Pith. sign in

REVIEW 4 major objections 5 minor 40 references

Minimal Batch Adaptive Learning Policy Engine for Real-Time Mid-Price Forecasting in High-Frequency Trading

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read ALPE, a batch-free reinforcement learning agent, is claimed to achieve the lowest mid-price forecasting error across 100 S&P 500 stocks.

desk verdict The Extended feature set feeds the target mid-price straight into the model, so the headline forecasting advantage is leakage, not skill. read the letter →

arxiv 2412.19372 v2 pith:FL6RGJ2G submitted 2024-12-26 q-fin.ST cs.LG

Adamantios Ntakaris , Gbenga Ibikunle This is my paper · ORCID
classification q-fin.STcs.LG
keywords reinforcementlearninghigh-frequencytradingmid-priceforecastinglimitorderbookS&P500stocksAdaptivePolicyEngineRRMSE
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 claims that ALPE, a reinforcement learning agent built for batch-free, event-by-event operation, forecasts the limit order book mid-price more accurately than a range of statistical and deep learning benchmarks. The evaluation uses Level 1 NASDAQ data for 100 S&P 500 stocks over September to November 2022, with six input feature sets and two error metrics, RMSE and the newly proposed RRMSE. The reported result is that ALPE achieves the lowest forecasting error in every setting, and that its advantage over ARIMA, Naive, MLP, CNN, LSTM, GRU, and RBFNN is statistically significant in most pairwise comparisons. The paper's case for caring is that a model requiring only the current LOB state, with no batch retraining, is better suited to the non-stationary, rapidly moving conditions of high-frequency trading.

What carries the argument

The load-bearing mechanism is the ALPE agent: a Markov decision process over LOB states with continuous actions bounded in $[-0.1, 0.14]$, a reward $R_t = -|a_t - p_t|\,(1 - \epsilon_t)$, a discount factor $\gamma = 0$, and an epsilon-greedy policy with adaptive decay $\epsilon_{t+1} = \max(0.0001, \epsilon_t \cdot 0.9996)$. The exploitation network is an MLP with eight hidden layers of 64 ReLU units, batch normalization, and a single linear output, trained for two epochs with the Adam optimizer. Feature importance transforms (MDI and GD) and the RRMSE metric are used to build input variants and to compare errors across stocks.

What would settle it

Re-run ALPE with the mid-price feature $u_2$ removed from all feature sets, or with every feature lagged one event so the target is $p_{t+1}$; if a trivial model that outputs $u_2$ matches ALPE's reported RMSE, then the zero-lag protocol leaks the answer.

Watch

Extended reading notes

Core claim

The central discovery reported is that a model-free, value-based RL agent with an eight-hidden-layer MLP policy network, adaptive epsilon decay, and a reward that penalizes distance from the true mid-price can outperform batch-trained alternatives on mid-price forecasting. The authors report consistent lowest RMSE and RRMSE scores across 100 stocks, all six feature sets, and all benchmark models, and they support this with a Friedman test followed by Conover post-hoc pairwise comparisons. The paper interprets this as evidence that minimal-batch, zero-lag reinforcement learning adapts to market shifts in real time and provides a practical route to low-latency forecasting.

Load-bearing premise

The protocol assumes the model's input at event $t$ does not contain the mid-price $p_t$ that is being predicted, but the Extended feature set includes $u_2 = (P^{ask}_t + P^{bid}_t)/2$, which is exactly $p_t$, so the reported errors depend on that assumption being satisfied.

Editorial extensions

If this is right

  • If ALPE's reported advantage is real, online reinforcement learning with a one-state look-back can beat batch-trained deep networks, which would make low-latency adaptive forecasting feasible in practice.
  • The adaptive epsilon-decay schedule, fixed across all stocks, offers a single, tuning-light exploration-exploitation balance that the paper argues is sufficient across a 100-stock universe.
  • RRMSE, by normalizing RMSE by the event mid-price, gives a stock-independent error scale that the paper shows is especially informative for lower-volume names.
  • A two-epoch training budget on a single LOB state means the policy network can be updated per event, a property the paper links to the approximately 10-event stationary windows observed in LOB data.
  • The statistically significant pairwise wins over most baselines, as reported, indicate the error reductions are unlikely to be chance fluctuations.

