Pith. sign in

REVIEW 5 major objections 6 minor 60 references

GPAE builds an autoencoder from Gaussian processes with random Fourier features and uses a latent density estimator during counterfactual search, reporting balanced, in-distribution counterfactuals on five tabular datasets.

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

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

A Gaussian-process autoencoder with a latent-space density estimator generates counterfactual examples for tabular data, with competitive or better scores on several evaluation metrics.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A promising but not-yet-trustworthy counterfactual method: the architecture and experiments are real, but the validity claim rests on a broken constraint, and the missing code/baselines make it hard to verify. the 5 major comments →

arxiv 2509.00884 v1 pith:3FK3NZ6O submitted 2025-08-31 cs.LG cs.AI

An Explainable Gaussian Process Auto-encoder for Tabular Data

Wei Zhang , Brian Barr , John Paisley This is my paper
classification cs.LG cs.AI
keywords Gaussian processrandom Fourier featuresautoencodercounterfactual explanationsalgorithmic recoursedensity estimationtabular datainterpretable machine learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper proposes an autoencoder built from Gaussian processes, approximated with random Fourier features, and a latent linear classifier that supervises the encoder. On top of this, it adds a learned density estimator in the latent space and uses it during counterfactual search: the search minimizes the change to the input while staying on the classifier's decision boundary and inside high-density regions of the target class. The paper reports that this setup, called GPAE, produces counterfactual samples that are simultaneously close to the query, diverse, stable, and in-distribution across five tabular datasets, while needing far fewer trainable parameters than neural-autoencoder baselines. A beta-selection algorithm chooses the strength of the density term by matching the counterfactual distribution to the data projected on the boundary. The practical payoff is an explainable-AI recipe for recourse: what to change in an application to flip an outcome, with immutable features fixed by a mask.

Core claim

GPAE replaces the usual neural encoder and decoder with Gaussian processes whose RBF kernel is approximated by random Fourier features. The encoder becomes a fixed random cosine layer followed by one trainable linear projection W_e, and the decoder the same with W_d; only the two projections and the latent classifier's boundary theta_c are learned. The paper's claim is that this parameter-light supervised autoencoder, searched with a density regularizer, generates counterfactuals that are valid, diverse, and in-distribution. The counterfactual for a query x is x+delta, where delta is found by minimizing ||delta||^2/2 - beta log p(f_e(x+delta)) subject to theta_c^T f_e(x+delta)=0; a mask mult

What carries the argument

The carrier of the argument is the random-Fourier-feature Gaussian process autoencoder coupled to a latent density estimator. The RFF map phi(x)=sqrt(2/S)[cos(z_s^T x + c_s)] is a fixed, untrained cosine layer; learning happens only in the linear projections W_e and W_d and the latent classifier weight theta_c. The density estimator p(lambda) proportional to exp(w^T phi(lambda)) N(lambda|mu,sigma) approximates the latent data density with an exponentiated RFF function under a Gaussian envelope, and its gradient steers the counterfactual search toward high-density target regions. The decision-boundary constraint theta_c^T f_e(x+delta)=0, plus the projection methodology for choosing beta, is w

Load-bearing premise

The search assumes that a latent code exactly on the decision boundary flips the classifier's predicted label to the target class; with a sigmoid output, that boundary is only the 50% probability contour, so label flipping is not logically entailed by the constraint.

What would settle it

On a tabular dataset, train GPAE and, for a held-out query, record counterfactuals at the converged solution of the constrained search, then evaluate the explained classifier's argmax label at those counterfactuals. If a substantial fraction are not assigned the target class, then the boundary constraint theta_c^T f_e(x+delta)=0 does not by itself make the sample valid, and the reported ~0.99 validity must be attributed to the density-regularized optimization rather than the equality constraint.

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

If this is right

  • Counterfactuals are generated directly in feature space, so immutable or non-actionable features can be kept fixed by masking gradients, without post-hoc replacement that creates out-of-distribution samples.
  • Because only two linear projections and a boundary vector are trained, the explainer is much lighter than deep autoencoder baselines and less prone to overfitting on small tabular sets.
  • The density regularizer and beta-selection procedure make the counterfactual distribution track the target class's latent distribution, which is what yields the reported low IM1/IM2 and high validity.
  • The same encoder doubles as a competitive classifier, so the model being explained is not an arbitrary black box; the latent space it supervises is the one being searched.
  • The masking mechanism extends the same search to fair-recourse settings where protected attributes must not be altered.

