Pith. sign in

REVIEW 5 major objections 5 minor 23 references

Counterfactual optimization for fault prevention in complex wind energy systems

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Counterfactual optimization can restore an anomalous wind-turbine transformer to a healthy state with minimal changes, with estimated savings around €3 million per year for a typical farm.

desk verdict Novel counterfactual-control formulation for wind turbines, but the classifier metrics are internally inconsistent and the savings estimate is not anchored to validated safety, so the empirical claims collapse. read the letter →

arxiv 2507.08849 v1 pith:IOBR3N2M submitted 2025-07-08 eess.SY cs.LGcs.SYmath.OC

Emilio Carrizosa , Martina Fischetti , Roshell Haaker , Juan Miguel Morales This is my paper · ORCID
classification eess.SYcs.LGcs.SYmath.OC MSC 90C1190C90
keywords CombinatorialoptimizationCounterfactualanalysisFaultpreventionWindenergyMachinelearningMathematicalmodelingMixed-integerquadraticprogrammingControlstrategy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that a machine-learning fault detector can be turned into a fault preventer: instead of just flagging a wind-turbine transformer as anomalous, the system should find the smallest feasible change to the controllable settings that makes the classifier call the component healthy again. This matters because offshore turbines are costly to visit and are usually shut down or heavily curtailed once something looks wrong, so a controller that restores safety with minimal power loss protects both the component and the revenue stream. The problem is framed as counterfactual optimization—find the nearest state, in normalized squared distance over power and temperatures, that the neural network labels 'good' with a confidence margin—and solved as a mixed-integer quadratic program with the classifier and learned temperature-response models embedded as constraints. Tests on one month of real farm data show the optimizer resolving most anomalous instants to feasible healthy states and recommending less drastic curtailment than the current alarm-based controller, with extra revenue estimated at about €11,500 for one turbine in that month, which the paper scales to more than €3 million per year for a 30-turbine farm.

What carries the argument

The central object is the counterfactual controller, model (7): a mixed-integer quadratic program whose decision variables are the controllable turbine features (power production, nacelle temperature, transformer temperature), whose objective is the normalized squared distance from the current anomalous measurement, and whose constraints embed the neural-network classifier's pre-sigmoid score with a confidence margin, learned gradient-boosted response models of the temperature controllers, bounds, integrality, and a power-curve upper limit. This machinery carries the argument because every claimed saving comes from the optimizer's ability to find a nearby state that the classifier calls healthy, rather than from a fixed rule or a uniform curtailment.

What would settle it

Put the controller on an actual test turbine for one month: at each detected anomaly, implement the optimizer's recommended power and temperature settings, then track transformer alarms, trips, and oil-temperature excursions in the following hours and days. If states the classifier labels 'good' still trigger alarms at the same rate as unmitigated anomalies, the safety premise fails; if the realized extra revenue per turbine is far from the estimated €11,500 in a month with a comparable number of anomalies, the savings claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a wind-turbine controller can be designed as a counterfactual optimizer: whenever the trained neural network marks the transformer state as anomalous, the controller solves a mixed-integer quadratic program that finds the nearest state the network classifies as good with a confidence margin, measuring distance by normalized squared changes to power, nacelle temperature, and transformer temperature. Feasibility is enforced by constraints that keep power within the warranted power curve, fix the uncontrollable features (ambient temperature, wind speed, time), and use gradient-boosted models of the manufacturer's black-box temperature controller to predict the nacelle and transformer temperatures that a chosen power level will produce. On the test month, 758 of 1197 instances needed no action, 407 were optimized, and 32 were infeasible (shutdown being the only safe option); the optimizer's recommended curtailment is less drastic than the existing alarm-driven controller's, and the paper estimates €11,500 of extra revenue for that turbine in May, i.e., about €300,000 per month for a 30-turbine park. The paper also shows the same machinery can be retuned to user risk preferences—accepting more false positives to catch more anomalies—and can be switched to a revenue-maximizing objective, yielding a further €1,500 per turbine over the baseline counterfactual strategy in the test month.

Load-bearing premise

The load-bearing premise is that the neural network's 'good' verdict is reliable enough that pushing a state just below its decision threshold truly makes the transformer safe; the paper reports only 22% balanced accuracy for that network (15% for the preference-tuned version), so if the classifier is systematically wrong about which states are healthy, the recommended counterfactuals may not prevent faults and the savings estimate collapses.

