Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Comprehend, Divide, and Conquer: Feature Subspace Exploration via Multi-Agent Hierarchical Reinforcement Learning

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 This paper proposes replacing one RL agent per feature with a hierarchy of agents that prune clusters, claiming better feature selection accuracy and average active-agent count O(log N) instead of O(N).

desk verdict Hybrid LLM/GMM state extraction for hierarchical RL feature selection is a real novelty, but the O(log N) efficiency proof and the 'all tasks' claim do not hold as written. read the letter →

arxiv 2504.17356 v3 pith:72TNJDDU submitted 2025-04-24 cs.AI cs.LG

Weiliang Zhang , Xiaohan Huang , Yi Du , Ziyue Qiao , Qingqing Long , Zhen Meng , Yuanchun Zhou , Meng Xiao This is my paper · ORCID
classification cs.AIcs.LG
keywords featureselectionhierarchicalreinforcementlearningmulti-agentsubspaceexplorationlargelanguagemodelsGaussianmixturetabulardatawrappermethods
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

Feature selection for tabular machine learning usually means scanning hundreds or thousands of features one by one, and RL-based wrappers multiply that cost by assigning one agent to each feature. This paper proposes HRLFS, which first represents every feature by both a fitted Gaussian mixture model and a semantic embedding of its name or description, clusters the features into a tree, and places an RL agent at each cluster node. A node that says 'drop' prunes its entire subtree, so many features can be rejected in one decision; a node that says 'select' delegates to its children. The paper's central claim is that this hierarchical delegation improves downstream model performance on all 21 tested datasets while cutting the average number of active agents from O(N) to O(log N) under the proof's assumptions, with measured active-agent reductions of 70.61% to 82.30% and runtime reductions of 35.49% to 55.26% against one-agent-per-feature.

What carries the argument

The load-bearing object is the hierarchical agent tree over feature clusters, built by agglomerative clustering on hybrid feature states. The state of each feature is a vector combining GMM parameters (capturing the shape of the feature's distribution) with a semantic embedding from an LLM (capturing what the feature means), so clustering groups features that are both statistically and semantically similar. In the decision process, an internal node's 'drop' action deactivates all its descendants, which is the pruning mechanism that turns per-feature decisions into per-subspace decisions; the 'select' action recurses to children. The recurrence $E(N) = 1 + 2pE(N/2)$ for the expected number of active agents, with $p$ the probability that a non-leaf agent delegates, is what the efficiency proof manipulates to obtain $O(\log N)$ at $p=1/2$.

What would settle it

Record the empirical delegation rate of trained non-leaf agents on a benchmark dataset. If the share of non-leaf agents that choose 'select' is consistently above $1/2$, the closed form $E(N) = ((2p)^{\log_2(N+1)} - 1)/(2p - 1)$ predicts active-agent counts that grow toward linear in $N$; for MNIST's 784 features, the $p=1/2$ prediction is roughly $2\log_2(785)\approx 19$ active agents, whereas the paper reports 156.73, so a small logging change during exploration would settle whether the logarithmic bound is actually observed.

Watch

Extended reading notes

Core claim

The discovery the paper argues for is that feature selection is best organized as hierarchical coarse-to-fine pruning rather than independent per-feature voting. The pipeline has three stages: comprehend, divide, conquer. Comprehend combines a Gaussian Mixture Model of each feature's value distribution with an LLM-produced semantic state from feature metadata; divide uses agglomerative clustering over those hybrid states to build a binary agent tree; conquer lets each internal agent choose between 'select' (delegate to children) and 'drop' (prune the whole subtree), with leaf agents deciding individual features. The paper claims that HRLFS outperforms filter, deep learning-based, and RL-based baselines on every task in its benchmark, and that the hierarchy's expected active agents follow $E(N)=1+2pE(N/2)$, giving $O(\log N)$ when the delegation probability is $p=1/2$, in contrast to the $O(N)$ agents of one-agent-per-feature methods.

Load-bearing premise

The load-bearing premise is Assumption 2 in Section 3.5: every non-leaf agent delegates to its children independently with a fixed probability $p = 1/2$; the paper treats this as a learned policy property and never measures it, so the average-case $O(\log N)$ claim stands or falls with an unverified constant.

Editorial extensions

If this is right

  • On the paper's 21 benchmark datasets, HRLFS is reported as the best-performing method across classification, multi-label classification, and regression tasks.
  • HRLFS is reported to cut active agents by 70.61% to 82.30% and runtime by 35.49% to 55.26% relative to one-agent-per-feature reinforcement learning.
  • If the $O(\log N)$ average-case bound holds, high-dimensional tabular feature selection becomes practical at feature counts where per-feature agent architectures are prohibitive.
  • The hierarchy granularity is a tunable lever: finer hierarchies improve performance and lower variance but activate more agents.
  • The method transfers best to datasets whose column names or descriptions carry genuine meaning, because the semantic state is what lets the clustering recognize related features.