Where Pith is reading between the lines

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

  • The beta-selection scheme, which projects latent codes onto the decision boundary and minimizes KL divergence between the projected data and the density estimate, could be reused by any latent-space counterfactual method that has a linear classifier, not just GPAE.
  • Because the density estimator is learned once in latent space, it could also serve as an out-of-distribution detector, flagging query points whose latent codes fall in low-density regions before an explanation is attempted.
  • The same architecture could incorporate non-RBF kernels through alternative random feature maps, making the density estimator and counterfactual search applicable to heterogeneous feature scales without architectural changes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes GPAE, a Gaussian-process autoencoder with random Fourier feature (RFF) approximations for generating counterfactual explanations on tabular data. The encoder/decoder are linear maps on RFF features; the latent space is supervised by a sigmoid classifier; and a learned density estimator is used in a constrained search over input perturbations to find counterfactuals that are close, diverse, and in-distribution. A separate procedure selects the density regularization strength by minimizing a KL divergence between projected data and the estimated density on the decision boundary. Experiments on five tabular datasets compare GPAE with logistic regression, Wachter, VAE-based methods, CounterNet, SAE, and a GP-inducing-point variant, reporting classification accuracy and counterfactual metrics including L2, diversity, instability, discriminative power, IM1/IM2, and validity. The paper claims that GPAE achieves a balanced performance with fewer learnable parameters and generates diversified, in-distribution counterfactuals.

Significance. If the validity mechanism were fully established, the paper would make a useful contribution: the RFF-based GP autoencoder is a lightweight alternative to neural autoencoders, the density-guided search is a principled way to encourage in-distribution counterfactuals, and the mask mechanism for immutable features is practically relevant. The empirical study is broad, uses standard benchmarks, and evaluates against several external baselines; IM1/IM2 are computed with independently trained autoencoders, which partially mitigates self-referentiality of the in-distribution claim. However, the central derivation connecting the optimization in §4.5 to the main claim of generating valid counterfactuals is incomplete: the stated equality constraint places counterfactuals at the classifier's decision boundary, which does not, in general, flip the predicted label. The reported validity values near 0.99 are therefore unexplained by the stated optimization. This gap, together with a factor-2 error in the reported gradient and a class-conditional gap in the density estimator, prevents the paper from being acceptable in its current form. The contributions are potentially salvageable with a corre

major comments (5)
  1. [§4.5, Eqs. (18)-(19) and Table 5] The claim that the boundary constraint 'naturally makes the counterfactual sample valid' is not justified. With the sigmoid classifier in Algorithm 2 (ŷ_i = sigmoid(λ_i^T θ_c)), the constraint θ_c^T f_e(x+δ)=0 gives p(target)=0.5. Under the standard threshold rule this is a tie, not a label flip from the original side; for queries whose original class is the positive class, the boundary point is still classified as the original class. Unlike the SAE baseline in §4.1, which explicitly uses step=2 to move beyond the boundary, Eq. (17) contains no term that pushes the logit past zero, and Eq. (18) is an equality, not an inequality. The reported Validity≈0.99 in Table 5 (e.g., GMC and FICO rows) is therefore not a consequence of the stated optimization. The authors should replace Eq. (18) with a strict inequality or margin constraint (e.g., θ_c^T f_e(x+δ) ≥ ε or ≤ −ε for the target side), ad
  2. [§4.5, Eq. (22) vs. Eq. (20)] The gradient in Eq. (22) has a factor-2 error in the Gaussian term of the density estimator. Differentiating the log-density term in Eq. (20), −β[w^T φ(f_e) − ½ f_e^T Σ^{-1} f_e + μ^T Σ^{-1} f_e], with respect to f_e gives −β[(∂φ)^T w − Σ^{-1} f_e + Σ^{-1} μ]. The coefficient of the μ^T Σ^{-1} term in Eq. (22) should therefore be 1, not 2. This changes the balance between the distance and density forces in the counterfactual search. The authors should correct Eq. (22) and confirm which gradient was actually implemented in the experiments.
  3. [§4.4-§4.5 and Highlights] The density estimator in Eqs. (12)-(14) is learned from all N observed latent codes without conditioning on class, and the search objective (17) uses this global density. The third Highlight states that the density estimator 'direct[s] the generated samples to be in-distribution of the target class,' but the construction does not target the target class specifically; it can equally favor the original class density near the boundary. The IM1/IM2 results suggest empirically that the generated counterfactuals lie near the target manifold, but this is not entailed by the stated density objective. The authors should either train a class-conditional density estimator, restrict the density model to target-class samples, or provide a clear argument for why the global density suffices.
  4. [§4.6, Eqs. (28)-(31)] The β-selection criterion minimizes KL(q(γ) || p(γ)), where p(γ) is the same parametric density estimator used in the search objective. This makes the 'optimal β' claim partly self-referential: the criterion can only measure how well the counterfactual distribution matches the model's own density estimate, not how well that estimate matches the true data density. The external IM1/IM2 metrics mitigate the impact, but the text should be revised to state that β is selected by internal density matching, and ideally the sensitivity of the final counterfactual metrics to β should be reported.
  5. [§4.5, Eq. (22)-(25)] The optimization algorithm is under-specified. The paper states that δ and η are optimized 'iteratively and interchangeably,' but does not give the update rule for η, the stopping criterion, or whether the equality constraint is enforced exactly at convergence. Given that the validity claim depends entirely on the final point satisfying the boundary condition, the manuscript should specify the primal-dual updates, the convergence tolerance, and how the final counterfactual is projected/rounded. This is needed to reproduce the reported validity numbers.
minor comments (6)
  1. [Throughout] There are numerous typos and formatting errors: 'Fgiure 4', 'W ac.', 'GP AE', 'Disucssion', 'Futhermore', 'high-stack fields', 'SA W', 'We All experiments', and 'the IM2 and IM2' in the caption of Figure 6. A careful proofread is needed.
  2. [Eq. (12)] The density estimator notation is inconsistent: Eq. (12) uses a scalar σ, while Eq. (20) uses a covariance matrix Σ. Please unify the notation and define the Gaussian envelope precisely.
  3. [§5.4, Table 5] The validity metric in Eq. (35) uses the explained classifier's predicted label. Since the paper's own classifier is a sigmoid, the tie-breaking convention at p=0.5 should be stated; otherwise the validity numbers are ambiguous.
  4. [Algorithm 2] Step 8 writes ŷ_i = sigmoid(λ_i^T θ_c), while Eq. (10) defines the logit as λ_i^T θ_c and Eqs. (19)-(24) also include a bias θ0. Please clarify whether the classifier has a bias term and, if so, include it consistently in the constraint and updates.
  5. [§5.2] The description of baselines is brief. In particular, the text says CARLA default parameters are used for all methods, but it is not stated whether the explained classifier is the same for all baselines and whether the same immutable-feature masking protocol is applied to methods that do not natively support it. This information is important for the fairness of the comparison.
  6. [§5.6] The ablation text says 'all the other metrics improve significantly' while Figure 6 shows Diversity decreasing; the wording should be adjusted to acknowledge the trade-off.