Editorial extensions

If this is right

  • Operators can act within the 10-minute data cadence: each anomalous timestamp is handed to the optimizer, which returns a concrete curtailment target rather than waiting for an alarm-triggered shutdown; the test month suggests roughly €11,500 extra revenue per turbine.
  • User risk preferences become a tunable dial: retraining the classifier with a deliberately skewed class balance raises the anomaly-detection rate and, even with that more conservative setting, the counterfactual controller still yields about €10,500 extra per turbine over the month.
  • The same model can be repurposed for a turbine manufacturer's perspective, letting the optimizer adjust temperature setpoints within a feasible band; the resulting control differs only modestly from the operator-only version, indicating flexibility without compromising feasibility.
  • Because the label can be redefined, any costly, labelable outcome—such as faults that required a crew visit—can be plugged into the same framework to produce a controller that minimizes visits.
  • A revenue-driven objective variant increases production further, adding about €1,500 per turbine per month on top of the distance-minimizing counterfactual, with the same safety constraints.

Reading between the lines

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

  • The revenue estimate is a linear extrapolation from one turbine in one month, and would shrink in months or farms with fewer anomalies, lower wind, or lower energy prices; the paper reports the assumptions but does not test them.
  • If the classifier's low balanced accuracy (22%, or 15% after preference tuning) means the 'good' region is not physically safe, the recommended counterfactuals could be unsafe despite being optimal; a robustness layer over the classifier's score or input measurements would be needed before field deployment.
  • The controller design transfers to other monitored assets with learned response models: the 10-minute cadence, the distance objective, and the embedded-classifier constraint are not transformer-specific, so gearboxes, blades, or generators could use the same template.
  • A direct way to validate the paper's physical-safety claim would be a field trial in which recommended power and temperature trajectories are actually executed and subsequent alarm counts are compared with the classifier's predictions; the paper only simulates the controller on historical data, so the offline savings need on-turbine confirmation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes a counterfactual optimization controller for oil-type transformers in offshore wind turbines. Given a neural-network classifier f that labels system states as good or anomalous, the controller solves a mixed-integer quadratic program (model (7)) to find a minimally distant state classified as good, subject to power-curve, temperature-emulator, and integrality constraints. The authors test the approach on real Vattenfall data for one turbine over the month of May, report that 407 of 1197 instances are optimized, claim extra revenue of about 11,500 EUR for that month and about 3 million EUR per year for a 30-turbine farm, and present extensions for user preferences and revenue-driven objectives.

Significance. The application of counterfactual optimization to energy-system control is a novel and potentially valuable direction, and the paper demonstrates a concrete optimization architecture in which trained machine-learning models are embedded as constraints in a solvable MIQP. The use of real industrial data and the explicit treatment of operational constraints (power curve, temperature limits, integrality) are strengths. However, the central claims that the recommended states are safe and that the approach yields large annual savings rest entirely on a classifier whose reported accuracy is internally inconsistent and, taken at face value, worse than random. Without a validated safety oracle or field evidence that the recommended actions prevent faults, the practical significance of the method is not established.

