Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Guided Random Forest and its application to data approximation

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

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

Pith's one-line read Guided Random Forest (GRAF) is an ensemble classifier that reuses random hyperplanes across the whole feature space; the paper claims it matches or beats random forest, gradient boosting, and AdaBoost on most of 115 benchmarks and doubles…

desk verdict A novel twist on oblique forests with a substantial empirical comparison, but the impurity function breaks the paper's own theoretical rationale and the key bound claim is not demonstrated. read the letter →

arxiv 1909.00659 v2 pith:CTQURFAB submitted 2019-09-02 cs.LG stat.ML

Prashant Gupta , Aashi Jindal , Jayadeva , Debarka Sengupta This is my paper · ORCID
classification cs.LGstat.ML
keywords guidedrandomforestensemblelearningobliquedecisiontreesglobalpartitioningboostingdataapproximationsensitivityscoregeneralizationerror
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes Guided Random Forest (GRAF), an ensemble classifier that builds each tree by repeatedly drawing random hyperplanes and reusing every hyperplane beyond the partition it was drawn for, so that split decisions are global rather than local. The authors aim to establish that this global partitioning is what tree ensembles lack relative to boosting, and that it lowers an upper bound on generalization error. They report experiments on 115 benchmark datasets in which GRAF is better than or comparable to random forest, gradient boosting, and AdaBoost on most datasets. GRAF also assigns each training point a sensitivity score, and the authors argue that keeping only the highest-sensitivity points approximates the full dataset well enough for downstream tasks. If the claims hold, the paper supplies a single method that is both a competitive ensemble classifier and a data-approximation tool.

What carries the argument

The mechanism is a guided partitioning loop. The impurity function $Z(\Omega_p)$ in equation (10) scores every current partition using global class counts; GRAF chooses the partition $\omega = \arg\max_{\Omega_i} Z(\Omega_i)$, draws a random hyperplane through the centroid of that partition, and then extends the same hyperplane to all other impure partitions. Samples accumulate a variable-length binary code as hyperplanes are added, and a leaf's code maps to the class posterior, with inverse-frequency correction for class imbalance. The sensitivity of a point is the number of hyperplanes needed to put its partition into a pure state, divided by the point's rank inside the partition, and then normalized class-wise; this is what turns the forest into a data-approximation tool.

What would settle it

Take the paper's $Z$ from equation (10) on a two-class problem with global counts $N_1=100$, $N_2=10000$, and a partition containing 100 class-1 samples and 1 class-2 sample. The parent $Z$ is approximately 0.0202, while splitting off 50 class-1 samples gives children whose $Z$ values sum to approximately 0.0204, violating the required inequality; any such node falsifies the claim that guided splitting increases overall purity and removes the route to the stated generalization-error bound.

Watch

Extended reading notes

Core claim

The central claim is that reusing a split plane across all impure partitions—rather than letting each tree region learn its own local splits—changes an ensemble of trees from a bagging-style average into something closer to boosting, and that this reduces an upper bound on generalization error. At each step GRAF selects the most impure partition $\omega$ under the impurity measure $Z$, draws a random hyperplane through it, and applies the same hyperplane to every other impure partition; leaves are described by the variable-length bit sequence each sample accumulates, and the tree returns class posteriors per leaf. On the paper's 115-dataset benchmark, GRAF is better than or comparable to AdaBoost on 94 datasets, random forest on 69, and gradient boosting on 67, including the high-class-count datasets. The paper also claims that the number of hyperplanes needed to purify a region defines a sensitivity score, and that subsampling the highest-sensitivity points can approximate the full data.

Load-bearing premise

The load-bearing premise is that splitting the most impure partition always increases overall purity, which requires the impurity measure $Z$ to satisfy $Z(\Omega_p) \ge Z(\Omega_{p0}) + Z(\Omega_{p1})$; the paper's equation (10) does not guarantee this, so the greedy motivation and the bound argument depend on a property the chosen measure can violate.