Reading between the lines

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

  • The paper's proof needs the delegation probability $p$ to be close to $1/2$, but it never reports measuring $p$ on trained agents; a direct measurement on the benchmark trees would turn the average-case complexity claim from a conditional result into an empirical one.
  • The observed active-agent counts, such as roughly 157 of 784 for MNIST, are far above the $2\log_2(N+1)$ curve, so the measured savings may reflect cluster-level pruning rather than the logarithmic bound itself.
  • A natural extension is to use the same hybrid state and hierarchy for feature transformation or feature generation, where the search space is also exponential and semantic grouping could prune it.
  • Because the framework only needs tabular features and their names, it could slot into automated ML pipelines as a preprocessing layer, but the LLM embedding cost would need to be amortized across many downstream evaluations.
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. This paper introduces HRLFS, a feature selection method that builds a hybrid state representation from GMM distributional features and LLM-based semantic embeddings, constructs a hierarchical clustering over features, and learns actor-critic policies for select/drop decisions at each tree node. The stated contributions are improved downstream performance relative to eight filter, embedded, deep-learning, and RL baselines on 21 tabular datasets, and an O(log N) average number of active agents compared with the O(N) one-agent-per-feature baseline. The paper also reports ablations on state representation, hierarchy granularity, clustering algorithm, policy backend, and hyperparameters.

Significance. The work addresses a real scalability bottleneck in RL-based feature selection, and the overall design is a plausible and potentially useful contribution: the evaluation covers diverse datasets, the ablations are informative, and the code and data are released. If the claims were fully supported, the hierarchy-plus-semantic-state architecture would be a meaningful step beyond one-agent-per-feature methods. However, the two headline claims are not established as stated: the universal performance superiority is contradicted by specific rows in Table 1, and the theoretical O(log N) efficiency result depends on an unvalidated assumption about the learned delegation probability, while the measured active-agent counts in Table 2 are more consistent with a constant-fraction-of-N behavior than with logarithmic scaling.

major comments (4)
  1. [Section 5.1, Table 1] The claim that 'HRLFS outperforms the classical, deep learning-based, and RL-based methods on all tasks' is contradicted by Table 1. On Cao, RLAS achieves 90.57 while HRLFS achieves 89.37±0.13, and on Openml_637, RLAS achieves 42.04 while HRLFS achieves 40.54±0.37. In addition, baseline standard deviations are not reported, so several wins of less than one point (e.g., Credit_Default and Ionosphere) cannot be distinguished from noise. The performance claim should be restricted to the datasets where the advantage is statistically supported, and the authors should provide baseline error bars or significance tests.
  2. [Section 3.5, Eq. (20) and Eq. (23)] The closed-form solution at p=1/2 is arithmetically incorrect. With E(1)=1 and E(N)=1+E(N/2), unrolling gives E(N)=log2(N+1)+1 for N=2^h−1. The displayed equality E(N)=log2(N+1)+1=2log2(N+1)+O(1) is false because the left- and right-hand sides differ by a term that grows logarithmically in N. The O(log N) conclusion survives the correction, but the theorem and the displayed equations must be fixed.
  3. [Section 3.5, Assumption 2; Section 5.9; Table 2] The O(log N) average-case result holds only if each non-leaf agent delegates with fixed probability p=1/2. This probability is a property of the learned policy, not a design parameter, and Section 5.9 validates only Assumption 1 (balance factor and tree height), never Assumption 2. The measured active-agent counts in Table 2 (16.46 of 57, 13.45 of 77, 104.39 of 400, and 156.73 of 784) are roughly constant fractions of N (17–29%), not logarithmic; for these N values an O(log N) process with p=1/2 would activate about 7–19 agents. The authors should measure the empirical delegation probability or effective branching factor and either prove the expected active-agent count under the observed value or substantially weaken the complexity claim.
  4. [Section 5.2, Table 2] The paper states that the reduction in active agents 'directly translates' into time savings, but the reported runtime reductions (35.49–55.26%) are far smaller than the active-agent reductions (70.61–82.30%). This indicates that runtime is not dominated by agent count; downstream model evaluation and state construction likely dominate. The efficiency claim should be rephrased, and the definition of 'active agents' (per decision step versus cumulative over training) should be stated explicitly, since the current table mixes both readings.
minor comments (5)
  1. [Table 2] The numeric entries are run together: '5616.46' should likely be '57 16.46', '7613.45' should likely be '77 13.45', '400104.39' should likely be '400 104.39', and '784156.73' should likely be '784 156.73'. This makes the table hard to read and easy to misinterpret.
  2. [Section 3.4] The statement that the method 'totally using 2N−1 agents to make log(N) decisions on N features' conflates the total number of agents with the number of active agents per decision step; please rephrase to distinguish per-step active agents from the total agent population.
  3. [Section 3.5, Assumption 2] Assumption 2 is introduced with p∈[0,1], but the theorem requires p=1/2; state explicitly that p=1/2 is an assumption that must be empirically validated, not a consequence of the framework.
  4. [Section 5.5] The claim that GPT-4's 'data screening and cleaning criteria reduce errors and biases' is not tested by the model comparison and should be removed or supported with evidence.
  5. [Section 3.2, Eq. (2)] The notation mixes the dataset-level log-likelihood with per-feature parameters θ*_i; please define Θ_i explicitly before the EM step.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: O(log N) proof is conditional on an unverified p=1/2 assumption, which is a correctness gap, not circularity.

