors at inference time.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-30 16:11 UTC pith:CKGPND5E

load-bearing objection Clear architecture idea for search-light Go, but the professional-strength and win-rate claims are asserted without any game-outcome evidence. the 5 major comments →

arxiv 2607.26946 v1 pith:CKGPND5E submitted 2026-07-29 cs.AI

Belief-Guided Decision Making with Uncertainty Gating in the Game of Go

Mehrad Yaghoubi , Azam Bastanfard , Abbas Jalilvand , Ashkan Rezaei This is my paper
classification cs.AI
keywords Computer GoBelief headuncertainty gatingsearch-free playMonte Carlo Tree SearchVision Transformerhallucinationepistemic uncertainty
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Standard Go engines lean on Monte Carlo Tree Search to fix the neural network’s mistakes. On ordinary GPUs that search is too slow, and without it the network often plays high-confidence blunders. This paper argues that the fix is architectural, not more search: keep a Policy head for choosing moves, but add a distinct Belief head that scores how stable and trustworthy the current position is. Memory layers track long-term threats and Ko, and a gate suppresses policy moves the Belief head flags as risky. Intelligence moves from runtime tree expansion into learned parametric intuition, so search-free play stays strong on limited hardware.

Core claim

Disentangling Policy from a supervised Belief head that models epistemic uncertainty and strategic stability, then gating the policy with that belief signal, internalizes the verification role usually left to MCTS and yields stronger search-free Go play with fewer hallucinations on consumer hardware.

What carries the argument

Belief-Guided Decision Model (BGDM): a Vision-Transformer backbone with a detached Belief head (global pool plus MLP critic) whose uncertainty signal gates Policy logits, trained first on expert/KataGo data then used as an independent filter at inference.

Load-bearing premise

A supervised Belief head plus gating can stand in for deep MCTS verification well enough to reach professional-level strength without search.

What would settle it

Play matched search-free games on the same backbone with and without Belief gating against a fixed opponent, and check whether win rate, blunder rate, and Elo actually rise as claimed; if gating does not improve game outcomes over the raw policy, the central claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Search-free Go bots can keep higher move quality on single consumer GPUs by rejecting overconfident policy errors before they are played.
  • Gradient detachment lets the Belief head act as a fixed critic, so policy training cannot rewrite the network’s sense of positional risk.
  • Memory over history (Transformer/GRU) becomes part of the belief state, helping detect Ko and slow multi-move attacks without tree expansion.
  • The same belief-plus-gate pattern is proposed as transferable to other domains with structural uncertainty, such as real-time strategy games.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If belief calibration truly tracks fatal traps, the same detached critic could regularize any overconfident discrete policy, not only Go.
  • The missing head-to-head Elo tables against search-starved AlphaZero-style baselines are the natural next measurement the architecture invites.
  • Gating on belief may trade some tactical sharpness for blunder avoidance, a quality-vs-creativity tension worth measuring in endgame suites.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes a Belief-Guided Decision Model (BGDM) for Go that disentangles a Policy head from a separate Belief head intended to model epistemic uncertainty and strategic stability. A Transformer backbone with memory, gradient detachment between Actor and Belief, and an uncertainty gating mechanism are introduced so that Belief can act as an internal critic/filter in place of deep MCTS. The stated goal is higher-quality search-free play on consumer GPUs (e.g., RTX 2060), with reduced hallucination and professional-level strength. Experiments in §IV report supervised pretraining dynamics, Belief-error vs Acc@1/Acc@5 correlation, and qualitative policy/belief heatmaps; the Discussion asserts end-game robustness and a 30% blunder reduction.

