Pith. sign in

REVIEW 5 major objections 5 minor 22 references

ZAPs shows that DeFi reward systems can cap rewards by a protocol's global volume share and layer anomaly detection on top, cutting adversarial capture by 30–90% while barely affecting legitimate users.

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 ↗

T0 review · deepseek-v4-flash

2026-08-01 00:08 UTC pith:QIKSRGYK

load-bearing objection A genuinely useful deployed DeFi reward-attribution framework, but the headline ensemble gain rests on isolation-forest evaluation that leaks its training negatives—fixable, but central. the 5 major comments →

arxiv 2607.27859 v1 pith:QIKSRGYK submitted 2026-07-30 q-fin.GN cs.LG

ZAPs: A Reward Attribution Framework for DeFi Ecosystems with Adversarial-Robust Scoring via Parallel Anomaly Ensemble Detection

Girish G N , Ashutosh Sahoo , Ajay Bhat , Akshay SP , Gurukiran S , Parag Paul , Dhanashekar Kandaswamy This is my paper
classification q-fin.GN cs.LG
keywords DeFi incentivesreward attributionsybil resistanceanomaly detectionisolation forestone-class autoencoderairdrop farmingcross-domain weighting
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.

ZAPs asks whether a reward-attribution mechanism can pay genuine economic contribution in proportion to ecosystem significance while resisting bots and sybils that manufacture that contribution. It answers yes by combining two structural safeguards — a composite activity score whose components are each bounded, and a two-layer weight that telescopes into a protocol's global volume share, capping any single wallet's extractable reward via Proposition 1 — with a four-layer behavioral defense whose core is a parallel anomaly ensemble. The headline empirical evidence is a held-out ROC-AUC of 0.923±0.013 for the ensemble versus 0.891±0.016 for the reconstruction model alone, with an unusual training-regime finding: fitting the isolation forest on the pooled population inverts its polarity (0.250 AUC) and destroys the ensemble gain, while fitting it on the benign population works. A sympathetic reader would care because billions in token distributions are currently paid out against naive volume and count heuristics that industrial farming exploits; ZAPs offers a deployed, measured alternative that is robust by construction, not by retrofit.

Core claim

The paper claims that reward attribution in DeFi can be made resistant to adversarial manufacturing by making the scoring function itself the defense. Each wallet gets a composite score with three bounded, additive components: percentile-anchored volume (capped at a high population percentile), engagement duration (capped), and a quality signal. A two-layer cross-domain weight multiplies protocol share within sector by sector share within the ecosystem, telescoping to the protocol's global volume share; Proposition 1 proves that no wallet can extract more than S·V_p/V_E from any protocol, regardless of how cheaply local dominance is bought. On the behavioral side, a four-layer stack (transac

What carries the argument

The central mechanism is a composable attribution algebra: a bounded per-(wallet, protocol) composite score s_{w,p} (Equation 1) whose components are individually capped via percentile anchoring and a time ceiling, multiplied by a telescoping cross-domain weight ω_p = V_p/V_σ(p) · V_σ(p)/V_E = V_p/V_E (Equation 4). This weight is what enacts Proposition 1's bound on extractable reward. The second key object is the parallel anomaly ensemble of Section 3.5.2: Detector A is a one-class reconstruction autoencoder with an over-complete latent space (10→64→16→64→10), trained only on malicious wallets so that low reconstruction error signals maliciousness; Detector B is an isolation forest, whose t

Load-bearing premise

The empirical results rest on a benign comparison set of only 383 wallets with a median of 499 transactions versus 21 for malicious wallets; if a representative benign panel changes which wallets are statistical outliers, the isolation forest's polarity inversion and the ensemble's +0.032 AUC gain could shrink or disappear.

What would settle it

Re-run the isolation forest on a benign sample frequency-matched to the malicious wallets on transaction count and activity distribution; if the 0.250 inversion does not reproduce, or if the ensemble gain at λ=0.90 vanishes, the central robustness claim fails. An independent labeled corpus of 1,073 farming (not exploit) wallets would also test whether the 0.923 AUC transfers from phishing/exploit attributions to the actual adversarial population of interest.