full rationale

The paper's downstream-performance claim is empirical: Table 1 compares HRLFS against externally defined baselines on held-out folds, so that claim is not equivalent to its own inputs. The efficiency claim is supported by a conditional theorem, not by a circular fit. Section 3.5 states 'Under the assumptions below, the average-case time complexity satisfies O(log N) when p=1/2' and then solves the recurrence E(N)=1+2pE(N/2). The logarithmic bound is a mathematical consequence of the assumed value p=1/2; it is not obtained by fitting p to the measured active-agent counts, nor does the proof use the conclusion 'O(log N)' as a premise. What is missing is empirical justification that the trained policy actually delegates with probability p=1/2: Section 5.9 validates only Assumption 1 (balance factor and tree height), and the active-agent reductions in Table 2 (70.61%-82.30%) are not a direct test of p and appear in tension with logarithmic scaling. That is a correctness and rigor gap, not a circularity. The self-citations (e.g., [14], [19], [20]) are contextual and not load-bearing for the core derivation; the main comparisons are against external baselines. Therefore no circular step is exhibited, and the circularity score is 0.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The method rests on several fitted hyperparameters and two structural assumptions in the complexity proof. The empirical performance claim is grounded in external benchmark comparisons, so it is not circular, but the efficiency theorem's conclusion is largely contained in the p=1/2 assumption. No invented entities such as particles or mediators are introduced; the hierarchical agent tree is a software architecture component.

free parameters (6)
  • k (GMM components per feature) = Maximum of BIC search per feature
    Controls granularity of distribution state; chosen by BIC on each feature's values (Section 3.2, Section 4.4).
  • alpha (reward weight) = 0.4
    Chosen via hyperparameter study on SpectF (Section 5.8), then fixed for all datasets; balances performance versus feature quantity.
  • lambda (quantity suppression) = 0.6
    Chosen via hyperparameter study on SpectF (Section 5.8), then fixed for all datasets.
  • gamma (discount factor) = 0.9
    Set in Section 4.4; standard choice, not tuned per dataset.
  • actor and critic learning rates = 0.001 and 0.01
    Set following reference [49] in Section 4.4.
  • agent hidden sizes, memory size, minibatch size = (64, 8), 400, 32
    Hyperparameters set in Section 4.4.
assumptions (6)
  • domain assumption The learned hierarchy approximates a perfect binary tree (Assumption 1).
    Used in Theorem 3.3 and 3.5 for the recurrence; empirically validated for 21 datasets in Section 5.9 but not guaranteed by the algorithm.
  • ad hoc to paper Each non-leaf agent independently delegates to children with fixed probability p=1/2 (Assumption 2).
    This is the key assumption producing O(log N); p is actually a property of the learned policy and is never shown to be 1/2. Section 3.5.
  • domain assumption GPT-4 and text-embedding-3-large embeddings capture semantically meaningful feature relationships for clustering.
    The method's core 'comprehend' step relies on LLM embeddings; only evaluated indirectly through ablations (Figure 7), with no external validation.
  • domain assumption GMM with BIC-selected k accurately represents each feature's distribution.
    Standard statistical modeling assumption used for state construction (Equations 1 and 2).
  • domain assumption The downstream Random Forest evaluation on a fixed 20% validation split gives a reliable reward signal.
    Reward is computed from this evaluation; variance across 5 runs is reported only for the proposed method.
  • standard math Law of total expectation and geometric series in the complexity proof.
    Used in Lemma 3.4 and Theorem 3.5; the final algebra contains an inconsistency (E(N)=log2(N+1)+1 versus stated 2log2(N+1)+O(1)).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Comprehend, Divide, and Conquer: Feature Subspace Exploration via Multi-Agent Hierarchical Reinforcement Learning." pith.science (2026). https://pith.science/paper/72TNJDDU

@misc{pith2026250417356,
  author       = {Pith},
  title        = {Pith review of: Comprehend, Divide, and Conquer: Feature Subspace Exploration via Multi-Agent Hierarchical Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/72TNJDDU}},
  note         = {Machine review of arXiv:2504.17356}
}
read the original abstract

Feature selection aims to preprocess the target dataset, find an optimal and most streamlined feature subset, and enhance the downstream machine learning task. Among filter, wrapper, and embedded-based approaches, the reinforcement learning (RL)-based subspace exploration strategy provides a novel objective optimization-directed perspective and promising performance. Nevertheless, even with improved performance, current reinforcement learning approaches face challenges similar to conventional methods when dealing with complex datasets. These challenges stem from the inefficient paradigm of using one agent per feature and the inherent complexities present in the datasets. This observation motivates us to investigate and address the above issue and propose a novel approach, namely HRLFS. Our methodology initially employs a Large Language Model (LLM)-based hybrid state extractor to capture each feature's mathematical and semantic characteristics. Based on this information, features are clustered, facilitating the construction of hierarchical agents for each cluster and sub-cluster. Extensive experiments demonstrate the efficiency, scalability, and robustness of our approach. Compared to contemporary or the one-feature-one-agent RL-based approaches, HRLFS improves the downstream ML performance with iterative feature subspace exploration while accelerating total run time by reducing the number of agents involved.

Figures

Figures reproduced from arXiv: 2504.17356 by the authors.