Significance. If the central claim held—that a calibrated, detached Belief head plus gating can substitute for MCTS verification well enough to raise search-free win rates to professional level on limited hardware—the result would matter for deployable game AI and, more broadly, for uncertainty-aware policy filtering without runtime search. The architectural separation of Policy from an independent critic and the explicit gradient-detachment protocol (III.C, Eq. 1) are reasonable design ideas. As written, however, the manuscript does not supply the game-outcome, Elo, baseline, ablation, or latency evidence needed to establish that significance; the contribution remains a proposal supported only by supervised move-matching metrics.

major comments (5)
  1. [Abstract; §I; §IV–VI] Abstract, §I, and §VI claim that BGDM “significantly improves search-free win rates,” “reduces hallucination,” and enables “professional-level play on limited hardware.” §IV–V report only supervised pretraining loss/accuracy curves (Fig. 5), Belief-error vs Acc@k correlation on expert/KataGo-style labels (Fig. 6), and qualitative heatmaps (Fig. 7). There are no win rates, Elo ratings, head-to-head matches against search-free Policy-Value or KataGo baselines, self-play strength curves, or any search-free playing evaluation. The load-bearing causal link—Belief calibration and gating → search-free strength—is therefore untested.
  2. [§V Discussion] §V states the agent reduces blunders by 30% due to more accurate perception of passing conditions. No definition of “blunder,” no counting protocol, no baseline comparison, and no table or figure support this number. An unsupported quantitative claim in the Discussion cannot stand in for missing outcome experiments.
  3. [§III.A; Abstract] The gating mechanism is central to the method (Abstract; §III.A “Control Relation”: Belief filters high-risk Policy moves). The manuscript never specifies the gate (threshold, risk rule, or how belief logits modify action selection at inference), nor does it report an ablation with vs. without gating under play or even under supervised top-k. Without that, the claim that gating filters overconfident policy errors is not evidenced.
  4. [§III.B–C; Algorithm 1; §IV] Algorithm 1 and §III.B describe only supervised pretraining of a belief_policy distribution via cross-entropy on expert actions (5 epochs, embed 128, 4 layers/heads). Gradient detachment (Eq. 1, §III.C) and the multi-head Policy/Value/Belief setup are not evaluated in RL/self-play, and no search-free inference protocol on RTX-class hardware (latency, nodes/sec, or strength at fixed budget) appears. The experimental section therefore does not instantiate the architecture’s intended use case as argued in §I against MCTS cost on consumer GPUs.
  5. [§III.A; Fig. 1; Abstract] Fig. 1 and the text disagree on heads and memory: the figure caption lists Policy, Value, and Territory Belief; the abstract promises Transformer/GRU for Ko and long-term dependency; the body emphasizes a Vision Transformer and an MLP Belief head after mean pooling, with Value largely undeveloped in the methods. This inconsistency makes the proposed model hard to reproduce and weakens the claim that Belief is a well-specified “internal simulator” distinct from standard value/ownership heads (cf. KataGo auxiliary targets discussed in §II.C).
minor comments (6)
  1. [Algorithm 1] Algorithm 1 is poorly typeset (line breaks mid-sentence, “belief_policy” vs. separate Belief head, incomplete evaluation block). It should be cleaned and aligned with the architecture description.
  2. [§III.C Eq. (1)] Eq. (1) writes ∇θ LTotal with LActor, LCritic, LBelief but does not define the individual losses or the stop-gradient operator formally; a clear expression for the detached update would help.
  3. [§II; References] References [37] and [38] (tennis video annotation; Ma’qeli script cellular automata) are unrelated to Computer Go or belief/uncertainty modeling and should be removed or justified.
  4. [§III.A; passim] Typos and wording: “Fixed2DPositionalEncoding,” “nosep,” “gati ng,” “Efficiency,” “halluci- nations” line breaks; “Vision Transformer backbone” vs. earlier CNN-centric Go literature could use a short related-work note on ViT-for-Go ([33] is cited but lightly used).
  5. [Fig. 6; §IV.B] Fig. 6 reports correlation between Belief error and Acc@k but no sample size, confidence intervals, or dataset split; axes and “Mean Belief Error” units are unspecified.
  6. [§III.A.1; Abstract] Input is described as 4 channels including “History,” but Ko handling and history length are not specified despite the abstract’s emphasis on Ko and memory.