Editorial extensions

If this is right

  • If global partitioning lowers the generalization error bound, then random-forest-style ensembles can be improved by reusing split geometry across regions instead of searching for locally optimal splits in each leaf.
  • Because GRAF trees output class-posterior estimates rather than raw votes, the ensemble can be combined additively in log space, the same functional form as boosting.
  • The sensitivity score gives each training point an importance value, so users can subsample by importance rather than uniformly and retain most of the accuracy, as the paper demonstrates on its simulated and benchmark datasets.
  • GRAF's treatment of each tree as a high-variance instance with weight updates that focus on the most impure partition positions the method between bagging and boosting, giving a single algorithm for both regimes.

Reading between the lines

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

  • The bound-reduction argument rests on the subadditivity of $Z$; if that property is not restored, GRAF's empirical success would still stand, but as a heuristic rather than as a consequence of the stated bound.
  • Because high-sensitivity points overlap with support-vector locations in the paper's experiments, sensitivity scores could plausibly serve as an active-learning acquisition rule, selecting points to label next, although the paper does not test this use.
  • The paper's own closing remark concedes that the sensitivity scheme does not provide a selection criterion for how many points to keep, which leaves the data-approximation claim underspecified at the point where the user must choose a subset size.
  • A direct extension would be to define sensitivity from each sample's marginal impurity reduction rather than from leaf size, which would make the approximation claim less sensitive to how the tree is grown.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Guided Random Forest (GRAF), an ensemble classifier in which oblique hyperplane splits are generated greedily for the most impure partition and then extended to other impure partitions, so that the resulting tree uses variable-length partition codes. The authors claim that this global partitioning reduces the generalization error bound, that GRAF outperforms random forest, gradient boosting, and AdaBoost, and that a sample sensitivity score derived from the number of hyperplanes needed to purify a partition can be used for data approximation. The manuscript reports bias/variance experiments on synthetic datasets, strength/correlation curves, a comparison on 115 UCI datasets, and data-approximation experiments on synthetic and UCI data.

Significance. The paper proposes a genuinely different ensemble construction: one oblique hyperplane is applied across multiple impure partitions rather than only to the local partition for which it was generated. The 115-dataset comparison is a substantial empirical effort, and the sensitivity-based data approximation idea is potentially useful. However, the central theoretical motivation is internally inconsistent, the claimed generalization-error-bound reduction is not actually evaluated, and the sensitivity scores depend on an arbitrary ordering. No code, seeds, or error bars are provided, which makes the empirical claims difficult to verify. If the theoretical issues are repaired and the empirical claims are substantiated, GRAF could be a competitive ensemble method with an interesting data-subsetting capability.

major comments (4)
  1. [Section 2, Eq. (10)] The impurity function Z is introduced with the requirement that if Omega_p = Omega_p0 union Omega_p1, then Z(Omega_p) >= Z(Omega_p0) + Z(Omega_p1), and the greedy partition selection in Eqs. (11)-(15) is motivated by an increase in overall purity. Equation (10) does not satisfy this property. For N1=100 and N2=10000, a partition with counts (100,1) has Z = 101*(1 - 1.00000001/1.00020001) = 0.020198, while splitting it into (50,0) and (50,1) gives Z values 0 and 51*(1 - 0.25000001/0.25010001) = 0.020396, respectively, so the children sum to 0.020396 > 0.020198. Thus splitting can increase total impurity, and the subadditivity premise on which the 'global partitioning reduces the generalization error bound' claim rests is false. A corrected impurity definition or a revised theoretical argument is needed.
  2. [Section 5.2, Eq. (30) and Fig. 6] The abstract states that global partitioning reduces the generalization error bound, but the paper never evaluates the bound PE* from Eq. (30) and never compares it with random forest or gradient boosting. Fig. 6 only plots strength and correlation for GRAF on synthetic datasets. A direct computation of the bound, at least on the same datasets, is required to support the central claim that the bound is reduced.
  3. [Section 6, Eq. (37)] Sensitivity scores are computed as theta_{omega x_i} = W(omega)/i after ranking points in a partition arbitrarily. Therefore the scores depend on an arbitrary permutation and are not a well-defined property of the data. The support-vector analogy and the data-approximation experiments in Figs. 7-9 are consequently not reproducible as specified. Please replace the arbitrary rank with an order-invariant definition or provide a justification for the chosen ordering.
  4. [Section 5.3] The claim that GRAF outperforms state-of-the-art bagging and boosting algorithms rests on an accuracy comparison over 115 UCI datasets, but no standard deviations, confidence intervals, statistical significance tests, seeds, or code are reported, and 'comparable' is not defined. The appendix table (Table A1) is also not included in the manuscript, so the result cannot be checked. Reporting these details is necessary before the empirical superiority conclusion can be accepted.
minor comments (5)
  1. [Section 2, Eqs. (13)-(14) and Section 4] The definition splits every impure partition phi by W_omega, but Section 4 states that bit assignment is skipped for partitions not dichotomized at a given height; please reconcile these two descriptions.
  2. [Section 3, Algorithm 1] The line 'Choose a random hypothesis using P(i)' is not sufficient to reproduce the hyperplane generation; Eq. (4) draws weights uniformly from feature ranges, not conditioned on P(i) as described.
  3. [Section 5.1] The synthetic datasets are described only as generated using Weka with the RandomRBF class; the specific parameters and the number of replicates should be listed for reproducibility.
  4. [Abstract and Section 7] The claim 'outperforms state of the art ... algorithms' is stronger than the result reported in Section 5.3, which says GRAF is better or comparable on a majority of datasets; please align the wording.
  5. [Section 6, Fig. 8] The method for sampling 'according to their sensitivities (P)' is not explained; it is unclear whether the scores are normalized and used as a probability distribution or converted to ranks.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: GRAF's empirical claims are tested against external baselines and its generalization-bound discussion uses Breiman's standard theorem.