Circularity Check

1 steps flagged

β-selection makes the in-distribution claim partly self-referential; the core GPAE derivation and external metrics remain independent.

specific steps
  1. fitted input called prediction [Section 4.6, Eqs. (28)-(30), Figure 4]
    "we search β by measuring the KL divergence between the approximate distribution and true distribution of the latent vectors on the decision boundary... The empirical distribution over N data samples on the decision boundary can be calculated as: q(γ) = (1/N) Σ N(γ|A^T λ_c^n, Σ)... The KL is minimized around β = 0.4."

    The counterfactual objective in Eq. (17) maximizes β log p(f_e(x+δ)), where p is the learned density estimator. Section 4.6 then selects β by minimizing KL(q(γ)||p(γ)): p(γ) in Eq. (28) is the same learned density estimator, and q(γ) in Eq. (29) is formed from the model's own projected latent/counterfactual codes λ_c^n. Thus β is chosen to make the counterfactual distribution match the model's own density estimate by construction. The claim that GPAE generates 'in-distribution' counterfactuals is therefore partly self-referential relative to p, rather than being established independently. This is mitigated by the external IM1/IM2 metrics, which use separately trained autoencoders, so the circularity is partial rather than total.

full rationale

The core GPAE construction is self-contained: the RFF-based encoder/decoder, the supervised latent-space classifier, and the learned density estimator are all fitted to training data, and the counterfactual search is a constrained optimization in feature space. The central counterfactual-quality claims are evaluated against external baselines and external metrics (IM1/IM2 with independently trained autoencoders), so the main derivation is not circular. The one partially circular element is the β-selection procedure in Section 4.6: the regularization weight is chosen by minimizing a KL divergence between the model's own density estimate and the model's own boundary-projected latent distribution, and the same density estimate is then maximized during counterfactual search. This makes the 'in-distribution' claim partly a consequence of the model's own construction. However, the density estimator itself is learned from real latent codes, and the external IM1/IM2 evaluation provides independent grounding. I did not count the validity issue around Eq. (18) as circularity: the claim that the boundary constraint θ_c^T f_e(x+δ)=0 'naturally makes the counterfactual sample valid' is a mathematical gap (logit=0 gives p=0.5, not a label flip), but it is not a definitional reduction of output to input. The paper's self-citations to [59] and [60] are background/baseline references and are not load-bearing for the main derivation. Overall, the central contribution has independent empirical content, so the circularity score is 3 rather than higher.