Circularity Check

2 steps flagged

Mild proxy circularity: supervised expert-move accuracy and Belief-error correlation are treated as if they were search-free win-rate / professional-strength results

specific steps
  1. fitted input called prediction [Abstract; §IV.B Hypothesis Validation; §V Discussion]
    "Experimental results demonstrate that this approach significantly improves search-free win rates and reduces hallucination, enabling professional-level play on limited hardware where massive MCTS is infeasible. ... As illustrated in Fig. 6, a lower belief error consistently leads to higher Acc@1 and Acc@5 scores. This indicates that the belief head provides a “baseline of reality,” preventing the policy from proposing high-confidence but strategically fatal moves. ... reducing blunders by 30%"

    Win rates, hallucination under play, blunders, and professional search-free strength are never measured (no Elo, match results, gating ablations, or hardware throughput). The only quantitative “validation” is supervised Belief-error vs. expert-move Acc@k on the same grounding distribution used to train the heads. Presenting that in-distribution proxy correlation—and an unsupported 30% figure—as experimental demonstration of search-free win-rate gains treats a fitted/related accuracy statistic as if it were the claimed outcome.

  2. fitted input called prediction [§III.B Algorithm 1; §IV.A–B; Fig. 6]
    "P (k) → Model(s) {Get belief_policy distribution} ... L → CrossEntropy(P (k), atarget) ... The reduction in Belief Error suggests that the shared Transformer backbone successfully learns representations that satisfy both move prediction and environmental stability assessment. ... a lower belief error consistently leads to higher Acc@1 and Acc@5 scores."

    Pretraining optimizes a belief_policy-style CE against expert actions on the shared SpatialTransformer; Policy accuracy is the same family of target. Under joint training, runs with lower Belief error will tend to have higher Acc@k by construction of shared-feature fit quality. The paper reads this correlation as causal evidence that Belief supplies an independent “baseline of reality” that prevents fatal Policy moves—i.e., a training-fit relationship is renamed as confirmation of the Belief-as-MCTS-substitute hypothesis.

full rationale

This is an architecture/ML-systems paper, not a first-principles derivation, so classic algebraic self-definition is largely absent. Gradient detachment (Eq. 1), the Belief/Policy split, and the gating story are design choices, not quantities defined in terms of the claimed outcome. Self-citations [36–38] (Polo RL, cellular automata, tennis annotation) are peripheral and not load-bearing for the Go strength claim. The only circular pressure is moderate and proxy-based: the Abstract/I/VI claim that experiments show improved search-free win rates, reduced hallucination, and professional-level play on limited hardware, while Section IV only reports supervised pretraining loss, Acc@1/Acc@5, Belief-error decay, and their correlation on expert/KataGo-style labels—plus an unmeasured “30% blunder reduction” in V. That correlation is partly expected by construction under a shared backbone trained on the same expert-move distribution (Algorithm 1 even names a belief_policy CE objective), so using it as validation that Belief has internalized MCTS-like foresight and will win search-free games is a fitted-proxy-as-prediction step. It does not make the architecture identical to the win-rate claim; score 3 reflects partial, non-central circularity rather than a forced derivation.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 2 invented entities

The central performance claim rests on modeling choices and untested transfer from supervised proxies to search-free strength, not on a closed derivation. Free parameters are ordinary network/optimiser choices; axioms are standard RL/Go assumptions plus paper-specific claims that Belief replaces MCTS and that expert-label correlation implies win-rate gains. Invented entities are architectural modules without external falsifiable handles beyond the missing game evaluations.