Figure 1
Figure 1. Comparison between HRLFS with other feature selection approaches. Among the various feature selection methodologies, reinforcement learning (RL)-based strategies [22] have received significant attention due to their ability to optimize feature subsets in an objective-directed manner with iterative feature subspace exploration. Despite these advances, RL-based feature selection methods encounter notable challenges wh… view at source ↗
Figure 2
Figure 2. Hybrid feature state extraction. log-likelihood function to maximize the likelihood between PDF to real distribution of 𝐹 , given as: 𝑙(Θ | 𝐹 ) = ∑︁ 𝑓 ∈𝐹 ln ∑︁ 𝑘 𝑖=1 𝑧𝑖𝑁 (𝑓 | 𝜇𝑖 , 𝜎𝑖) ! , (2) where Θ = {𝜃𝑖 } 𝑛 𝑖=1 denotes all the parameters for 𝑛 features. By that, we apply the expectation maximization (EM) algorithm to optimize the loss function and obtain the final parameter 𝜃 ∗ 𝑖 as the distribution state of the … view at source ↗
Figure 3
Figure 3. The prompt to generate feature description using dataset metadata. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Construction of agent hierarchy by incremental clustering. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Detail of the iteration and optimization with the hierarchical agents. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Comparison of HRLFS and baseline methods in downstream task performance and time consumption. HRLFS−ℎ , with time consumption comparable to traditional statistical methods (KBest, MCDM, and LASSONet) and much superior to methods based on RL (RLAS and MARLFS), deep lear…
Figure 7
Figure 7. Figure 7: The impact of different state representation methods. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: The impact of multi-agent hierarchical reinforcement learning architecture. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: The performance on downstream tasks and the proportion of active agents with changes in the granularity of the agent [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Comparison of different feature state extraction models. [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: The impact of different clustering algorithms. [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: The impact of different policy model. 5.7 Analysis of Reinforcement Learning Backend Selection This experiment aims to answer the question: Whether the policy model used in HRLFS is superior to other existing models? We compared ActorCritic [34] (Ours) with DQN [58], …
Figure 13
Figure 13. Figure 13: The impact of hyperparameter 𝜆 and 𝛼. In this section, we analysis two essential hyperparameter of HRLFS, i.e., 𝜆 (Equation 5) and 𝛼 (Equation 6). A higher 𝜆 encourages the selection of fewer features, while an increase 𝛼 prioritizes performance over feature compactne…
Figure 14
Figure 14. Figure 14: Empirical study of HRLFS’s hierarchical structure compared with the perfect binary tree. 6 Related Work Feature Selection. Feature selection methods are broadly categorized into three approaches: filter-based, wrapper￾based, and embedded-based methods [1]. Filter-base…

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. Knowledge-Guided Biomarker Identification for Label-Free Single-Cell RNA-Seq Data: A Reinforcement Learning Perspective

    q-bio.GN 2025-01 reject novelty 6.0 of 10

    RIGPS combines ensemble pre-filtering with multi-agent reinforcement learning to select compact gene panels, reporting improved clustering and annotation across 24 scRNA-seq datasets, though the final selection step m...

Reference graph

Works this paper leans on

81 extracted references · 70 canonical work pages · cited by 1 Pith paper

  1. [1]

    Feature selection: A data perspective,

    J. Li, K. Cheng, S. Wang, F. Morstatter, R. P. Trevino, J. Tang, and H. Liu, “Feature selection: A data perspective, ”ACM computing surveys (CSUR), vol. 50, no. 6, pp. 1–45, 2017

    2017

  2. [2]

    Towards data-centric ai: A comprehensive survey of traditional, reinforcement, and generative approaches for tabular data transformation,

    D. Wang, Y. Huang, W. Ying, H. Bai, N. Gong, X. Wang, S. Dong, T. Zhe, K. Liu, M. Xiaoet al., “Towards data-centric ai: A comprehensive survey of traditional, reinforcement, and generative approaches for tabular data transformation, ”arXiv preprint arXiv:2501.10555, 2025

    arXiv 2025

  3. [3]

    Recent advances in feature selection and its applications,

    Y. Li, T. Li, and H. Liu, “Recent advances in feature selection and its applications, ”Knowledge and Information Systems, vol. 53, pp. 551–577, 2017

    work page 2017

  4. [4]

    Feature selection for high-dimensional data—a pearson redundancy based filter,

    J. Biesiada and W. Duch, “Feature selection for high-dimensional data—a pearson redundancy based filter, ” inComputer recognition systems 2. Springer, 2008, pp. 242–249

    work page 2008

  5. [5]

    Feature selection for high-dimensional data: A fast correlation-based filter solution,

    L. Yu and H. Liu, “Feature selection for high-dimensional data: A fast correlation-based filter solution, ” inProceedings of the 20th international conference on machine learning (ICML-03), 2003, pp. 856–863

    work page 2003

  6. [6]

    A fast hybrid feature selection based on correlation-guided clustering and particle swarm optimization for high-dimensional data,

    X.-F. Song, Y. Zhang, D.-W. Gong, and X.-Z. Gao, “A fast hybrid feature selection based on correlation-guided clustering and particle swarm optimization for high-dimensional data, ”IEEE Transactions on Cybernetics, vol. 52, no. 9, pp. 9573–9586, 2021

    work page 2021

  7. [7]

    Decision tree classifier for network intrusion detection with ga-based feature selection,

    G. Stein, B. Chen, A. S. Wu, and K. A. Hua, “Decision tree classifier for network intrusion detection with ga-based feature selection, ” inProceedings of the 43rd annual Southeast regional conference-Volume 2, 2005, pp. 136–141

    work page 2005

  8. [8]

    Feature subset selection in large dimensionality domains,

    I. A. Gheyas and L. S. Smith, “Feature subset selection in large dimensionality domains, ”Pattern recognition, vol. 43, no. 1, pp. 5–13, 2010

    work page 2010

Show all 81 references
  1. [9]

    Consistent feature selection for analytic deep neural networks,

    V. C. Dinh and L. S. Ho, “Consistent feature selection for analytic deep neural networks, ”Advances in Neural Information Processing Systems, vol. 33, pp. 2420–2431, 2020