full rationale

I examined the claimed derivation chain. The method is constructed from random hyperplanes and an impurity function; no parameter is fitted to benchmark test labels and then reported as a prediction. The generalization-error-bound discussion cites Breiman's standard bound PE* <= rho(1-s^2)/s^2 (Eq. 30), an external theorem independent of GRAF's own construction. The strength/correlation measurements are empirical summaries, not the bound itself, and the paper does not claim to have derived the bound from its own outputs. The comparison on 115 UCI datasets is against external algorithms (Random Forest, Gradient Boosting, AdaBoost) under Fernandez-Delgado's protocol, so the headline 'comparable or better' claim is externally falsifiable. The boosting analogy in Section 3 is an interpretation of the algorithm's sample-weight updates, not an input that is renamed as an output. The sensitivity/support-vector analogy in Section 6 is an empirical overlap and accuracy comparison, not a definitional equivalence. The one substantive weakness is mathematical rather than circular: Eq. (10) does not satisfy the subadditivity condition asserted for Z in Section 2, so the 'overall purity increases' rationale is unsupported; this is a correctness gap, not a case where a prediction equals its input by construction. No load-bearing self-citation chain appears; the references to Breiman, Freund-Schapire, Friedman, and Fernandez-Delgado et al. are external. Therefore no circular step is found.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The paper's central claims rest on a small set of design choices: a particular impurity measure, the global reuse of hyperplanes, and a ranking-based sensitivity definition. The impurity measure is presented as if it were subadditive, but it is not; this is the most serious gap. The sensitivity ranking depends on arbitrary order within partitions. Hyperparameters L and M are tuned per dataset, which is standard but means the reported accuracies are post-tuning.

free parameters (4)
  • Number of estimators L = tuned from {100,200,500,1000,2000}
    Tuned by 5-fold cross-validation on training data; not a free parameter in a derivation.
  • Feature subspace size M = tuned from {log2(d), sqrt(d), d/2, d}
    Tuned per dataset via cross-validation.
  • Minimum samples per split = 2 to 5
    Tuned via cross-validation; part of the algorithm's hyperparameter set.
  • Sensitivity ranking divisor = i (arbitrary rank)
    Eq. (37) divides partition importance by the sample's arbitrary rank within the partition; this choice makes sensitivity scores order-dependent and is not a fitted constant but an ad hoc modeling choice.
assumptions (5)
  • ad hoc to paper The impurity function Z is subadditive: Z(Omega) >= Z(Omega0) + Z(Omega1) for any split.
    Stated in Sec. 2 above Eq. (10) as a property of Z, but Eq. (10) does not satisfy it in general; counterexample with imbalanced class sizes. Greedy selection of the most impure partition depends on this.
  • ad hoc to paper Applying a hyperplane generated for one partition to other impure partitions improves overall purity.
    The method extends W_omega to all impure partitions in Eq. (14) without proving that each extension helps; in implementation they skip non-dichotomizing splits, which changes the stated behavior.
  • ad hoc to paper The arbitrary ranking of samples within a partition does not affect the usefulness of sensitivity scores.
    Eq. (37) divides by rank i, so the sensitivity of a sample depends on its arbitrary position in the partition; this is a modeling assumption that the data approximation claim relies on.
  • standard math Breiman's strength-correlation generalization bound applies to GRAF.
    Used as external benchmark in Eq. (30); standard result, not derived here.
  • standard math Random hyperplanes with bias through the partition mean dichotomize any partition almost surely.
    Continuous weight sampling implies zero probability of a point lying exactly on the hyperplane; standard.
invented entities (1)
  • Sensitivity score s_i independent evidence
    purpose: A per-sample importance score used for data approximation and support-vector analogy.
    The score is defined by the algorithm (Eqs. 37-40) and compared with support vectors on multiple datasets, giving a falsifiable handle. However, its definition depends on arbitrary ranking, so the independent evidence is weak.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Guided Random Forest and its application to data approximation." pith.science (2026). https://pith.science/paper/CTQURFAB