Axiom & Free-Parameter Ledger

6 free parameters · 6 axioms · 0 invented entities

The central claim rests on the RFF approximation, the assumed density functional form, and the boundary-validity assumption; none of these are derived from first principles in the paper. Several hyperparameters are left unspecified in the main experiments.

free parameters (6)
  • Latent dimension d
    Central to the autoencoder; the main experiments do not report the value used, only ablation studies vary it (Figure 6).
  • RFF dimension S
    Width of the random Fourier features; the main experiments do not report the value, ablation in Figure 7 varies it.
  • Kernel width b = 1 (fixed in Eq. 2), but Algorithm 2 lists b_i
    Eq. (2) fixes b=1 because embeddings can scale, but Algorithm 2 still uses per-feature widths b_i, leaving the actual value used in experiments unclear.
  • Regularization rate beta = around 0.4 on LCD (Figure 4)
    Selected by minimizing KL between the counterfactual distribution and the model's own density estimate (Section 4.6); effectively a hand-chosen internal criterion, not an externally validated constant.
  • Density estimator envelope parameters mu, Sigma
    The Gaussian envelope N(lambda|mu,sigma) in Eq. (12) and (20) requires estimates of the latent mean and covariance; the paper does not state how these are computed.
  • Search step size and max iterations
    The dual ascent in Eqs. (21)-(24) requires step sizes, number of alternating updates, and initialization of delta; none are specified.
axioms (6)
  • standard math Random Fourier feature approximation converges to the RBF kernel as S approaches infinity.
    Section 3, Definition 2: phi(x)^T phi(x') -> exp(-1/2||x-x'||^2) as S->infinity. The model relies on this for both the encoder/decoder and the density estimator.
  • standard math Gaussian processes are valid priors for the encoder and decoder functions.
    Section 4.2, Eqs. (6)-(8): fe and fd are defined as Gaussian processes with RFF approximation; the GP interpretation is standard.
  • domain assumption The density functional form p(lambda) = e^{f(lambda)} N(lambda|mu,sigma) / Z is an adequate model for the latent distribution.
    Section 4.4, Eq. (12): the paper assumes this exponentiated-GP-with-Gaussian-envelope form can represent the latent distribution, with no comparison to alternative density models.
  • ad hoc to paper A latent code exactly on the decision boundary counts as a valid counterfactual.
    Section 4.5, Eq. (18): enforcing theta^T fe(x+delta)=0 is assumed to flip the label; the paper does not derive how a boundary point becomes a valid counterfactual.
  • domain assumption Minimizing KL(q_beta|p) selects the correct regularization rate beta.
    Section 4.6: the KL between the counterfactual distribution and the model's learned density on the decision boundary is assumed to be the right criterion for in-distribution counterfactuals.
  • domain assumption The classifier being explained is the linear head on the encoder's latent space.
    Section 4.3: the paper explains the GPAE's own classifier (fc(lambda)=lambda^T theta_c), not an arbitrary black-box; the relevance of the explanation depends on this being the classifier of interest.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of An Explainable Gaussian Process Auto-encoder for Tabular Data." pith.science (2026). https://pith.science/paper/3FK3NZ6O

@misc{pith2026250900884,
  author       = {Pith},
  title        = {Pith review of: An Explainable Gaussian Process Auto-encoder for Tabular Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3FK3NZ6O}},
  note         = {Machine review of arXiv:2509.00884}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Explainable machine learning has attracted much interest in the community where the stakes are high. Counterfactual explanations methods have become an important tool in explaining a black-box model. The recent advances have leveraged the power of generative models such as an autoencoder. In this paper, we propose a novel method using a Gaussian process to construct the auto-encoder architecture for generating counterfactual samples. The resulting model requires fewer learnable parameters and thus is less prone to overfitting. We also introduce a novel density estimator that allows for searching for in-distribution samples. Furthermore, we introduce an algorithm for selecting the optimal regularization rate on density estimator while searching for counterfactuals. We experiment with our method in several large-scale tabular datasets and compare with other auto-encoder-based methods. The results show that our method is capable of generating diversified and in-distribution counterfactual samples.

Figures

Figures reproduced from arXiv: 2509.00884 by Brian Barr, John Paisley, Wei Zhang.