    2020

  2. [10]

    Deep feature selection: theory and application to identify enhancers and promoters,

    Y. Li, C.-Y. Chen, and W. W. Wasserman, “Deep feature selection: theory and application to identify enhancers and promoters, ”Journal of Computational Biology, vol. 23, no. 5, pp. 322–336, 2016

    2016

  3. [11]

    Lassonet: Neural networks with feature sparsity,

    I. Lemhadri, F. Ruan, and R. Tibshirani, “Lassonet: Neural networks with feature sparsity, ” inInternational conference on artificial intelligence and statistics. PMLR, 2021, pp. 10–18. Manuscript submitted to ACM 21

    2021

  4. [12]

    Scihorizon: Benchmarking ai-for-science readiness from scientific data to large language models,

    C. Qin, X. Chen, C. Wang, P. Wu, X. Chen, Y. Cheng, J. Zhao, M. Xiao, X. Dong, Q. Longet al., “Scihorizon: Benchmarking ai-for-science readiness from scientific data to large language models, ” inProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Min...

    2025

  5. [13]

    Gut microbiota and tuberculosis,

    Y. Liu, L. Yang, M. Meskini, A. Goel, M. Opperman, S. S. Shyamal, A. Manaithiya, M. Xiao, R. Ni, Y. Anet al., “Gut microbiota and tuberculosis, ” iMeta, p. e70054

  6. [14]

    Beyond discrete selection: Continuous embedding space optimization for generative feature selection,

    M. Xiao, D. Wang, M. Wu, P. Wang, Y. Zhou, and Y. Fu, “Beyond discrete selection: Continuous embedding space optimization for generative feature selection, ” in2023 IEEE International Conference on Data Mining (ICDM). IEEE, 2023, pp. 688–697

    2023

  7. [15]

    Knowledge-guided gene panel selection for label-free single-cell rna-seq data: A reinforcement learning perspective,

    M. Xiao, W. Zhang, X. Huang, H. Zhu, M. Wu, X. Li, and Y. Zhou, “Knowledge-guided gene panel selection for label-free single-cell rna-seq data: A reinforcement learning perspective, ”IEEE Transactions on Computational Biology and Bioinformatics, pp. 1–14, 2025

    2025

  8. [16]

    Revolutionizing biomarker discovery: Leveraging generative ai for bio-knowledge-embedded continuous space exploration,

    W. Ying, D. Wang, X. Hu, J. Qiu, J. Park, and Y. Fu, “Revolutionizing biomarker discovery: Leveraging generative ai for bio-knowledge-embedded continuous space exploration, ” inProceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024, ...

    2024

  9. [17]

    Advances, challenges and opportunities in creating data for trustworthy ai,

    W. Liang, G. A. Tadesse, D. Ho, L. Fei-Fei, M. Zaharia, C. Zhang, and J. Zou, “Advances, challenges and opportunities in creating data for trustworthy ai, ”Nature Machine Intelligence, vol. 4, no. 8, pp. 669–677, 2022

    2022

  10. [18]

    Knowledge hierarchy guided biological-medical dataset distillation for domain llm training,

    X. Cai, C. Wang, Q. Long, Y. Zhou, and M. Xiao, “Knowledge hierarchy guided biological-medical dataset distillation for domain llm training, ”arXiv preprint arXiv:2501.15108, 2025

    2025

  11. [19]

    Collaborative multi-agent reinforcement learning for automated feature transformation with graph-driven path optimization,

    X. Huang, D. Wang, Z. Ning, Z. Qiao, Q. Long, H. Zhu, Y. Du, M. Wu, Y. Zhou, and M. Xiao, “Collaborative multi-agent reinforcement learning for automated feature transformation with graph-driven path optimization, ”arXiv preprint arXiv:2504.17355, 2025

    2025 arXiv

  12. [20]

    m-kailin: Knowledge-driven agentic scientific corpus distillation framework for biomedical large language models training,

    M. Xiao, X. Cai, Q. Long, C. Wang, Y. Zhou, and H. Zhu, “m-kailin: Knowledge-driven agentic scientific corpus distillation framework for biomedical large language models training, ”arXiv preprint arXiv:2504.19565, 2025

    2025

  13. [21]

    Tabular data-centric ai: Challenges, techniques and future perspectives,

    Y. Fu, D. Wang, H. Xiong, and K. Liu, “Tabular data-centric ai: Challenges, techniques and future perspectives, ” inProceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024, pp. 5522–5525

    2024

  14. [22]

    Efficient reinforced feature selection via early stopping traverse strategy,

    K. Liu, P. Wang, D. Wang, W. Du, D. O. Wu, and Y. Fu, “Efficient reinforced feature selection via early stopping traverse strategy, ” in2021 IEEE International Conference on Data Mining (ICDM). IEEE, 2021, pp. 399–408

    2021

  15. [23]

    Automating feature subspace exploration via multi-agent reinforcement learning,

    K. Liu, Y. Fu, P. Wang, L. Wu, R. Bo, and X. Li, “Automating feature subspace exploration via multi-agent reinforcement learning, ” inProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, pp. 207–215

    2019

  16. [24]

    Zhang, Z

    K. Zhang, Z. Yang, and T. Başar,Multi-Agent Reinforcement Learning: A Selective Overview of Theories and Algorithms. Cham: Springer International Publishing, 2021, pp. 321–384. [Online]. Available: https://doi.org/10.1007/978-3-030-60990-0_12