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

If this is right

  • If the central claim holds, any volume-based reward program can structurally eliminate the niche-protocol farming exploit by applying the telescoping weight of Equation 4, since extractable reward at a protocol is bounded by that protocol's global volume share.
  • The isolation-forest training regime becomes a deployment requirement: fit the detector on the benign population, never on pooled production traffic, or the detector actively inverts and degrades the ensemble.
  • The graduated penalty structure implies that false-positive harm is bounded: a legitimate wallet misclassified by one tier keeps roughly half its reward instead of losing everything, which is what allows the 1–8% legitimate-user impact.
  • The framework's composability means the same attribution algebra can be extended to other permissionless incentive systems (airdrops, points, liquidity mining) with the same adversarial assumptions.
  • In live production, the framework reports a 56% reduction in sybil allocation and a 49% increase in quality-wallet participation, which suggests measurable economic benefit beyond detection accuracy.

Where Pith is reading between the lines

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

  • Editorial inference: the polarity-inversion finding may generalize beyond DeFi — any unsupervised anomaly detector used in an adversarial, class-imbalanced setting should be fit on the population that defines 'normal' for the purpose at hand, rather than on whatever data is available.
  • Editorial inference: because the paper admits the quality dimension q_{w,p} is static in the reported deployment, the actual distributional statistics likely understate what the full framework would achieve if sector-specific quality signals were activated.
  • Editorial inference: a natural testable extension is to apply ZAPs-style weighting to quadratic funding or retroactive public-goods funding, where the same tension between capital-weighted and participant-weighted allocation exists and where a protocol-level volume-share cap could be replaced by a different ecosystem-significance metric.
  • Editorial inference: the reported 30–90% adversarial capture reduction comes from constructed scenarios; a stronger follow-up would be to run a live randomized trial comparing ZAPs against a naive volume formula on matched populations, measuring both sybil capture and legitimate-user retention.

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 / 5 minor

Summary. The paper presents ZAPs, a reward-attribution framework for DeFi that combines a percentile-normalized composite activity score, a two-layer cross-domain volume weighting, and a four-layer adversarial detection stack with graduated penalties. The central claims are (i) a structural bound on extractable rewards at any protocol (Proposition 1), (ii) an anomaly ensemble that reaches 0.923 ROC-AUC versus 0.891 for the reconstruction model alone when the isolation forest is fit on benign wallets, (iii) a discovered polarity inversion of the isolation forest under pooled fitting, (iv) simulated adversarial reward-capture reductions of 30–90% with 1–8% legitimate-user impact, and (v) live production improvements. The paper is candid about several limitations, including a small activity-skewed benign set, a static quality dimension, and unquantified post-distribution memory, but the central empirical claims are weakened by evaluation-design issues described below.

Significance. If the claims hold, ZAPs would be a useful deployed artifact for a practically important problem: sybil-resistant and adversarial-robust reward attribution in DeFi. The paper's strengths are its explicit architectural description, the honest statement of several limitations in Section 6, and the identification of an interesting empirical phenomenon—the polarity inversion of an isolation forest under pooled fitting (§4.3, §5.2). However, the headline ensemble gain is not currently supported because of a training/evaluation overlap and test-set-based selection of λ. The structural contribution (Proposition 1) is correct but reduces to the definition of the telescoping weight. The work is therefore more of an engineering report with a promising but incompletely validated detection result than a fully established scientific claim at this stage.