major comments (5)
  1. [Section 2.2.2, Figure 4] The reported test metrics are mutually inconsistent. Under the standard definition, balanced accuracy equals (TPR+TNR)/2, so a balanced accuracy of 22% implies TPR+TNR = 0.44, and hence any weighted average of TPR and TNR—including standard accuracy—cannot exceed 0.44. The reported standard accuracy of 66% is therefore impossible. This is not a cosmetic discrepancy; it means the true discriminative performance of f is unknowable from the paper, and every downstream claim built on f is thereby called into question.
  2. [Section 3, Eq. (7c); Section 4.1] The safety claim is circular and unsupported. Constraint (7c) forces the counterfactual into the region that f labels as good, so the optimizer's output is classified healthy by construction. The only evidence that this region corresponds to physically safe transformer states is the classifier's test performance, which—at 22% balanced accuracy—is far below random. No independent validation against actual faults or future alarms is provided; the purple alarm intervals in Figures 7–9 are not used to test whether the red/green recommended states prevent subsequent failures. The statement in Section 4.1 that safety 'could have been maintained' is therefore not established.
  3. [Section 4.1; Abstract] The economic extrapolation is not supported. The estimate of 11,500 EUR for one turbine in May is multiplied by 30 turbines and annualized to 'more than 3 million EUR per year.' No evidence shows that May is representative of other months or that the 407 optimized instances are typical; the authors themselves acknowledge in Section 7 that savings depend on the number of anomalies and failures. Since the classifier is unreliable, the power-production differences generating the revenue estimate may be artifacts of misclassification rather than safe operational changes. The abstract presents the 3-million-EUR figure as a headline result without these caveats.
  4. [Section 3.1.2, Eqs. (7d)–(7e)] The XGBoost temperature emulators n and t are embedded as hard constraints, but their accuracy is reported only as two RMSE values. The paper does not state whether these values are computed on a held-out test set, nor the scale of the normalized temperatures, so the constraints cannot be assessed. More importantly, no test demonstrates that following the recommended (x_P, x_TN, x_TT) actually produces those temperatures through the manufacturer's black-box controller. The claim in Section 3.1.2 that the optimized strategy is 'implementable by the current controller and safe for the component' is therefore unverified.
  5. [Section 5] The preference-tuning experiment further weakens the evidence. The revised neural network has a balanced accuracy of 15%, which is worse than random on both classes. Using this model as the objective oracle in model (7) cannot provide meaningful counterfactual safety, and the claimed extra gain of about 10,500 EUR for the month inherits all the validity problems of the base classifier. This section therefore does not demonstrate a useful mechanism for incorporating user preferences; it demonstrates only that the optimization pipeline can be re-run when the classifier changes.
minor comments (5)
  1. [Abstract; Section 1.3] The phrase '3 millioneper year' should read '3 million euros per year' (missing currency symbol and spacing), and the claim should be qualified as discussed in Major Comment 3.
  2. [Section 3, after Eq. (6)] The sentence 'constraints (3) imposes that the counterfactual be integral' mixes singular and plural and refers to the wrong equation; it should refer to constraint (6).
  3. [Section 3.1.2] The text says 'Figures 10 visualize the predicted values,' but the referenced figures are numbered 6(a) and 6(b); the figure numbering in this section is inconsistent.
  4. [Section 5] The statement that the same confusion matrix results 'no matter if we optimize the f1 score or the average precision score' is unclear, since the choice of scoring metric normally affects model selection and hence the confusion matrix.
  5. [Section 6] The added constraint x_P ≤ x̃_P + π x̃_P only bounds upward deviations of power from the baseline; the stated intention to limit the change in component status would require a two-sided bound, especially if negative energy prices are considered.

Circularity Check

1 steps flagged · score 6.0 of 10

The claimed safety/revenue result is a restatement of the classifier constraint that defines the counterfactual target; no independent validation of the classifier's 'good' region is provided.

  1. self definitional [Section 3, model (7), constraints (7b)-(7c); also Section 3.1.3 and Section 4.1]
    "We write a mathematical model to determine a x = [x P, xTN, xTT, xWS, xD, xTA] that is the most similar to x∗, but for which f(x) = 0 (that is, instance x is predicted as good). ... constraint (3) ensures that the counterfactual is predicted to belong to the class 0 (i.e., not an anomaly)."

    The healthy state is defined as f(x)=0, and constraint (7c) (y=f(x) ≤ s(τ)−ε) forces every solution of model (7) into that class by construction. The paper then reads this constrained output as an empirical safety result: 'our counterfactual controller shows that safety could have been maintained with a lower curtailment' (Section 4.1) and builds the 3 M€/year savings estimate on it. The only evidence that f's good region corresponds to physical health is that the labels used to train f are themselves 'internally defined by the company using Machine Learning techniques' (Section 2.2), i.e., another model's output.

full rationale

The core counterfactual loop is not circular in the algebraic sense: model (7) is a genuine optimization problem and 407 feasible solutions with power increases were computed. The circularity is semantic and load-bearing: the 'healthy/safe' label is the classifier output f, (7c) guarantees f(x)≤s(τ)−ε, and the paper then treats that guarantee as restored safety and converts it to revenue. Because the classifier was trained on company-ML-defined labels rather than on verified fault outcomes, and because its reported metrics are not credible (a 22% balanced accuracy implies TPR+TNR=0.44, so no class-weighted accuracy can reach the claimed 66%), f cannot serve as an external oracle. The revenue extrapolation from one turbine-month to 30 turbines-year also assumes away seasonal and turbine heterogeneity, but that is an extrapolation issue rather than a circularity. The paper does contain some independent content: the early-detection comparison against the turbine alarm system (Figure 8) and the reverse-engineering of the temperature controller (RMSE values) are external to the self-definitional loop. On balance, the central safety claim and the headline savings figure rest on the same classifier-defined target, warranting a 6 rather than a 0-2.