Figure 1
Figure 1. Figure 1: The framework of our GPAE model. The RFF mapping ( [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A visual example for illustrative purposes. Left: A rendering of the VAE’s [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The 2D visualization of density estimation of LCD dataset. (a)The scatter plot [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The study on the selection of β in LCD data set. Left: the KL divergence between true density and estimated density in latent space vs β. The KL is minimized around β = 0.4. The KL increases with lower or higher β, indicating less matching between the estimated density and true density of the learned latent representation. Right: the visualization of 2D latent space of the target class (upper left corner) … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparisons between GPAE and others methods on LCD dataset. [PITH_FULL_IMAGE:figures/full_fig_p033_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The relationship between different evaluation metrics and latent dimensions for the [PITH_FULL_IMAGE:figures/full_fig_p034_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Evaluation of the LCD dataset using different metrics as a function of RFF [PITH_FULL_IMAGE:figures/full_fig_p035_7.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

60 extracted references · 45 canonical work pages · 9 internal anchors

  1. [1]

    Neural additive models: Interpretable machine learning with neural nets

    Rishabh Agarwal, Levi Melnick, Nicholas Frosst, Xuezhou Zhang, Ben Lengerich, Rich Caruana, and Geoffrey E Hinton. Neural additive models: Interpretable machine learning with neural nets. Advances in Neural Information Processing Systems, 2021

    work page 2021

  2. [2]

    The inverse classification problem

    Charu C Aggarwal, Chen Chen, and Jiawei Han. The inverse classification problem. Journal of Computer Science and Technology , 25(3):458–468, 2010

    work page 2010

  3. [3]

    Variational autoencoder based anomaly detection using reconstruction probability

    Jinwon An and Sungzoon Cho. Variational autoencoder based anomaly detection using reconstruction probability. Special Lecture on IE, 2(1): 1–18, 2015

    work page 2015

  4. [4]

    Getting a clue: A method for explaining uncertainty estimates

    Javier Antor´ an, Umang Bhatt, Tameem Adel, Adrian Weller, and Jos´ e Miguel Hern´ andez-Lobato. Getting a clue: A method for explaining uncertainty estimates. arXiv preprint arXiv:2006.06848 , 2020

    Pith/arXiv arXiv 2006

  5. [5]

    Node-gam: Neural generalized additive model for interpretable deep learning

    Chun-Hao Chang, Rich Caruana, and Anna Goldenberg. Node-gam: Neural generalized additive model for interpretable deep learning. arXiv preprint arXiv:2106.01613, 2021. 35

    Pith/arXiv arXiv 2021

  6. [6]

    Kernel methods for deep learning

    Youngmin Cho and Lawrence Saul. Kernel methods for deep learning. In Advances in Neural Information Processing Systems , 2009

    work page 2009

  7. [7]

    Deep gaussian processes

    Andreas Damianou and Neil D Lawrence. Deep gaussian processes. In Artificial intelligence and statistics , pages 207–215. PMLR, 2013

    work page 2013

  8. [8]

    Multi-Objective Counterfactual Explanations

    Susanne Dandl, Christoph Molnar, Martin Binder, and Bernd Bis- chl. Multi-objective counterfactual explanations. arXiv preprint arXiv:2004.11165, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2004

  9. [9]

    Explanations based on the missing: Towards contrastive explanations with pertinent negatives

    Amit Dhurandhar, Pin-Yu Chen, Ronny Luss, Chun-Chen Tu, Paishun Ting, Karthikeyan Shanmugam, and Payel Das. Explanations based on the missing: Towards contrastive explanations with pertinent negatives. In Advances in neural information processing systems , pages 592–603, 2018

    work page 2018

  10. [10]

    Model Agnostic Contrastive Explanations for Structured Data

    Amit Dhurandhar, Tejaswini Pedapati, Avinash Balakrishnan, Pin- Yu Chen, Karthikeyan Shanmugam, and Ruchir Puri. Model ag- nostic contrastive explanations for structured data. arXiv preprint arXiv:1906.00117, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1906

  11. [11]

    Cruds: Counterfactual recourse using disentangled subspaces

    Michael Downs, Jonathan L Chu, Yaniv Yacoby, Finale Doshi-Velez, and Weiwei Pan. Cruds: Counterfactual recourse using disentangled subspaces. ICML WHI , 2020:1–23, 2020

    work page 2020

  12. [12]

    Random forest explainability using counterfactual sets

    Rub´ en R Fern´ andez, Isaac Mart ´ ın de Diego, V ´ ıctor Ace˜ na, Alberto Fern´ andez-Isabel, and Javier M Moguerza. Random forest explainability using counterfactual sets. Information Fusion, 63:196–207, 2020

    work page 2020

  13. [13]

    Gener- ative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Gener- ative adversarial nets. Advances in neural information processing systems, 27:2672–2680, 2014

    work page 2014

  14. [14]