major comments (5)
  1. [§4.3 / Table 4] The isolation forest is trained on the benign population and evaluated on the same benign set. The evaluation set is 'the held-out 30% (322 wallets) plus the full benign set' of 383 wallets, and the benign-fit isolation forest is explicitly fit on the benign population. This is transductive leakage: the forest has seen the very benign wallets it is scored against. The reported 0.638 AUC for the benign-fit forest and the +0.032 ensemble gain therefore measure the detector on its own training negatives, not on unseen data. A proper split must hold out a benign test fold from the forest's training set. Without this, the claim that the ensemble beats its strongest member and the claim that benign-fit 'inverts polarity' are not established. This is the most load-bearing weakness in the empirical support.
  2. [§4.3 / Figure 5] The optimal fusion weight λ* is chosen on the same evaluation set used to report the ensemble AUC. The text reports 'best over sweep' and λ*=0.90, and Figure 5 plots the sweep over the labeled evaluation data. Selecting the fusion weight on the test set and then reporting test performance at that selected value inflates the reported gain and can explain part or all of the +0.032 ROC-AUC difference. The evaluation should use nested cross-validation or a separate validation split for λ, with the test set touched only once.
  3. [§6 / §4.3] The benign comparison set consists of only 383 wallets with median 499 transactions versus a malicious median of 21 transactions. The paper itself notes that the magnitude 'should be expected to depend on the benign sampling frame.' This activity skew is not merely a quantitative detail: the isolation forest's polarity inversion is driven by the benign set being heavy-activity outliers relative to the pooled population, and the standalone 0.638 AUC is measured on that skewed in-sample set. In addition, the labeled corpus is 'drawn from public exploit and phishing attributions,' not airdrop farming, so Table 4 measures transfer to a proxy population. The robustness claims for farming therefore rest on an acknowledged proxy and a non-representative comparison set. A representative benign panel and, ideally, labeled farming wallets are needed before the central robustness claim can be accep
  4. [§3.3 / Proposition 1] The proposition is correct but definitional. Its proof uses only sw,p ≤ S and the telescoped identity ωp = Vp/VE from Equation (4). The statement is therefore a restatement of the normalization choices, not an independent structural result. The paper presents this as 'closing the most common structural exploit' and as a formal contribution; this overstates the content. The mechanism may still be useful as a design choice, but the proposition should be presented as a design identity that follows from the definition of ωp, and the discussion should make clear that it does not by itself bound total ecosystem-level extraction, only the per-protocol attribution for a single wallet.
  5. [§4.7 / §6] The production campaign results are reported as point estimates without uncertainty, sample sizes, baseline definitions, or statistical tests: '56% reduction in sybil allocation, 49% increase in quality-wallet participation, 50% reduction in sell pressure.' Section 6 states that the post-distribution memory mechanism is unquantified and no ablation isolates its contribution. Given that Layer 3 is a component of the full stack, the production attribution of effects to the framework is not supported. The authors should either provide campaign-level detail with error bars and matched baselines or explicitly label these as indicative observations rather than validated effect estimates.
minor comments (5)
  1. [§3.2–3.5] Many hyperparameters are withheld: α, β, γ, τ, T*, λ, penalty breakpoints, and clustering thresholds. Withholding parameters is understandable for a deployed system, but it prevents independent replication. At minimum, report a sensitivity analysis over plausible ranges, or release anonymized feature vectors and evaluation code.
  2. [§4.3 / Table 4] The table reports λ=1.00 as 'equivalent to PA alone' but does not include a λ=0 row for the isolation forest alone. Since Section 5.2 discusses the cost of removing the reconstruction model, the λ=0 standalone AUCs should be shown explicitly.
  3. [§4.4 / Table 5] The simulation results are point estimates. They should be accompanied by variance across seeds or scenario perturbations, especially because the 'change' percentages are the central quantitative evidence for the asymmetry claim.
  4. [§5.2] The statement that removing the reconstruction model 'costs 0.285' is not directly readable from Table 4. Add the λ=0 row or compute the difference from the shown numbers explicitly.
  5. [Equation (11)] The coefficient of variation uses sample mean in the denominator. For attributes with near-zero means this is unstable; state the handling of such cases (e.g., a floor or exclusion).

Circularity Check

2 steps flagged

Proposition 1 reduces to the definition of ω_p = V_p/V_E, and the benign-fit isolation forest is scored on the same benign wallets used to train it; the ensemble gain is therefore not independent of its fitting data.

specific steps
  1. self definitional [Section 3.3, Eq. (4)-(5), Proposition 1 and its Proof]
    "The effective weight is the product of two nested shares: ωp = Vp/Vσ(p) · Vσ(p)/VE = Vp/VE. ... Total attribution ... is Rw = µ(w) Σ_p s_{w,p} ωp. Proposition 1 ... s_{w,p} ωp ≤ S·Vp/VE ... Proof ... Substituting the telescoped form of Equation (4) yields s_{w,p} ωp ≤ S Vp/VE."

    The claimed bound is exactly the definition of ω_p inserted into the reward formula. Because each component of s_{w,p} is bounded by S by Equations (2)-(3), the 'proof' is a restatement of the chosen normalization. The paper advertises Proposition 1 as proving that niche-protocol farming is bounded by global volume share, but that property was put in by definition when the reward was defined as s_{w,p}ω_p with ω_p=V_p/V_E. No independent derivation is supplied.