Reading between the lines

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

  • We infer that the reported errors may largely measure how faithfully ALPE copies the mid-price feature $u_2$ from its input, since the target at event $t$ is also present as a feature; removing $u_2$ or shifting the target to $p_{t+1}$ would test whether any genuine predictive skill remains.
  • RRMSE could serve as a broadly useful normalized error metric for comparing forecasting models across assets of different price scales, but its value in this paper is entangled with the same target-leak issue because the denominator is the predicted quantity.
  • A stock-specific or volatility-dependent epsilon decay schedule is a natural extension; the paper itself notes the fixed decay factor is not ideal for low-volume stocks, and our reading suggests this is where ALPE's gains are least consistent.
  • The event-driven, one-state protocol could be reframed as a one-event-ahead forecasting benchmark, which would connect this RL setup to the large body of order-flow and mid-price movement prediction tasks.
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 / 5 minor

Summary. The paper proposes ALPE, a reinforcement-learning agent for event-by-event mid-price forecasting using Level 1 limit order book data from 100 S&P 500 stocks over September-November 2022. The agent is compared against Naive, ARIMA, MLP, CNN, LSTM, GRU, and RBFNN across six feature sets using RMSE and RRMSE, and the paper claims that ALPE consistently achieves the lowest forecasting error. The central claim is not supported by the stated evaluation design: the target mid-price p_t appears directly as an input feature in the Extended set, the exploration action in Eq. (17) injects p_t into the agent's output, and the training target in Eq. (24) is a function of the network's own current output. These issues mean the reported error numbers do not measure predictive skill under the zero-lag protocol described in the manuscript.

Significance. If the forecasting claim were valid, the paper would provide a useful large-scale comparison of an adaptive, batch-free mid-price forecaster across 100 stocks, with a potentially useful normalized error metric in RRMSE. The GD/MDI feature-importance pipeline and the event-by-event protocol are also relevant to the HFT forecasting literature. However, the paper provides no code or data, and the load-bearing evaluation is compromised by direct target leakage through the features and the action construction. The manuscript therefore does not currently establish that ALPE has predictive skill, and the quantitative claims in the abstract, Section 4, and Section 5 are not interpretable as forecasting performance as stated.

major comments (4)
  1. [§3.1, §3.2, Table 1] The target variable is included in the model input. Equation (1) defines p_t = (P^ask_t + P^bid_t)/2, and Table 1 defines the Extended feature u2 = (P^ask_1 + P^bid_1)/2, which is exactly p_t at the same event. Under the zero-lag, event-by-event protocol described in §3.1, where the state at event t is used to predict the mid-price at event t, the Extended feature sets give every model direct access to the target. The reported RMSE and RRMSE values therefore measure target reconstruction and leakage, not forecasting error. This invalidates the central performance claim as stated.
  2. [§3.3.1, Eq. (17)] The exploration action injects the contemporaneous target into the agent's output. Equation (17) sets a_t = p_t + α^e_t whenever exploration fires, so ALPE's reported output is partly or wholly constructed from the true mid-price at the same event. This leakage affects even the Simple feature sets and gives ALPE an advantage that has nothing to do with learned forecasting ability. The reward function in Eq. (19) also evaluates the action against p_t, reinforcing the same contemporaneous dependence.
  3. [§3.3.3, Eqs. (24)-(25)] The training target is self-referential. Equation (24) defines f_{π,target} as an expression that includes f_π(s_t, a_t; θ_ALPE), the current network output, and Eq. (25) minimizes the squared difference between the network output and this target. Because the target moves with the estimator, the loss does not define a fixed external regression target and can be made small by construction. This undermines the claim that the network learns a policy or value function from the reward signal.
  4. [§5 vs. Table 3] The headline quantitative claim in the conclusion is inconsistent with the reported table. Section 5 states that for Amazon under Extended GD, ALPE achieved an RRMSE of 2.484E-04, but Table 3 reports an RRMSE of 2.944E-04 for ALPE under Exte GD. This discrepancy in the paper's central example, together with the target-leakage issues above, makes it impossible to take the numerical superiority claims at face value without access to code and data.
minor comments (5)
  1. [§3.1] The text refers to 'our proposed DQR model,' but the model introduced in the paper is ALPE; this appears to be a leftover from an earlier draft.
  2. [Table 3] For Amazon Simple, the GRU row reports an RRMSE of 9.206E-03 alongside an RMSE of 1.048E-01; this is an order of magnitude larger than the other RRMSE values and is inconsistent with the stated relationship RRMSE = RMSE/p_t, suggesting a typographical error.
  3. [Table A.9 and Table 2] The ticker 'OALPEY' in Table A.9 appears to be a typo for ORLY, and 'L VS' in Table 2 and Table A.6 appears to be a typo for LVS.
  4. [Throughout] There are numerous typographical and formatting errors, including 'definied' in Eq. (2), 'kernalised' in §3.2, 'V AR' in Section 2, and inconsistent use of 'Extensive' versus 'Extended.' These should be corrected in any revision.
  5. [Eq. (17) and surrounding text] The notation around the action is confusing: α_t, α^e_t, and the action bounds a_min, a_max are introduced without making clear whether α_t is a parameter or an action variable, and Eq. (17) mixes the exploration perturbation with the learned policy output. Clarifying this notation would help, but it does not resolve the target-leakage problem.