    Interpretable credit ap- plication predictions with counterfactual explanations

    Rory Mc Grath, Luca Costabello, Chan Le Van, Paul Sweeney, Far- bod Kamiab, Zhao Shen, and Freddy Lecue. Interpretable credit ap- plication predictions with counterfactual explanations. arXiv preprint arXiv:1811.05245, 2018. 36

    Pith/arXiv arXiv 2018

  15. [15]

    Local rule-based explanations of black box decision systems

    Riccardo Guidotti, Anna Monreale, Salvatore Ruggieri, Dino Pedreschi, Franco Turini, and Fosca Giannotti. Local rule-based explanations of black box decision systems. arXiv preprint arXiv:1805.10820 , 2018

    Pith/arXiv arXiv 2018

  16. [16]

    Rocoursenet: Robust training of a prediction aware recourse model

    Hangzhi Guo, Feiran Jia, Jinghui Chen, Anna Squicciarini, and Amulya Yadav. Rocoursenet: Robust training of a prediction aware recourse model. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , pages 619–628, 2023

    work page 2023

  17. [17]

    Counternet: End- to-end training of prediction aware counterfactual explanations

    Hangzhi Guo, Thanh H Nguyen, and Amulya Yadav. Counternet: End- to-end training of prediction aware counterfactual explanations. In Pro- ceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages 577–589, 2023

    work page 2023

  18. [18]

    Global explanations of neural networks

    Mark Ibrahim, Melissa Louie, Ceena Modarres, and John Paisley. Global explanations of neural networks. In AAAI/ACM Conference on AI, Ethics, and Society , 2019

    work page 2019

  19. [19]

    Towards realistic individual recourse and actionable explanations in black-box decision making systems

    Shalmali Joshi, Oluwasanmi Koyejo, Warut Vijitbenjaronk, Been Kim, and Joydeep Ghosh. Towards realistic individual recourse and actionable explanations in black-box decision making systems. arXiv preprint arXiv:1907.09615, 2019

    Pith/arXiv arXiv 1907

  20. [20]

    Dace: Distribution-aware counterfactual explanation by mixed-integer linear optimization

    Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, and Hiroki Arimura. Dace: Distribution-aware counterfactual explanation by mixed-integer linear optimization. In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI-20, Christian Bessiere (Ed.). International Joint Conferences on Artificial Intelligence Organi- za...

    work page 2020

  21. [21]

    Model-agnostic counterfactual explanations for consequential decisions

    Amir-Hossein Karimi, Gilles Barthe, Borja Balle, and Isabel Valera. Model-agnostic counterfactual explanations for consequential decisions. In International Conference on Artificial Intelligence and Statistics, pages 895–905. PMLR, 2020

    work page 2020

  22. [22]

    Algorithmic Recourse: from Counterfactual Explanations to Interventions

    Amir-Hossein Karimi, Bernhard Sch¨ olkopf, and Isabel Valera. Algorith- mic recourse: from counterfactual explanations to interventions. arXiv preprint arXiv:2002.06278, 2020. 37

    work page internal anchor Pith review Pith/arXiv arXiv 2002

  23. [23]

    Good Counterfactuals and Where to Find Them: A Case-Based Technique for Generating Counterfactuals for Explainable AI (XAI)

    Mark T Keane and Barry Smyth. Good counterfactuals and where to find them: A case-based technique for generating counterfactuals for explainable ai (xai). arXiv preprint arXiv:2005.13997 , 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2005

  24. [24]

    Auto-encoding variational bayes

    Diederik P Kingma and Max Welling. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 , 2013

    Pith/arXiv arXiv 2013

  25. [25]

    Improved variational inference with inverse autore- gressive flow

    Durk P Kingma, Tim Salimans, Rafal Jozefowicz, Xi Chen, Ilya Sutskever, and Max Welling. Improved variational inference with inverse autore- gressive flow. Advances in neural information processing systems , 29, 2016

    work page 2016

  26. [26]

    Gram- mar variational autoencoder

    Matt J Kusner, Brooks Paige, and Jos´ e Miguel Hern´ andez-Lobato. Gram- mar variational autoencoder. In International Conference on Machine Learning, pages 1945–1954. PMLR, 2017

    work page 1945

  27. [27]

    Generalized inverse classification

    Michael T Lash, Qihang Lin, Nick Street, Jennifer G Robinson, and Jeffrey Ohlmann. Generalized inverse classification. In Proceedings of the 2017 SIAM International Conference on Data Mining , pages 162–170. SIAM, 2017

    work page 2017

  28. [28]