free parameters (4)
  • Transformer width/depth/heads (embed 128, 4 layers, 4 heads) = 128-dim, 4 layers, 4 heads
    Architecture capacity chosen in Algorithm 1; no ablation showing necessity for the claimed search-free strength.
  • Pretraining hyperparameters (lr, batch, epochs) = lr=1e-4, B=64, E=5
    Adam η=1e-4, batch 64, 5 epochs set in Algorithm 1; govern the supervised Belief grounding the method depends on.
  • Belief gating threshold / risk rule
    Inference filter that suppresses Policy when Belief indicates high risk is core to the method (III.A.5) but never given a numeric rule or learned schedule.
  • Multi-task loss weighting (Policy vs Belief vs Value)
    Joint training is asserted; balancing weights are not specified despite citing gradient interference concerns ([13]).
axioms (5)
  • domain assumption Standard Go rules and the need to handle Ko/history make pure Markov features insufficient, motivating memory.
    Invoked in I and III to justify Transformer/GRU history channels.
  • domain assumption MCTS is the main corrector of raw policy error in AlphaZero-style agents and is too costly on consumer GPUs.
    Framing assumption in Abstract and II.A; standard in the field though hardware numbers are not measured here.
  • ad hoc to paper A supervised Belief head can act as an internal simulator/independent critic that replaces MCTS verification via gating.
    Core methodological axiom in I and III.A–C; not derived and not validated by game outcomes.
  • ad hoc to paper Gradient detachment makes Belief an objective critic and prevents confirmation bias between Policy and Belief.
    Stated in III.C (Eq. 1 discussion); plausible design choice, not proven necessary or sufficient.
  • ad hoc to paper Lower Belief error on expert data implies higher-quality tactical decisions and fewer search-free blunders.
    Hypothesis Validation IV.B and Discussion V treat correlation with Acc@k as support for win-rate/blunder claims.
invented entities (2)
  • Belief head (strategic stability / epistemic uncertainty critic) no independent evidence
    purpose: Provide an internal filter distinct from Policy and scalar Value to suppress hallucinated moves without MCTS.
    Named novelty in Abstract and III.A.4; outputs and training target are underspecified relative to standard value/ownership heads.
  • Belief-Guided Decision Model (BGDM) with uncertainty gating no independent evidence
    purpose: End-to-end architecture that shifts verification from runtime search into parametric intuition.
    System label in III; gating mechanism is described qualitatively only.

pith-pipeline@v1.2.0-daily-grok45 · 12393 in / 3832 out tokens · 81304 ms · 2026-07-30T16:11:17.718614+00:00 · methodology

0 comments
read the original abstract

Recent advancements in Computer Go, driven by AlphaZero and MuZero, rely heavily on Monte Carlo Tree Search (MCTS) to correct the errors of the neural network policy. While effective on massive computational clusters, this dependence creates a critical bottleneck on consumer-grade hardware, where the computational cost of tree management severely limits inference rates. Furthermore, without deep search, these models suffer from hallucination, proposing moves with high confidence that are strategically fatal. This paper introduces a novel Belief-Guided architecture that disentangles the Policy head from a distinct Belief head. Unlike traditional value functions, the Belief head acts as an internal simulator and independent critic, modeling epistemic uncertainty and strategic stability. By integrating memory mechanisms (Transformer/GRU) to handle long-term dependencies and the Ko rule, and utilizing a gating mechanism to filter overconfident policy errors, our model shifts the burden of intelligence from runtime search to parametric "intuition." Experimental results demonstrate that this approach significantly improves search-free win rates and reduces hallucination, enabling professional-level play on limited hardware where massive MCTS is infeasible.

Figures

Figures reproduced from arXiv: 2607.26946 by Abbas Jalilvand, Ashkan Rezaei, Azam Bastanfard, Mehrad Yaghoubi.

Figure 7
Figure 7. Figure 7: Visualizing the Decision Process. (Left) Board State. (Center) Policy Heatmap. (Right) Belief Value and Top [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 10 linked inside Pith

  1. [1]

    Silver, D., Huang, A., Maddison, C. et al. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529, 484–489

    2016

  2. [2]

    Silver, D., Schrittwieser, J., Simonyan, K. et al. (2017). Mas -tering the game of Go without human knowledge. Nature, 550, 354–359

    2017

  3. [3]

    Silver, D., et al. (2017). Mastering chess and shogi by self-play with a general reinforcement learning algorithm. arXiv preprint arXiv:1712.01815

    Pith/arXiv arXiv 2017

  4. [4]

    Schrittwieser, J., Antonoglou, I., Hubert, T. et al. (2020). Mastering Atari, Go, chess and shogi by planning with a learned model. Nature, 588, 604 – 609

    2020

  5. [5]

    Danihelka, I., Guez, A., Schrittwieser, J., & Silver, D. (2022). Policy improvement by planning with Gumbel. International Conference on Learning Representations (ICLR)