Circularity Check

3 steps flagged · score 8.0 of 10

The target mid-price is fed into ALPE as Extended feature u2 and as the exploration action, so the reported forecasting error advantage is substantially a leakage artefact rather than predictive skill.

  1. self definitional [Section 3.1, Eq. (1); Section 4, Table 1]
    "The mid-price is defined as the average of the best bid and ask LOB prices: pt = pAskt + pBidt /2, (1)... Extended Basic u2 = (Pask1 +Pbid1 )/2 ... These include the mid-price ( u2), which is directly connected to our forecasting objective."

    Equation (1) defines the forecast target p_t as (P^ask_t+P^bid_t)/2. Table 1 defines the Extended feature u2 as (P^ask_1+P^bid_1)/2, and the text states that u2 is the mid-price 'directly connected to our forecasting objective.' Under the event-by-event protocol described in Section 3.1, where the state is the current LOB information and the objective is to forecast p_t, u2 is literally the target. Any model using the Extended feature set can copy this input; the reported RMSE/RRMSE for Extended inputs therefore measures reconstruction of an input feature, not an out-of-sample forecast. No lag or shift between feature time t and target time t is specified, so the reduction follows from the paper's own equations.

  2. fitted input called prediction [Section 3.3.1, Eq. (17)]
    "at = ( pt + αet, ξ ≤ ϵt (exploration) fπ(st, αt; θALP E), otherwise (exploitation) (17)"

    During exploration, ALPE's action is set to the true current mid-price p_t plus a small exploration offset. The reward R_t = -|a_t - p_t|(1-ϵ_t) and the reported prediction error are then minimized essentially for free whenever exploration fires, because the 'prediction' contains the target. This effect holds even for the Simple feature set, which does not include u2. Thus ALPE's advantage over competitors lacking this direct target injection is artificially forced by the action definition rather than by learned forecasting skill.