    Comparison-based inverse classification for interpretability in machine learning

    Thibault Laugel, Marie-Jeanne Lesot, Christophe Marsala, Xavier Re- nard, and Marcin Detyniecki. Comparison-based inverse classification for interpretability in machine learning. In International Conference on Information Processing and Management of Uncertainty in Knowledge- Based Systems, pages 100–111. Springer, 2018

    work page 2018

  29. [29]

    Supervised autoencoders: Improving generalization performance with unsupervised regularizers

    Lei Le, Andrew Patterson, and Martha White. Supervised autoencoders: Improving generalization performance with unsupervised regularizers. Advances in Neural Information Processing Systems , 31:107–117, 2018

    work page 2018

  30. [30]

    Deep neural networks as gaussian processes

    Jaehoon Lee, Yasaman Bahri, Roman Novak, Samuel S Schoenholz, Jeffrey Pennington, and Jascha Sohl-Dickstein. Deep neural networks as gaussian processes. arXiv preprint arXiv:1711.00165 , 2017

    Pith/arXiv arXiv 2017

  31. [31]

    FOCUS: Flexible Optimizable Counterfactual Explanations for Tree Ensembles

    Ana Lucic, Harrie Oosterhuis, Hinda Haned, and Maarten de Rijke. Focus: Flexible optimizable counterfactual explanations for tree ensembles.arXiv preprint arXiv:1911.12199, 2019. 38

    work page internal anchor Pith review Pith/arXiv arXiv 1911

  32. [32]

    Preserving causal constraints in counterfactual explanations for machine learning classifiers

    Divyat Mahajan, Chenhao Tan, and Amit Sharma. Preserving causal constraints in counterfactual explanations for machine learning classifiers. arXiv preprint arXiv:1912.03277 , 2019

    Pith/arXiv arXiv 1912

  33. [33]

    Explaining machine learning classifiers through diverse counterfactual explanations

    Ramaravind K Mothilal, Amit Sharma, and Chenhao Tan. Explaining machine learning classifiers through diverse counterfactual explanations. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, pages 607–617, 2020

    2020

  34. [34]

    Bayesian learning for neural networks , volume 118

    Radford M Neal. Bayesian learning for neural networks , volume 118. Springer Science & Business Media, 2012

    2012

  35. [35]

    Countergan: Generating realistic counterfactuals with residual generative adversarial nets

    Daniel Nemirovsky, Nicolas Thiebaut, Ye Xu, and Abhishek Gupta. Countergan: Generating realistic counterfactuals with residual generative adversarial nets. arXiv preprint arXiv:2009.05199 , 2020

    Pith/arXiv arXiv 2009

  36. [36]

    Learning model-agnostic counterfactual explanations for tabular data

    Martin Pawelczyk, Klaus Broelemann, and Gjergji Kasneci. Learning model-agnostic counterfactual explanations for tabular data. In Proceed- ings of The Web Conference 2020 , pages 3126–3132, 2020

    work page 2020

  37. [37]

    Carla: a python library to benchmark algorithmic recourse and counterfactual explanation algorithms

    Martin Pawelczyk, Sascha Bielawski, Johannes van den Heuvel, Tobias Richter, and Gjergji Kasneci. Carla: a python library to benchmark algorithmic recourse and counterfactual explanation algorithms. arXiv preprint arXiv:2108.00783, 2021

    Pith/arXiv arXiv 2021

  38. [38]

    Causality

    Judea Pearl. Causality. Cambridge university press, 2009

    2009

  39. [39]

    Neural oblivious decision ensembles for deep learning on tabular data

    Sergei Popov, Stanislav Morozov, and Artem Babenko. Neural oblivious decision ensembles for deep learning on tabular data. arXiv preprint arXiv:1909.06312, 2019

    Pith/arXiv arXiv 1909

  40. [40]

    Face: feasible and actionable counterfactual explanations

    Rafael Poyiadzi, Kacper Sokol, Raul Santos-Rodriguez, Tijl De Bie, and Peter Flach. Face: feasible and actionable counterfactual explanations. In Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , pages 344–350, 2020

    work page 2020

  41. [41]

    Variational Autoencoder for Deep Learning of Images, Labels and Captions

    Yunchen Pu, Zhe Gan, Ricardo Henao, Xin Yuan, Chunyuan Li, Andrew Stevens, and Lawrence Carin. Variational autoencoder for deep learning of images, labels and captions. arXiv preprint arXiv:1609.08976 , 2016. 39

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  42. [42]

    Neural basis models for interpretability

    Filip Radenovic, Abhimanyu Dubey, and Dhruv Mahajan. Neural basis models for interpretability. Advances in Neural Information Processing Systems, 2022

    work page 2022

  43. [43]