@misc{pith2026190900659,
  author       = {Pith},
  title        = {Pith review of: Guided Random Forest and its application to data approximation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CTQURFAB}},
  note         = {Machine review of arXiv:1909.00659}
}
read the original abstract

We present a new way of constructing an ensemble classifier, named the Guided Random Forest (GRAF) in the sequel. GRAF extends the idea of building oblique decision trees with localized partitioning to obtain a global partitioning. We show that global partitioning bridges the gap between decision trees and boosting algorithms. We empirically demonstrate that global partitioning reduces the generalization error bound. Results on 115 benchmark datasets show that GRAF yields comparable or better results on a majority of datasets. We also present a new way of approximating the datasets in the framework of random forests.

Figures

Figures reproduced from arXiv: 1909.00659 by the authors.

Figure 1
Figure 1. An overview of the creation of high variance instances in GRAF. Every instance consists of sub-spacing the dataset in a uniformly sampled [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Partitioning of space in GRAF is represented by a tree. A partition, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 5
Figure 5. Bias-variance analysis with an increasing samples in a training [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figures from the paper (6 more)
Figure 3
Figure 3. Figure 3: Bias-variance analysis with an increasing number of estimators [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: Bias-variance analysis with an increasing number of dimensions [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Strength and correlation analysis with increasing number of [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: Performance evaluation of Random forest [3] and GRAF, with [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 7
Figure 7. Figure 7: illustrates that when only 25% of the total points are sampled, samples with the highest sensitivities adequately approximate the regions with the highest confusion [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: An analogy between support vectors and points with high sen [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]

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. Learning ON Large Datasets Using Bit-String Trees

    cs.LG 2025-08 unverdicted novelty 4.0 of 10

    The metadata announces hashing and cancer-genomics methods, but the full text is a different neural SDE paper, making the listed paper's claims unverifiable.

Reference graph

Works this paper leans on

20 extracted references · 19 canonical work pages · cited by 1 Pith paper

  1. [1]

    Ensemble methods in machine learning,

    T. G. Dietterich, “Ensemble methods in machine learning,” in International workshop on multiple classifier systems. Springer, 2000, pp. 1–15

    work page 2000

  2. [2]

    Do we need hundreds of classifiers to solve real world classifica- tion problems?

    M. Fern ´andez-Delgado, E. Cernadas, S. Barro, and D. Amorim, “Do we need hundreds of classifiers to solve real world classifica- tion problems?” The Journal of Machine Learning Research , vol. 15, no. 1, pp. 3133–3181, 2014

    work page 2014

  3. [3]

    Random forests,

    L. Breiman, “Random forests,” Machine learning, vol. 45, no. 1, pp. 5–32, 2001

    2001

  4. [4]

    Random rotation ensembles,

    R. Blaser and P . Fryzlewicz, “Random rotation ensembles,” The Journal of Machine Learning Research , vol. 17, no. 1, pp. 126–151, 2016

    work page 2016

  5. [5]