1 more flagged steps
  1. other [Section 3.3.3, Eqs. (24)-(25)]
    "fπ,target (st, at; θALP E) = Rt − |αt − fπ(st, at; θALP E)| ×(1 − ϵt), (24) ... L(θALP E) = (fπ,target (st, at; θALP E) − fπ(st, at; θALP E))2, (25)"

    The training target fπ,target contains the current network output fπ(st, at; θALP E) on its right-hand side. The loss then minimizes the difference between fπ and a quantity that itself is shifted by |α_t - fπ|(1-ϵ_t); as the prediction changes, the target changes with it. This is a self-referential regression target rather than an independent ground-truth label, so the network is not trained against a fixed mid-price value. The reported performance therefore cannot be interpreted as evidence of convergence to an externally defined mid-price.

full rationale

The paper's central empirical claim is that ALPE consistently achieves the lowest forecasting error for mid-price movements, yet the manuscript's own equations make the target available as an input. Equation (1) defines p_t as the average of the best ask and bid prices; Table 1 defines the Extended feature u2 as exactly that average and labels it as the mid-price 'directly connected to our forecasting objective.' With the zero-lag, event-by-event protocol described in Section 3.1, the Extended feature set contains the forecast target. Independently, Eq. (17) injects p_t into ALPE's action whenever exploration fires, so even the Simple feature set gives ALPE direct access to the target during exploration. Eq. (24) compounds the problem by constructing the training target from the network's own current output. Together, these reductions mean that the reported RMSE/RRMSE values do not establish predictive skill: a substantial part of ALPE's 'forecast' can be read off the input or action definition. The self-citations to the authors' prior work on MDI/GD features and RRMSE are not the source of circularity; those methods are defined and applied in this paper. There is also an internal inconsistency in the reported numbers, since the conclusion quotes an Amazon Extended-GD RRMSE of 2.484E-04 while Table 3 reports 2.944E-04, and no code or data release is provided to check the target alignment. These concerns do not affect the circularity finding: the leakage is structural, following from Eqs. (1), Table 1, and Eq. (17), not from speculation about implementation details.

Assumptions & free parameters 6 free parameters · 3 assumptions · 1 invented entities

The central empirical claim rests on several modeling choices, the most important being that a zero-lag mid-price is a legitimate forecasting target. Since the target is a feature in the input, that assumption fails. Other entries are standard domain and modeling assumptions.

free parameters (6)
  • epsilon_decay = 0.9996
    Chosen to balance exploration and exploitation; authors note it should ideally be stock-specific but fixed across stocks for fair comparison.
  • epsilon_min = 0.0001
    Floor for exploration probability; defined in Section 3.3.1.
  • action_bounds = [-0.1, 0.1]
    Exploration adjustment range tied to tick size; chosen by hand.
  • MLP_topology = 8 hidden layers x 64 neurons, 2 epochs
    Selected by grid search over up to 10 layers, 4-1024 neurons, with and without batch norm, Adam vs GD; tuned to maximize average performance across the majority of stocks.
  • GD_feature_importance_learning_rate = 0.001
    Reported effective for most stocks; T=10 iterations and delta=0.001 for numerical stability.
  • baseline_window = 10 LOB states
    Competitors trained on current plus nine historical LOB states; chosen based on authors' prior observation of stationarity alternation every ~10 events.
assumptions (3)
  • ad hoc to paper Zero-lag event-by-event regression: state at event t may be used to predict mid-price at the same event t.
    This is the defining premise of the protocol in Section 3.1, and it fails because Eq 1 makes p_t a function of the state features, including u2.
  • domain assumption Immediate-reward Markov formulation: setting gamma=0 and using only the current LOB state captures the HFT forecasting problem.
    Section 3.3.1; no delayed rewards or temporal dependencies are modeled, which may be too restrictive for forecasting.
  • domain assumption Baseline comparison fairness: training competitors on 10 LOB states while ALPE uses 1 state is an appropriate comparison.
    Section 3.1; the comparison mixes batch and online protocols, and no online versions of the baselines are included.
invented entities (1)
  • ALPE agent
    purpose: Proposed online RL regressor for mid-price forecasting
    A model architecture, not a physical entity; no external falsifiable prediction outside the reported benchmark.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Minimal Batch Adaptive Learning Policy Engine for Real-Time Mid-Price Forecasting in High-Frequency Trading." pith.science (2026). https://pith.science/paper/FL6RGJ2G

@misc{pith2026241219372,
  author       = {Pith},
  title        = {Pith review of: Minimal Batch Adaptive Learning Policy Engine for Real-Time Mid-Price Forecasting in High-Frequency Trading},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FL6RGJ2G}},
  note         = {Machine review of arXiv:2412.19372}
}
read the original abstract