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

The central claim rests on several domain assumptions: company labels as ground truth, the classifier as a safety oracle, the XGBoost temperature emulators as faithful black-box models, and linear extrapolation of savings. No new physical entities are introduced. The free parameters are mostly user-set margins and prices that directly influence the reported economic figures.

free parameters (6)
  • epsilon (confidence margin) = not specified
    Added in constraint (7c) to ensure the counterfactual is classified as good with a margin; no value or tuning procedure is given.
  • m_delta_T (max temperature variation) = 30 degrees in Section 4.2
    User-set limit on drastic temperature changes; chosen arbitrarily in the manufacturer-perspective experiment.
  • pi (revenue slack bound) = 10%
    Limits the revenue-driven solution's deviation from the baseline counterfactual; chosen ad hoc in Section 6.
  • energy price r = 100 EUR/MWh
    Assumed fixed revenue per MWh proposed by the industrial partner; drives all savings estimates and is not tied to a specific farm.
  • under-sampling ratio = 25% more anomalies than good in Section 5
    Alters the training distribution to reflect user preference; chosen ad hoc.
  • NN hyperparameters = not fully specified
    Neural network trained with sklearn and K-fold CV; exact architecture, regularization, and threshold tuning are not reported.
assumptions (6)
  • domain assumption Company ML-generated anomaly labels are taken as ground truth for safe and unsafe states.
    Section 2.1 and 2.2: labels are produced by Vattenfall's internal ML model; the paper's classifier is trained to mimic them. If these labels are noisy or not causally linked to faults, the counterfactual target is meaningless.
  • ad hoc to paper The classifier f's decision boundary is a valid safety oracle; a point with f(x)=0 is actually safe.
    Constraint (7c) defines the counterfactual target solely through f. The paper's own confusion matrix gives balanced accuracy 22%, so f may not reliably identify safe states.
  • domain assumption The XGBoost emulators n(x) and t(x) accurately reproduce the manufacturer's black-box temperature controller.
    Section 3.1.2: n and t are fitted to historical data with RMSE 0.585 and 1.142 (normalized). No validation shows the fitted relations hold for out-of-sample or novel counterfactual setpoints.
  • domain assumption Suggested control setpoints are implementable and will lead to the modeled temperature responses in practice.
    Section 4 assumes the operator can set power and the black-box controller adjusts temperatures as modeled; no field test or closed-loop verification is provided.
  • domain assumption One-month, one-turbine revenue difference extrapolates linearly to all months and all turbines in a 30-turbine farm.
    Section 4.1 and 7: 11,500 EUR in May for one turbine times 30 turbines and 12 months gives 3M per year, ignoring seasonality, turbine heterogeneity, anomaly frequency, and price variation.
  • standard math Standard mathematical assumptions of the mixed-integer quadratic program.
    Model (7) is a MIQP with a convex quadratic objective; Gurobi is assumed to solve instances to optimality within the 15-minute time limit.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Counterfactual optimization for fault prevention in complex wind energy systems." pith.science (2026). https://pith.science/paper/IOBR3N2M

@misc{pith2026250708849,
  author       = {Pith},
  title        = {Pith review of: Counterfactual optimization for fault prevention in complex wind energy systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IOBR3N2M}},
  note         = {Machine review of arXiv:2507.08849}
}
read the original abstract

Machine Learning models are increasingly used in businesses to detect faults and anomalies in complex systems. In this work, we take this approach a step further: beyond merely detecting anomalies, we aim to identify the optimal control strategy that restores the system to a safe state with minimal disruption. We frame this challenge as a counterfactual problem: given a Machine Learning model that classifies system states as either good or anomalous, our goal is to determine the minimal adjustment to the system's control variables (i.e., its current status) that is necessary to return it to the good state. To achieve this, we leverage a mathematical model that finds the optimal counterfactual solution while respecting system specific constraints. Notably, most counterfactual analysis in the literature focuses on individual cases where a person seeks to alter their status relative to a decision made by a classifier, such as for loan approval or medical diagnosis. Our work addresses a fundamentally different challenge: optimizing counterfactuals for a complex energy system, specifically an offshore wind turbine oil type transformer. This application not only advances counterfactual optimization in a new domain but also opens avenues for broader research in this area. Our tests on real world data provided by our industrial partner show that our methodology easily adapts to user preferences and brings savings in the order of 3 million euros per year in a typical farm.

Figures

Figures reproduced from arXiv: 2507.08849 by the authors.