    2021 doi

  17. [25]

    Autogfs: Automated group-based feature selection via interactive reinforcement learning,

    W. Fan, K. Liu, H. Liu, A. Hariri, D. Dou, and Y. Fu, “Autogfs: Automated group-based feature selection via interactive reinforcement learning, ” in Proceedings of the 2021 SIAM International Conference on Data Mining (SDM). SIAM, 2021, pp. 342–350

    2021

  18. [26]

    Autofs: Automated feature selection via diversity-aware interactive reinforcement learning,

    W. Fan, K. Liu, H. Liu, P. Wang, Y. Ge, and Y. Fu, “Autofs: Automated feature selection via diversity-aware interactive reinforcement learning, ” in 2020 IEEE International Conference on Data Mining (ICDM). IEEE, 2020, pp. 1008–1013

    2020

  19. [27]

    Exploring large language models for feature selection: A data-centric perspective,

    D. Li, Z. Tan, and H. Liu, “Exploring large language models for feature selection: A data-centric perspective, ”arXiv preprint arXiv:2408.12025, 2024

    2024 arXiv

  20. [28]

    Causal feature selection for responsible machine learning,

    R. Moraffah, P. Sheth, S. Vishnubhatla, and H. Liu, “Causal feature selection for responsible machine learning, ”arXiv preprint arXiv:2402.02696, 2024

    2024 arXiv

  21. [29]

    Self-organizing feature maps identify proteins critical to learning in a mouse model of down syndrome,

    C. Higuera, K. J. Gardiner, and K. J. Cios, “Self-organizing feature maps identify proteins critical to learning in a mouse model of down syndrome, ” PloS one, vol. 10, no. 6, p. e0129126, 2015

    2015

  22. [30]

    E. A. Feinberg and A. Shwartz,Handbook of Markov decision processes: methods and applications. Springer Science & Business Media, 2012, vol. 40

    2012

  23. [31]

    G. H. Dunteman,Principal components analysis. Sage, 1989, vol. 69

    1989

  24. [32]

    Hierarchical grouping to optimize an objective function,

    J. H. Ward Jr, “Hierarchical grouping to optimize an objective function, ”Journal of the American statistical association, vol. 58, no. 301, pp. 236–244, 1963

    1963

  25. [33]

    Prioritized experience replay,

    T. Schaul, “Prioritized experience replay, ”arXiv preprint arXiv:1511.05952, 2015

    2015 arXiv

  26. [34]

    Soft actor-critic algorithms and applications,

    T. Haarnoja, A. Zhou, K. Hartikainen, G. Tucker, S. Ha, J. Tan, V. Kumar, H. Zhu, A. Gupta, P. Abbeelet al., “Soft actor-critic algorithms and applications, ”arXiv preprint arXiv:1812.05905, 2018

    2018 arXiv

  27. [35]

    A natural policy gradient,

    S. M. Kakade, “A natural policy gradient, ”Advances in neural information processing systems, vol. 14, 2001

    2001

  28. [36]

    A performance-driven benchmark for feature selection in tabular deep learning,

    V. Cherepanova, R. Levin, G. Somepalli, J. Geiping, C. B. Bruss, A. G. Wilson, T. Goldstein, and M. Goldblum, “A performance-driven benchmark for feature selection in tabular deep learning, ”Advances in Neural Information Processing Systems, vol. 36, pp. 41 956–41 979, 2023

    2023

  29. [37]

    Gene expression omnibus: Ncbi gene expression and hybridization array data repository,

    R. Edgar, M. Domrachev, and A. E. Lash, “Gene expression omnibus: Ncbi gene expression and hybridization array data repository, ”Nucleic acids research, vol. 30, no. 1, pp. 207–210, 2002

    2002

  30. [38]

    Uci dataset download,

    Public, “Uci dataset download, ” [EB/OL], 2022, https://archive.ics.uci.edu/

    2022

  31. [39]

    Kaggle dataset download,

    J. Howard, “Kaggle dataset download, ” [EB/OL], 2022, https://www.kaggle.com/datasets

    2022

  32. [40]

    Openml dataset download,

    Public, “Openml dataset download, ” [EB/OL], 2022, https://www.openml.org

    2022

  33. [41]

    Libsvm dataset download,

    L. Chih-Jen, “Libsvm dataset download, ” [EB/OL], 2022, https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/

    2022

  34. [42]

    Group-wise reinforcement feature generation for optimal and explainable representation space reconstruction,

    D. Wang, Y. Fu, K. Liu, X. Li, and Y. Solihin, “Group-wise reinforcement feature generation for optimal and explainable representation space reconstruction, ” inProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 1826–1834

    2022

  35. [43]

    Traceable group-wise self-optimizing feature transformation learning: A dual optimization perspective,

    M. Xiao, D. Wang, M. Wu, K. Liu, H. Xiong, Y. Zhou, and Y. Fu, “Traceable group-wise self-optimizing feature transformation learning: A dual optimization perspective, ”ACM Transactions on Knowledge Discovery from Data, vol. 18, no. 4, pp. 1–22, 2024. Manuscript submitted to AC...