    2022

  6. [6]

    Kocsis, L., & Szepesvári, C. (2006). Bandit Based Monte-Carlo Planning. Machine Learning: ECML 2006, 4212, 282 -293

    2006

  7. [7]

    B., et al

    Browne, C. B., et al. (2012). A Survey of Monte Carlo Tree Search Methods. IEEE Transactions on Computational Intelli -gence and AI in Games, 4(1), 1–43

    2012

  8. [8]

    Świechowski, M., Godlewski, K., Sawicki, B. et al. (2023). Monte Carlo Tree Search: a review of recent modifications and applications. Artificial Intelligence Review, 56, 2497 –2562

    2023

  9. [9]

    Zhang, G., Peng, Y., & Xu, Y. (2022). An Efficient Dynamic Sampling Policy for Monte Carlo Tree Search. 2022 Winter Simulation Conference (WSC), 2760–2771

    2022

  10. [10]

    Buckman, J., Hafner, D., Tucker, G., Brevdo, E., & Lee, H. (2018). Sample-Efficient Reinforcement Learning with Stochas-tic Ensemble Value Expansion. Advances in Neural Information Processing Systems (NeurIPS), 31

    2018

  11. [11]

    Wu, D.J. (2019). Accelerating Self -Play Learning in Go. arXiv preprint arXiv:1902.10565

    Pith/arXiv arXiv 2019

  12. [12]

    Husna, A., & Müller, M. (2025). Analysing KataGo: A Com-parative Evaluation Against Perfect Play in the Game of Go. Computers and Games: 12th International Conference (CG 2024), 43 –53

    2025

  13. [13]

    Yu, T., Kumar, S., Gupta, A., Levine, S., Hausman, K., & Finn, C. (2020). Gradient surgery for multi -task learning. 34th International Conference on Neural Information Processing Systems (NeurIPS)

    2020

  14. [14]

    Wu, T.-R., et al. (2025). MiniZero: Comparative Analysis of AlphaZero and MuZero on Go, Othello, and Atari Games. IEEE Transactions on Games, 17(1), 125 –137

    2025

  15. [15]

    Cipolla, R., Gal, Y., & Kendall, A. (2018). Multi -task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics. 2018 IEEE/CVF CVPR, 7482–7491

    2018

  16. [16]

    Vlastelica, M., Blaes, S., Pinneri, C., & Martius, G. (2023). Mind the Uncertainty: Risk-Aware and Actively Explor-ing Model -Based Reinforcement Learning. arXiv preprint arXiv:2309.05582

    Pith/arXiv arXiv 2023

  17. [17]

    Jafferjee, T., Imani, E., Talvitie, E.J., White, M., & Bowl-ing, M. (2020). Hallucinating Value: A Pitfall of Dyna-style Planning with Imperfect Environment Models. arXiv preprint arXiv:2006.04363

    Pith/arXiv arXiv 2020

  18. [18]

    Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017). On calibration of modern neural networks. 34th International Conference on Machine Learning (ICML), 70, 1321–1330

    2017

  19. [19]

    Gal, Y., & Ghahramani, Z. (2016). Dropout as a Bayesian approximation: representing model uncertainty in deep learning. 33rd International Conference on Machine Learning (ICML), 48, 1050–1059

    2016

  20. [20]

    Wang, J., Zhu, T., Li, H., Hsueh, C.-H., & Wu, I.-C. (2018). Belief-State Monte Carlo Tree Search for Phantom Go. IEEE Transactions on Games, 10(2), 139 –154

    2018

  21. [21]

    Liu, Y., et al. (2025). AlphaGo Moment for Model Architecture Discovery. arXiv preprint arXiv:2507.18074

    Pith/arXiv arXiv 2025

  22. [22]

    Haarnoja, T., Zhou, A., Abbeel, P., & Levine, S. (2018). Soft Actor -Critic: Off -Policy Maximum Entropy Deep Rein -forcement Learning with a Stochastic Actor. arXiv preprint arXiv:1801.01290