Figure 1
Figure 1. Input features over time. The data correspond to real measurements and come with a label that indicates [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Transformer data used to build our Machine Learning models and their split into training/validation and test [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Under-sampling technique to balance the training set: By comparing with [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Confusion Matrix for the Neural Network classifier [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: An example of a power curve showing the (non-linear) relation between power production and wind speed [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: We use XGBoost estimators to mimic the black-box temperature controller behavior. The temperature values [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Power production: Comparison between the historical power production at the turbine (in blue) and the [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: ML can help detecting faults in advance and Operations Research can suggest the optimal reaction to move [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Power production: Comparison between the historical power production at the turbine (in blue) and the [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: Transformer temperature: Comparison between the historical transformer temperature at the turbine (in blue) [PITH_FULL_IMAGE:figures/full_fig_p020_10.png]
Figure 11
Figure 11. Figure 11: Confusion Matrix for our Neural Network classifier that focuses on reducing false-negatives (bottom-left [PITH_FULL_IMAGE:figures/full_fig_p021_11.png]
Figure 12
Figure 12. Figure 12: Power production: comparison between the historical of the power production at the turbine (in blue) and [PITH_FULL_IMAGE:figures/full_fig_p021_12.png]
Figure 13
Figure 13. Figure 13: Power production: comparison between the historical power production at the turbine (in blue) and the [PITH_FULL_IMAGE:figures/full_fig_p021_13.png]
Figure 14
Figure 14. Figure 14: Power production: comparison between the historical power production at the turbine (in blue), standard [PITH_FULL_IMAGE:figures/full_fig_p022_14.png]
Figure 15
Figure 15. Figure 15: Comparison of power productions with the different types of controllers. Zoom in the period 2-3 of May [PITH_FULL_IMAGE:figures/full_fig_p023_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 20 canonical work pages

  1. [1]

    On the computation of counterfactual explanations – a survey

    Artelt, A., Hammer, B., 2019. On the computation of counterfactual explanations – a survey. arXiv:1911.07749

    arXiv 2019

  2. [2]

    Mathematical optimization modelling for group counterfactual explanations

    Carrizosa, E., Ramírez-Ayerbe, J., Romero Morales, D., 2024. Mathematical optimization modelling for group counterfactual explanations. European Journal of Operational Research 319, 399–412

    work page 2024

  3. [3]

    Multi-objective counterfactual explana- tions, in: Bäck, T., Preuss, M., Deutz, A., Wang, H., Doerr, C., Emmerich, M., Trautmann, H

    Dandl, S., Molnar, C., Binder, M., Bischl, B., 2020. Multi-objective counterfactual explana- tions, in: Bäck, T., Preuss, M., Deutz, A., Wang, H., Doerr, C., Emmerich, M., Trautmann, H. (Eds.), Parallel Problem Solving from Nature – PPSN XVI, Springer International Pub- lishing, Cham. pp. 448–469

    work page 2020

  4. [4]

    Exploring the Trade-off between Plausibility, Change Intensity and Adversarial Power in Counterfactual Explanations using Multi-objective Optimization

    Del Ser, J., Barredo-Arrieta, A., Díaz-Rodríguez, N., Herrera, F., Holzinger, A., 2022. Ex- ploring the trade-off between plausibility, change intensity and adversarial power in counter- factual explanations using multi-objective optimization.arXiv:2205.10232

    work page Pith review arXiv 2022

  5. [5]

    On the optimized design of next-generation wind farms

    Fischetti, M., 2021. On the optimized design of next-generation wind farms. European Journal of Operational Research 291, 862–870

    work page 2021

  6. [6]

    Machine learning meets mathematical optimization to predict the optimal production of offshore wind parks

    Fischetti, M., Fraccaro, M., 2019. Machine learning meets mathematical optimization to predict the optimal production of offshore wind parks. Computers and Operations Research 106, 289–297

    work page 2019

  7. [7]

    Deep neural networks and mixed integer linear optimization

    Fischetti, M., Jo, J., 2018. Deep neural networks and mixed integer linear optimization. Constraints 23, 296–309

    work page 2018

  8. [8]

    Vattenfall opti- mizes offshore wind farm design

    Fischetti, M., Kristoffersen, J.R., Hjort, T., Monaci, M., Pisinger, D., 2020. Vattenfall opti- mizes offshore wind farm design. INFORMS Journal on Applied Analytics 50, 80–94

    work page 2020

Show all 23 references
  1. [9]

    Counterfactual explanations and how to find them: literature review and benchmarking