    Random features for large-scale kernel machines

    Ali Rahimi and Benjamin Recht. Random features for large-scale kernel machines. In J. Platt, D. Koller, Y. Singer, and S. Roweis, editors, Advances in Neural Information Processing Systems , volume 20. Curran Associates, Inc., 2008

    work page 2008

  44. [44]

    Generating Counterfactual and Contrastive Explanations using SHAP

    Shubham Rathi. Generating counterfactual and contrastive explanations using shap. arXiv preprint arXiv:1906.09293 , 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1906

  45. [45]

    ”why should i trust you?” explaining the predictions of any classifier

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. ”why should i trust you?” explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , 2016

    work page 2016

  46. [46]

    Efficient search for diverse coherent explanations

    Chris Russell. Efficient search for diverse coherent explanations. In Pro- ceedings of the Conference on Fairness, Accountability, and Transparency, pages 20–28, 2019

    work page 2019

  47. [47]

    Explaingan: Model explanation via decision boundary crossing transformations

    Pouya Samangouei, Ardavan Saeedi, Liam Nakagawa, and Nathan Sil- berman. Explaingan: Model explanation via decision boundary crossing transformations. In Proceedings of the European Conference on Computer Vision (ECCV) , pages 666–681, 2018

    work page 2018

  48. [48]

    Certifai: Coun- terfactual explanations for robustness, transparency, interpretability, and fairness of artificial intelligence models

    Shubham Sharma, Jette Henderson, and Joydeep Ghosh. Certifai: Coun- terfactual explanations for robustness, transparency, interpretability, and fairness of artificial intelligence models. arXiv preprint arXiv:1905.07857, 2019

    Pith/arXiv arXiv 1905

  49. [49]

    Learning important features through propagating activation differences

    Avanti Shrikumar, Peyton Greenside, and Anshul Kundaje. Learning important features through propagating activation differences. In Inter- national Conference on Machine Learning , 2017

    work page 2017

  50. [50]

    Axiomatic attribution for deep networks

    Mukund Sundararajan, Ankur Taly, and Qiqi Yan. Axiomatic attribution for deep networks. In International Conference on Machine Learning , 2017. 40

    work page 2017

  51. [51]

    Variational learning of inducing variables in sparse gaussian processes

    Michalis Titsias. Variational learning of inducing variables in sparse gaussian processes. In Artificial intelligence and statistics , pages 567–574. PMLR, 2009

    work page 2009

  52. [52]

    Interpretable predictions of tree-based ensembles via actionable feature tweaking

    Gabriele Tolomei, Fabrizio Silvestri, Andrew Haines, and Mounia Lal- mas. Interpretable predictions of tree-based ensembles via actionable feature tweaking. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining , pages 465–474, 2017

    work page 2017

  53. [53]

    Actionable recourse in linear classification

    Berk Ustun, Alexander Spangher, and Yang Liu. Actionable recourse in linear classification. In Proceedings of the Conference on Fairness, Accountability, and Transparency, pages 10–19, 2019

    work page 2019

  54. [54]

    Interpretable Counterfactual Explanations Guided by Prototypes

    Arnaud Van Looveren and Janis Klaise. Interpretable counterfactual explanations guided by prototypes. arXiv preprint arXiv:1907.02584 , 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  55. [55]

    Interpretable counterfactual explanations guided by prototypes

    Arnaud Van Looveren and Janis Klaise. Interpretable counterfactual explanations guided by prototypes. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases , pages 650–665. Springer, 2021

    work page 2021

  56. [56]

    Counterfactual explanations without opening the black box: Automated decisions and the gdpr

    Sandra Wachter, Brent Mittelstadt, and Chris Russell. Counterfactual explanations without opening the black box: Automated decisions and the gdpr. Harv. JL & Tech. , 31:841, 2017

    work page 2017

  57. [57]

    Measurable Counterfactual Local Explanations for Any Classifier

    Adam White and Artur d’Avila Garcez. Measurable counterfactual local explanations for any classifier. arXiv preprint arXiv:1908.03020 , 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1908

  58. [58]

    Deep bayesian nonparametric tracking

    Aonan Zhang and John Paisley. Deep bayesian nonparametric tracking. In International Conference on Machine Learning , pages 5833–5841. PMLR, 2018

    work page 2018

  59. [59]

    An interpretable deep clas- sifier for counterfactual generation

    Wei Zhang, Brian Barr, and John Paisley. An interpretable deep clas- sifier for counterfactual generation. In Proceedings of the Third ACM International Conference on AI in Finance , pages 36–43, 2022

    work page 2022

  60. [60]

    Gaussian process neural additive models

    Wei Zhang, Brian Barr, and John Paisley. Gaussian process neural additive models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 16865–16872, 2024. 41

    work page 2024

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.