    Pith/arXiv arXiv 2018

  23. [23]

    Gal, Y. (2016). Uncertainty in Deep Learning. PhD Thesis, University of Cambridge

    2016

  24. [24]

    Z., Konidaris, G

    Lovering, C., Forde, J. Z., Konidaris, G. D., Pavlick, E., & Littman, M. L. (2022). Evaluation Beyond Task Performance: Analyzing Concepts in AlphaZero in Hex. arXiv preprint arXiv:2211.14673

    Pith/arXiv arXiv 2022

  25. [25]

    Tian, Y., et al. (2019). ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero. International Conference on Machine Learning (ICML)

    2019

  26. [26]

    Zhang, S., et al. (2024). Look Before You Leap: Enhancing Attention and Vigilance Regarding Harmful Content with GuidelineLLM. arXiv preprint arXiv:2412.10423

    Pith/arXiv arXiv 2024

  27. [27]

    Ye, W., Liu, S.-W., Kurutach, T., Abbeel, P., & Gao, Y. (2021). Mastering Atari Games with Limited Data. arXiv preprint arXiv:2111.00210

    Pith/arXiv arXiv 2021

  28. [28]

    Schrittwieser, J., et al. (2021). Online and offline reinforcement learning by planning with a learned model. 35th Interna-tional Conference on Neural Information Processing Systems (NeurIPS)

    2021

  29. [29]

    Chow, Y., Ghavamzadeh, M., Janson, L., & Pavone, M. (2018). Risk-Constrained Reinforcement Learning with Percentile Risk Criteria. Journal of Machine Learning Research, 18(167), 1–51

    2018

  30. [30]

    G., Dabney, W., & Munos, R

    Bellemare, M. G., Dabney, W., & Munos, R. (2017). A Distri-butional Perspective on Reinforcement Learning. 34th Interna-tional Conference on Machine Learning (ICML), 449–458

    2017

  31. [31]

    M., Broekens, J., Plaat, A., & Jonker, C

    Moerland, T. M., Broekens, J., Plaat, A., & Jonker, C. M. (2023). Model-based Reinforcement Learning: A Survey. Foun-dations and Trends in Machine Learning, 16(1), 1–118

    2023

  32. [32]

    Shi, C., Yang, K., Yang, J., & Shen, C. (2024). Transformers as game players: provable in-context game-playing capabilities of pre-trained models. 38th Conference on Neural Information Processing Systems (NeurIPS)

    2024

  33. [33]

    -H., Kao, C

    Hsieh, Y. -H., Kao, C. -C., & Yuan, S. -M. (2025). Imitating Human Go Players via Vision Transformer. Algorithms, 18(2), 61

    2025

  34. [34]

    S., et al

    Shai, A. S., et al. (2024). Transformers represent belief state geometry in their residual stream. arXiv preprint arXiv:2405.15943

    Pith/arXiv arXiv 2024

  35. [35]

    S., et al

    Hu, E. S., et al. (2024). The Belief State Transformer. Interna-tional Conference on Learning Representations (ICLR)

    2024

  36. [36]

    Toward competitive multi- agents in Polo game based on reinforcement learning

    Movahedi, Z., Bastanfard, A. Toward competitive multi- agents in Polo game based on reinforcement learning. Multimed Tools https://doi.org/10.1007/s11042-021- 10968-z

    doi:10.1007/s11042-021-

  37. [37]

    Seyyed Amir Hadi Minoofam, Mohammad Mahdi Dehshibi, Azam Bastanfard, and Parvin Eftekhari. 2012. Ad-hocMa’qeli script generation using block cellular automata. J. Cell. Autom. 7, 4 (2012), 321–334

    2012

  38. [38]

    Improving the Accuracy of the Annotation Algorithm in Pattern-Based Tennis Game Video,

    A. Bastanfard and D. Amirkhani, "Improving the Accuracy of the Annotation Algorithm in Pattern-Based Tennis Game Video," 2021 29th Iranian Conference on Electrical Engineering (ICEE) , Tehran, Iran, 10.1109/ICEE52715.2021.9544273

    arXiv 2021