    Guidotti, R., 2024. Counterfactual explanations and how to find them: literature review and benchmarking. Data Mining and Knowledge Discovery 38, 2770–2824

    2024

  2. [10]

    Factual and counterfactual explanations for black box decision making

    Guidotti, R., Monreale, A., Giannotti, F., Pedreschi, D., Ruggieri, S., Turini, F., 2019. Factual and counterfactual explanations for black box decision making. IEEE Intelligent Systems 34, 14–23

    2019

  3. [11]

    Gurobi machine learning manual

    Gurobi, 2024. Gurobi machine learning manual. URL:https:// gurobi-machinelearning.readthedocs.io/en/stable/index.html

    2024

  4. [12]

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

    Joshi, S., Koyejo, O., Vijitbenjaronk, W., Kim, B., Ghosh, J., 2019. Towards realis- tic individual recourse and actionable explanations in black-box decision making systems. arXiv:1907.09615. 25

    2019 arXiv

  5. [13]

    Dace: Distribution-aware coun- terfactual explanation by mixed-integer linear optimization, in: Bessiere, C

    Kanamori, K., Takagi, T., Kobayashi, K., Arimura, H., 2020. Dace: Distribution-aware coun- terfactual explanation by mixed-integer linear optimization, in: Bessiere, C. (Ed.), Proceed- ings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI-20...

    2020

  6. [14]

    Distribution-aware counterfac- tual explanation by mixed-integer linear optimization

    Kanamori, K., Takagi, T., Kobayashi, K., Arimura, H., 2021. Distribution-aware counterfac- tual explanation by mixed-integer linear optimization. Transactions of the Japanese Society for Artificial Intelligence 36

    2021

  7. [15]

    Model-agnostic counterfactual explana- tions for consequential decisions, in: Chiappa, S., Calandra, R

    Karimi, A., Barthe, G., Balle, B., Valera, I., 2020. Model-agnostic counterfactual explana- tions for consequential decisions, in: Chiappa, S., Calandra, R. (Eds.), The 23rd International Conference on Artificial Intelligence and Statistics, AISTATS 2020, 26-28 August 2020, On...

    2020

  8. [16]

    Applications of machine learning to machine fault diagnosis: A review and roadmap

    Lei, Y ., Yang, B., Jiang, X., Jia, F., Li, N., Nandi, A.K., 2020. Applications of machine learning to machine fault diagnosis: A review and roadmap. Mechanical Systems and Signal Processing 138, 106587

    2020

  9. [17]

    Find- ing regions of counterfactual explanations via robust optimization

    Maragno, D., Kurtz, J., Rober, T.E., Goedhart, R., Birbil, S.I., den Hertog, D., 2024. Find- ing regions of counterfactual explanations via robust optimization. INFORMS Journal on Computing 36, 1316–1334

    2024

  10. [18]

    Mining pareto-optimal counterfac- tual antecedents with a branch-and-bound model-agnostic algorithm

    Raimundo, Marcos M.and Nonato, L.G., Poco, J., 2024. Mining pareto-optimal counterfac- tual antecedents with a branch-and-bound model-agnostic algorithm. Data Min Knowl Disc 36, 2942–2974

    2024

  11. [19]

    Synthesizing action sequences for modifying model decisions

    Ramakrishnan, G., Lee, Y .C., Albarghouthi, A., 2020. Synthesizing action sequences for modifying model decisions. Proceedings of the AAAI Conference on Artificial Intelligence 34, 5462–5469

    2020

  12. [20]

    scikit-learn: Machine learning in python

    ScikitLearn, 2025. scikit-learn: Machine learning in python. URL:https:// scikit-learn.org/stable/index.html

    2025

  13. [21]

    A systematic analysis of performance measures for clas- sification tasks

    Sokolova, M., Lapalme, G., 2009. A systematic analysis of performance measures for clas- sification tasks. Information Processing and Management 45, 427–437

    2009

  14. [22]

    Vashishtha, G., Chauhan, S., Sehri, M., Zimroz, R., Dumond, P., Kumar, R., Gupta, M.K.,

  15. [2025]

    Measurement 242, 116216

    A roadmap to fault diagnosis of industrial machines via machine learning: A brief review. Measurement 242, 116216. 26 Acknowledgements This research has been funded under Grant QUALIFICA (PROGRAMA: AYUDAS A AC- CIONES COMPLEMENTARIAS DE I+D+i) by Junta de Andalucía grant numbe...

    doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.