High-frequency trading (HFT) has transformed modern financial markets, making reliable short-term price forecasting models essential. In this study, we present a novel approach to mid-price forecasting using Level 1 limit order book (LOB) data from NASDAQ, focusing on 100 U.S. stocks from the S&P 500 index during the period from September to November 2022. Expanding on our previous work with Radial Basis Function Neural Networks (RBFNN), which leveraged automated feature importance techniques based on mean decrease impurity (MDI) and gradient descent (GD), we introduce the Adaptive Learning Policy Engine (ALPE) - a reinforcement learning (RL)-based agent designed for batch-free, immediate mid-price forecasting. ALPE incorporates adaptive epsilon decay to dynamically balance exploration and exploitation, outperforming a diverse range of highly effective machine learning (ML) and deep learning (DL) models in forecasting performance.

Figures

Figures reproduced from arXiv: 2412.19372 by the authors.

Figure 1
Figure 1. The figure provides an overview of our experimental protocol, which is structured around two [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. RRMSE scores for all six input datasets (i.e., [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. Percentage error reduction between RMSE and RRMSE across stocks, plotted against total [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Performance-based volume profiling: The graph shows the volume distribution per stock, [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 37 canonical work pages

  1. [1]

    Online High-Frequency Trading Stock Forecasting with Automated Feature Clustering and Radial Basis Function Neural Networks

    A. Ntakaris, G. Ibikunle, Online high-frequency trading stock forecasting with auto- mated feature clustering and radial basis function neural networks, arXiv preprint arXiv:2412.16160 (2024)

    work page Pith review arXiv 2024

  2. [2]

    Alshawarbeh, A

    E. Alshawarbeh, A. T. Abdulrahman, E. Hussam, Statistical modeling of high fre- quency datasets using the arima-ann hybrid, Mathematics 11 (22) (2023) 4594

    work page 2023

  3. [3]

    H. Dong, X. Guo, H. Reichgelt, R. Hu, Predictive power of arima models in forecast- ing equity returns: a sliding window method, Journal of Asset Management 21 (6) (2020) 549–566

    work page 2020

  4. [4]

    A. H. Bukhari, M. A. Z. Raja, M. Sulaiman, S. Islam, M. Shoaib, P. Kumam, Frac- tional neuro-sequential arfima-lstm for financial market forecasting, IEEE Access 8 (2020) 71326–71338

    work page 2020

  5. [5]

    Zhu, G.-Y

    R. Zhu, G.-Y. Zhong, J.-C. Li, Forecasting price in a new hybrid neural network model with machine learning, Expert Systems with Applications 249 (2024) 123697

    work page 2024

  6. [6]

    R. Liu, H. Liu, H. Huang, B. Song, Q. Wu, Multimodal multiscale dynamic graph con- volution networks for stock price prediction, Pattern Recognition 149 (2024) 110211

    work page 2024

  7. [7]

    T. Yin, C. Liu, F. Ding, Z. Feng, B. Yuan, N. Zhang, Graph-based stock correlation and prediction for high-frequency trading systems, Pattern Recognition 122 (2022) 108209. doi:https://doi.org/10.1016/j.patcog.2021.108209. 20

    arXiv 2022

  8. [8]

    Dingli, K

    A. Dingli, K. S. Fournier, Financial time series forecasting-a deep learning approach, International Journal of Machine Learning and Computing 7 (5) (2017) 118–122

    work page 2017

Show all 40 references
  1. [9]

    A. U. Haq, A. Zeb, Z. Lei, D. Zhang, Forecasting daily stock trend using multi-filter feature selection and deep learning, Expert Systems with Applications 168 (2021) 114444

    2021

  2. [10]

    Arifovic, X.-z

    J. Arifovic, X.-z. He, L. Wei, Machine learning and speed in high-frequency trading, Journal of Economic Dynamics and Control 139 (2022) 104438

    2022

  3. [11]