    2024

  36. [44]

    A comparative study on feature selection in text categorization,

    Y. Yang, J. O. Pedersenet al., “A comparative study on feature selection in text categorization, ” inicml, vol. 97. Citeseer, 1997, p. 35

    1997

  37. [45]

    Ensemble of feature selection algorithms: a multi-criteria decision-making approach,

    A. Hashemi, M. B. Dowlatshahi, and H. Nezamabadi-pour, “Ensemble of feature selection algorithms: a multi-criteria decision-making approach, ” International Journal of Machine Learning and Cybernetics, vol. 13, no. 1, pp. 49–69, 2022

    2022

  38. [46]

    Sequential attention for feature selection,

    T. Yasuda, M. Bateni, L. Chen, M. Fahrbach, G. Fu, and V. Mirrokni, “Sequential attention for feature selection, ” inThe Eleventh International Conference on Learning Representations, 2023. [Online]. Available: https://openreview.net/forum?id=TTLLGx3eet

    2023

  39. [47]

    Composite feature selection using deep ensembles,

    F. Imrie, A. Norcliffe, P. Liò, and M. van der Schaar, “Composite feature selection using deep ensembles, ”Advances in Neural Information Processing Systems, vol. 35, pp. 36 142–36 160, 2022

    2022

  40. [48]

    Reinforcement learning guided auto-select optimization algorithm for feature selection,

    H. Zhang, X. Yue, and X. Gao, “Reinforcement learning guided auto-select optimization algorithm for feature selection, ”Expert Systems with Applications, vol. 268, p. 126320, 2025

    2025

  41. [49]

    A finite-time analysis of two time-scale actor-critic methods,

    Y. F. Wu, W. Zhang, P. Xu, and Q. Gu, “A finite-time analysis of two time-scale actor-critic methods, ”Advances in Neural Information Processing Systems, vol. 33, pp. 17 617–17 628, 2020

    2020

  42. [50]

    New embedding models and api updates,

    OpenAI, “New embedding models and api updates, ” https://openai.com/index/new-embedding-models-and-api-updates/, accessed: 2025-04-17

    2025

  43. [51]

    Gpt-4 is openai’s most advanced system, producing safer and more useful responses,

    ——, “Gpt-4 is openai’s most advanced system, producing safer and more useful responses, ” https://openai.com/blog/gpt-4/, accessed: 2025-04-17

    2025

  44. [52]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding, ” inProceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologi...

    2019

  45. [53]

    Towards general text embeddings with multi-stage contrastive learning,

    Z. Li, X. Zhang, Y. Zhang, D. Long, P. Xie, and M. Zhang, “Towards general text embeddings with multi-stage contrastive learning, ”arXiv preprint arXiv:2308.03281, 2023

    2023 arXiv

  46. [54]

    Gemini embedding: Generalizable embeddings from gemini,

    J. Lee, F. Chen, S. Dua, D. Cer, M. Shanbhogue, I. Naim, G. H. Ábrego, Z. Li, K. Chen, H. S. Veraet al., “Gemini embedding: Generalizable embeddings from gemini, ”arXiv preprint arXiv:2503.07891, 2025

    2025 arXiv

  47. [55]

    Some methods for classification and analysis of multivariate observations,

    J. MacQueen, “Some methods for classification and analysis of multivariate observations, ” inProceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics, vol. 5. University of California press, 1967, pp. 281–298

    1967

  48. [56]

    A density-based algorithm for discovering clusters in large spatial databases with noise,

    M. Ester, H.-P. Kriegel, J. Sander, X. Xuet al., “A density-based algorithm for discovering clusters in large spatial databases with noise, ” inkdd, vol. 96, no. 34, 1996, pp. 226–231

    1996

  49. [57]

    A tutorial on spectral clustering,

    U. Von Luxburg, “A tutorial on spectral clustering, ”Statistics and computing, vol. 17, pp. 395–416, 2007

    2007

  50. [58]

    Human-level control through deep reinforcement learning,

    V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovskiet al., “Human-level control through deep reinforcement learning, ”nature, vol. 518, no. 7540, pp. 529–533, 2015

    2015

  51. [59]

    Deep reinforcement learning with double q-learning,

    H. Van Hasselt, A. Guez, and D. Silver, “Deep reinforcement learning with double q-learning, ” inProceedings of the AAAI conference on artificial intelligence, vol. 30, no. 1, 2016

    2016

  52. [60]

    Dueling network architectures for deep reinforcement learning,

    Z. Wang, T. Schaul, M. Hessel, H. Hasselt, M. Lanctot, and N. Freitas, “Dueling network architectures for deep reinforcement learning, ” inInternational conference on machine learning. PMLR, 2016, pp. 1995–2003

    2016

  53. [61]

    Avl trees with relaxed balance,

    K. S. Larsen, “Avl trees with relaxed balance, ” inProceedings of 8th International Parallel Processing Symposium. IEEE, 1994, pp. 888–893

    1994

  54. [62]

    Feature clustering based support vector machine recursive feature elimination for gene selection,

    X. Huang, L. Zhang, B. Wang, F. Li, and Z. Zhang, “Feature clustering based support vector machine recursive feature elimination for gene selection, ” Applied Intelligence, vol. 48, pp. 594–607, 2018

    2018

  55. [63]

    Kernel feature selection via conditional covariance minimization,

    J. Chen, M. Stern, M. J. Wainwright, and M. I. Jordan, “Kernel feature selection via conditional covariance minimization, ”Advances in neural information processing systems, vol. 30, 2017

    2017

  56. [64]

    Traceable automatic feature transformation via cascading actor-critic agents,

    M. Xiao, D. Wang, M. Wu, Z. Qiao, P. Wang, K. Liu, Y. Zhou, and Y. Fu, “Traceable automatic feature transformation via cascading actor-critic agents, ” inProceedings of the 2023 SIAM International Conference on Data Mining (SDM). SIAM, 2023, pp. 775–783