    On oblique random forests,

    B. H. Menze, B. M. Kelm, D. N. Splitthoff, U. Koethe, and F. A. Hamprecht, “On oblique random forests,” in Joint European Con- ference on Machine Learning and Knowledge Discovery in Databases . Springer, 2011, pp. 453–469

    work page 2011

  6. [6]

    Nonlinear boosting projections for ensemble construction,

    N. Garc ˜Aa-Pedrajas, C. Garc ˜Aa-Osorio, and C. Fyfe, “Nonlinear boosting projections for ensemble construction,” Journal of Machine Learning Research, vol. 8, no. Jan, pp. 1–33, 2007

    work page 2007

  7. [7]

    Rotation forest: A new classifier ensemble method,

    J. J. Rodriguez, L. I. Kuncheva, and C. J. Alonso, “Rotation forest: A new classifier ensemble method,” IEEE transactions on pattern analysis and machine intelligence, vol. 28, no. 10, pp. 1619–1630, 2006

    work page 2006

  8. [8]

    An experimental study on rotation forest ensembles,

    L. I. Kuncheva and J. J. Rodr ´ıguez, “An experimental study on rotation forest ensembles,” in International workshop on multiple classifier systems. Springer, 2007, pp. 459–468

    work page 2007

Show all 20 references
  1. [9]

    An empirical evaluation of rotation-based ensemble classifiers for customer churn predic- tion,

    K. W. De Bock and D. Van den Poel, “An empirical evaluation of rotation-based ensemble classifiers for customer churn predic- tion,” Expert Systems with Applications , vol. 38, no. 10, pp. 12 293– 12 301, 2011

    2011

  2. [10]

    Greedy function approximation: a gradient boost- ing machine,

    J. H. Friedman, “Greedy function approximation: a gradient boost- ing machine,” Annals of statistics, pp. 1189–1232, 2001

    2001

  3. [11]

    Boosting the margin: A new explanation for the effectiveness of voting methods,

    R. E. Schapire, Y. Freund, P . Bartlett, W. S. Lee et al. , “Boosting the margin: A new explanation for the effectiveness of voting methods,” The annals of statistics , vol. 26, no. 5, pp. 1651–1686, 1998

    1998

  4. [12]

    A decision-theoretic generalization of on-line learning and an application to boosting,

    Y. Freund and R. E. Schapire, “A decision-theoretic generalization of on-line learning and an application to boosting,” Journal of computer and system sciences, vol. 55, no. 1, pp. 119–139, 1997

    1997

  5. [13]

    Bias, variance, and arcing classifiers,

    L. Breiman, “Bias, variance, and arcing classifiers,” 1996

    1996

  6. [14]

    Bias plus variance decomposition for zero-one loss functions,

    R. Kohavi, D. H. Wolpert et al., “Bias plus variance decomposition for zero-one loss functions,” in ICML, vol. 96, 1996, pp. 275–83

    1996

  7. [15]

    A unified bias-variance decomposition,

    P . Domingos, “A unified bias-variance decomposition,” in Proceed- ings of 17th International Conference on Machine Learning , 2000, pp. 231–238

    2000

  8. [16]

    Variance and bias for general loss functions,

    G. M. James, “Variance and bias for general loss functions,” Machine Learning, vol. 51, no. 2, pp. 115–135, 2003

    2003

  9. [17]

    Weka: Practical machine learning tools and techniques with java implementations,

    R. Dimov, M. Feld, D. M. Kipp, D. A. Ndiaye, and D. D. Heck- mann, “Weka: Practical machine learning tools and techniques with java implementations,” AI Tools SeminarUniversity of Saarland, WS, vol. 6, no. 07, 2007

    2007

  10. [18]

    The random subspace method for constructing decision forests,

    I. Barandiaran, “The random subspace method for constructing decision forests,” IEEE transactions on pattern analysis and machine intelligence, vol. 20, no. 8, 1998

    1998

  11. [19]

    Shape quantization and recognition with randomized trees,

    Y. Amit and D. Geman, “Shape quantization and recognition with randomized trees,” Neural computation, vol. 9, no. 7, pp. 1545–1588, 1997

    1997

  12. [20]

    UCI machine learning repository,

    D. Dheeru and E. Karra Taniskidou, “UCI machine learning repository,” 2017. [Online]. Available: http://archive.ics.uci.edu/ ml

    2017

Pith tools

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