    M. K. Mangat, E. Reschenhofer, T. Stark, C. Zwatz, High-frequency trading with machine learning algorithms and limit order book data, Data Science in Finance and Economics 2 (4) (2022) 437–463

    2022

  4. [12]

    Moews, G

    B. Moews, G. Ibikunle, Predictive intraday correlations in stable and volatile market environments: Evidence from deep learning, Physica A: Statistical Mechanics and its Applications 547 (2020) 124392

    2020

  5. [13]

    J. A. Sirignano, Deep learning for limit order books, Quantitative Finance 19 (4) (2019) 549–570

    2019

  6. [14]

    Nousi, A

    P. Nousi, A. Tsantekidis, N. Passalis, A. Ntakaris, J. Kanniainen, A. Tefas, M. Gab- bouj, A. Iosifidis, Machine learning for forecasting mid-price movements using limit order book data, IEEE Access 7 (2019) 64722–64736

    2019

  7. [15]

    Ibikunle, B

    G. Ibikunle, B. Moews, K. Rzayev, Can machine learning unlock new insights into high-frequency trading?, arXiv preprint arXiv:2405.08101 (2024)

    2024 arXiv

  8. [16]

    M¨ akinen, J

    Y. M¨ akinen, J. Kanniainen, M. Gabbouj, A. Iosifidis, Forecasting jump arrivals in stock prices: new attention-based network architecture using limit order book data, Quantitative Finance 19 (12) (2019) 2033–2050

    2019

  9. [17]

    J. W. Lee, Stock price prediction using reinforcement learning, in: ISIE 2001. 2001 IEEE International Symposium on Industrial Electronics Proceedings (Cat. No. 01TH8570), Vol. 1, IEEE, 2001, pp. 690–695

    2001

  10. [18]

    Y. Deng, F. Bao, Y. Kong, Z. Ren, Q. Dai, Deep direct reinforcement learning for financial signal representation and trading, IEEE Transactions on Neural Networks and Learning Systems 28 (3) (2016) 653–664

    2016

  11. [19]

    Y. Wang, Y. Miao, A. C. Wong, N. P. Granger, C. Michler, Domain-adapted learning and interpretability: Drl for gas trading, arXiv preprint arXiv:2301.08359 (2023)

    2023 arXiv

  12. [20]

    S. Sun, R. Wang, B. An, Reinforcement learning for quantitative trading, ACM Transactions on Intelligent Systems and Technology 14 (3) (2023) 1–29

    2023

  13. [21]

    Karpe, J

    M. Karpe, J. Fang, Z. Ma, C. Wang, Multi-agent reinforcement learning in a realistic limit order book market simulation, in: Proceedings of the first ACM international conference on AI in finance, 2020, pp. 1–7

    2020

  14. [22]

    Philip, Estimating permanent price impact via machine learning, Journal of Econometrics 215 (2) (2020) 414–449

    R. Philip, Estimating permanent price impact via machine learning, Journal of Econometrics 215 (2) (2020) 414–449. 21

    2020

  15. [23]

    Y. Cao, J. Zhai, Estimating price impact via deep reinforcement learning, Interna- tional Journal of Finance & Economics 27 (4) (2022) 3954–3970

    2022

  16. [24]

    X.-Z. He, S. Lin, Reinforcement learning equilibrium in limit order markets, Journal of Economic Dynamics and Control 144 (2022) 104497

    2022

  17. [25]

    Tsantekidis, N

    A. Tsantekidis, N. Passalis, A. Tefas, Modeling limit order trading with a continuous action policy for deep reinforcement learning, Neural Networks 165 (2023) 506–515

    2023

  18. [26]

    W. Fan, K. Liu, H. Liu, Y. Ge, H. Xiong, Y. Fu, Interactive reinforcement learning for feature selection with decision tree in the loop, IEEE Transactions on Knowledge and Data Engineering 35 (2) (2021) 1624–1636

    2021

  19. [27]