    2023

  57. [65]

    Deeppink: reproducible feature selection in deep neural networks,

    Y. Lu, Y. Fan, J. Lv, and W. Stafford Noble, “Deeppink: reproducible feature selection in deep neural networks, ”Advances in neural information processing systems, vol. 31, 2018

    2018

  58. [66]

    Effective nonlinear feature selection method based on hsic lasso and with variational inference,

    K. Koyama, K. Kiritoshi, T. Okawachi, and T. Izumitani, “Effective nonlinear feature selection method based on hsic lasso and with variational inference, ” inInternational Conference on Artificial Intelligence and Statistics. PMLR, 2022, pp. 10 407–10 421

    2022

  59. [67]

    Few-shot learning for feature selection with hilbert-schmidt independence criterion,

    A. Kumagai, T. Iwata, Y. Ida, and Y. Fujiwara, “Few-shot learning for feature selection with hilbert-schmidt independence criterion, ”Advances in Neural Information Processing Systems, vol. 35, pp. 9577–9590, 2022

    2022

  60. [68]

    Reinforcement learning: An introduction,

    R. S. Sutton, “Reinforcement learning: An introduction, ”A Bradford Book, 2018

    2018

  61. [69]

    A partially-supervised reinforcement learning framework for visual active search,

    A. Sarkar, N. Jacobs, and Y. Vorobeychik, “A partially-supervised reinforcement learning framework for visual active search, ”Advances in Neural Information Processing Systems, vol. 36, pp. 12 245–12 270, 2023

    2023

  62. [70]

    Gpt-4 technical report,

    O. et al., “Gpt-4 technical report, ” 2024. [Online]. Available: https://arxiv.org/abs/2303.08774

    2024 arXiv

  63. [71]

    Large language models are semi-parametric reinforcement learning agents,

    D. Zhang, L. Chen, S. Zhang, H. Xu, Z. Zhao, and K. Yu, “Large language models are semi-parametric reinforcement learning agents, ”Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  64. [72]

    Fastft: Accelerating reinforced feature transformation via advanced exploration strategies,

    T. He, X. Huang, Y. Du, Q. Long, Z. Qiao, M. Wu, Y. Fu, Y. Zhou, and M. Xiao, “Fastft: Accelerating reinforced feature transformation via advanced exploration strategies, ” in2025 IEEE 41st International Conference on Data Engineering (ICDE). IEEE Computer Society, 2025, pp. 4120–4133

    2025

  65. [73]

    Hierarchical reinforcement learning: A comprehensive survey,

    S. Pateria, B. Subagdja, A.-h. Tan, and C. Quek, “Hierarchical reinforcement learning: A comprehensive survey, ”ACM Computing Surveys (CSUR), vol. 54, no. 5, pp. 1–35, 2021

    2021

  66. [74]

    Hierarchical reinforcement learning,

    B. Hengst, “Hierarchical reinforcement learning, ”Encyclopedia of machine learning, pp. 495–502, 2011. Manuscript submitted to ACM 23

    2011

  67. [75]

    Probabilistic subgoal representations for hierarchical reinforcement learning,

    V. H. Wang, T. Wang, W. Yang, J. Kämäräinen, and J. Pajarinen, “Probabilistic subgoal representations for hierarchical reinforcement learning, ” in Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. OpenReview.net, 2024. [On...

    2024

  68. [76]

    Learning for decentralized control of multiagent systems in large, partially-observable stochastic environments,

    M. Liu, C. Amato, E. Anesta, J. Griffith, and J. How, “Learning for decentralized control of multiagent systems in large, partially-observable stochastic environments, ” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 30, 2016

    2016

  69. [77]

    Hierarchical multi-agent reinforcement learning,

    M. Ghavamzadeh, S. Mahadevan, and R. Makar, “Hierarchical multi-agent reinforcement learning, ”Autonomous Agents and Multi-Agent Systems, vol. 13, pp. 197–229, 2006

    2006

  70. [78]

    Multi-agent reinforcement learning with hierarchical coordination for emergency responder stationing,

    A. Sivagnanam, A. Pettet, H. Lee, A. Mukhopadhyay, A. Dubey, and A. Laszka, “Multi-agent reinforcement learning with hierarchical coordination for emergency responder stationing, ” inForty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-...

    2024

  71. [79]

    Rethinking decision transformer via hierarchical reinforcement learning,

    Y. Ma, J. Hao, H. Liang, and C. Xiao, “Rethinking decision transformer via hierarchical reinforcement learning, ” inForty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. OpenReview.net, 2024. [Online]. Available: https://openre...

    2024

  72. [80]

    Option-critic in cooperative multi-agent systems,

    J. Chakravorty, N. Ward, J. Roy, M. Chevalier-Boisvert, S. Basu, A. Lupu, and D. Precup, “Option-critic in cooperative multi-agent systems, ”arXiv preprint arXiv:1911.12825, 2019

    1911 arXiv

  73. [81]

    Hierarchical cooperative multi-agent reinforcement learning with skill discovery,

    J. Yang, I. Borovikov, and H. Zha, “Hierarchical cooperative multi-agent reinforcement learning with skill discovery, ” inProceedings of the 19th International Conference on Autonomous Agents and MultiAgent Systems, 2020, pp. 1566–1574. Manuscript submitted to ACM

    2020

Pith tools

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