    I. K. Thajeel, K. Samsudin, S. J. Hashim, F. Hashim, Dynamic feature selection model for adaptive cross site scripting attack detection using developed multi-agent deep q learning model, Journal of King Saud University-Computer and Information Sciences 35 (6) (2023) 101490

    2023

  20. [28]

    Q. Liu, Q. Zhou, R. Yang, J. Wang, Robust representation learning by clustering with bisimulation metrics for visual reinforcement learning with distractions, in: Proceed- ings of the AAAI Conference on Artificial Intelligence, Vol. 37, 2023, pp. 8843–8851

    2023

  21. [29]

    Fathinezhad, P

    F. Fathinezhad, P. Adibi, B. Shoushtarian, H. Baradaran Kashani, J. Chanussot, Soft dimensionality reduction for reinforcement data clustering, World Wide Web 26 (5) (2023) 3027–3054

    2023

  22. [30]

    Elbaz, A

    K. Elbaz, A. Zhou, S.-L. Shen, Deep reinforcement learning approach to optimize the driving performance of shield tunnelling machines, Tunnelling and Underground Space Technology 136 (2023) 105104

    2023

  23. [31]

    Gauci, E

    J. Gauci, E. Conti, Y. Liang, K. Virochsiri, Y. He, Z. Kaden, V. Narayanan, X. Ye, Z. Chen, S. Fujimoto, Horizon: Facebook’s open source applied reinforcement learn- ing platform, arXiv preprint arXiv:1811.00260 (2018)

    2018 arXiv

  24. [32]

    Y. Fu, D. Wu, B. Boulet, Reinforcement learning based dynamic model combination for time series forecasting, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 36, 2022, pp. 6639–6647

    2022

  25. [33]

    Saadallah, K

    A. Saadallah, K. Morik, Online ensemble aggregation using deep reinforcement learn- ing for time series forecasting, in: 2021 IEEE 8th International Conference on Data Science and Advanced Analytics (DSAA), IEEE, 2021, pp. 1–8

    2021

  26. [34]

    Kuremoto, M

    T. Kuremoto, M. Obayashi, K. Kobayashi, Forecasting time series by sofnn with reinforcement learning, in: Proceedings of the 27th Annual International Symposium on Forecasting, Neural Forecasting Competition (NN3), New York, NY, USA, 2007, pp. 24–27

    2007

  27. [35]

    Hirata, T

    T. Hirata, T. Kuremoto, M. Obayashi, S. Mabu, K. Kobayashi, Forecasting real time series data using deep belief net and reinforcement learning., J. Robotics Netw. Artif. Life 4 (4) (2018) 260–264. 22

    2018

  28. [36]

    Zhuang, V

    D. Zhuang, V. J. Gan, Z. D. Tekler, A. Chong, S. Tian, X. Shi, Data-driven predictive control for smart hvac system in iot-integrated buildings with time-series forecasting and reinforcement learning, Applied Energy 338 (2023) 120936

    2023

  29. [37]

    F. Liu, C. Quek, G. S. Ng, Neural network model for time series prediction by reinforcement learning, in: Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005., Vol. 2, IEEE, 2005, pp. 809–814

    2005

  30. [38]

    Ntakaris, G

    A. Ntakaris, G. Mirone, J. Kanniainen, M. Gabbouj, A. Iosifidis, Feature engineering for mid-price prediction with deep learning, IEEE Access 7 (2019) 82390–82412

    2019

  31. [39]

    Ntakaris, M

    A. Ntakaris, M. Gabbouj, J. Kanniainen, Optimum output long short-term memory cell for high-frequency trading forecasting, arXiv preprint arXiv:2304.09840 (2023)

    2023 arXiv

  32. [40]

    D. P. Kingma, Adam: A method for stochastic optimization, arXiv preprint arXiv:1412.6980 (2014). Appendix A. Table A.1: RMSE and RRMSE Scores for BAC, BRK, and EIX. Stock Set Model RMSE RRMSE Set RMSE RRMSEBAC Simple Naive 6.011E-01±7.694E-03 5.349E-03±8.458E-05 Exte 6.003E-01...

    2014 arXiv

Pith tools

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