  2. fitted input called prediction [Section 4.3, Table 4 / Figure 5; see also Section 5.2]
    "The reconstruction model is retrained one-class on a 70% fold of the malicious corpus; the held-out 30% (322 wallets) plus the full benign set form the evaluation set ... Fit on the benign population alone, as a proper normality model, the same detector reaches 0.638 ... only the benign-fit configuration produces an ensemble that beats its strongest member: 0.923 against 0.891, a gain of +0.032 ROC-AUC at λ* = 0.90."

    The isolation forest's 'benign population' training regime in this section is the same 383-wallet benign set that is included in the evaluation set. The 0.638 AUC and the +0.032 ensemble gain are therefore measured on data the forest was fit on, at least for the benign class; this is transductive leakage, not a held-out prediction. The choice of λ* as the best over the sweep on this same evaluation set further inflates the reported peak. The limitation section acknowledges the benign set's activity skew but not this train/evaluation overlap, so the headline ensemble improvement over the reconstruction model is not independently established.

full rationale

The paper's structural result (Proposition 1) is self-definitional: the bound s_{w,p}ω_p ≤ S V_p/V_E follows immediately from setting ω_p=V_p/V_E in Eq. (4) and bounding s_{w,p} by S in Eq. (1); calling it a proof adds no information. The empirical ensemble claim is compromised by using the same benign wallets to fit the isolation forest and to score it, plus selecting λ* on the test sweep, so the +0.032 gain over the reconstruction model is not a clean out-of-sample result. The zScore self-citations [6,7] are not load-bearing because the quality term q_{w,p} is held at a calibrated baseline for all users in the reported deployment; this lowers the circularity score. The remaining experiments (adversarial simulations, production metrics) are not themselves circular, though they inherit the ensemble's evaluation weakness. Overall, two load-bearing steps reduce to their own inputs/construction, so the central claims are only partially independent.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 0 invented entities

The framework's headline bound is a definitional consequence of its own weighting, and the empirical claims rest on self-collected data with several disclosed domain assumptions (proxy labels, skewed benign set). The quality signal is the authors' own zScore work but is held constant in the reported deployment, so it does not add independent evidence.

free parameters (6)
  • α, β, γ (composite score mixing coefficients) = withheld
    Eq. (1): 'calibrated empirically and withheld'; affect all distributional statistics; Proposition 1 only needs positivity and sum-to-one.
  • τ (high percentile anchor in volume normalization) = withheld
    Eq. (2): a 'high fixed τ'; the value determines where the volume clip binds and shapes the reward distribution.
  • T* (engagement ceiling) = withheld
    Eq. (3): 'design-chosen ceiling'; determines how quickly engagement saturates.
  • λ (ensemble fusion weight) = λ*=0.90 (reported; production value withheld)
    Eq. (10); chosen as best over the sweep on the evaluation set in §4.3, so the reported 0.923 AUC is a selected maximum over this parameter.
  • Penalty tier breakpoints and multipliers = ~50% (tier 3), ~80% (tier 4); breakpoints withheld
    §3.5.5: numeric breakpoints are withheld; these determine the legitimate-user cost and the distribution of tiers.
  • Sybil clustering thresholds (CV_max, D_max, funder gate Ψ) = withheld
    Algorithm 1 and Eq. (11); the validation scan excludes 39 high-degree funders; thresholds are not published.
axioms (5)
  • domain assumption Dollar volume at a protocol is the primary signal of economic contribution (after normalization)
    Eqs. (1)-(2): the whole scoring function builds on volume; non-volume signals are only an engagement term and a constant quality floor in the reference deployment.
  • domain assumption High-percentile normalization is a valid reference for differentiation
    §3.2: the choice of a high percentile anchors the clip; the ablation in §5.1 shows max/mean normalization fail, but no formal justification is given.
  • domain assumption A first-funder group with low behavioral CV indicates coordinated sybil activity
    §3.5.4, Eq. (11): shared provenance plus low CV is treated as evidence of one operator; no false-positive rate or validation on labeled families is provided.
  • domain assumption The labeled malicious corpus (exploit/phishing attributions) transfers to airdrop farming
    §4.3 and §6: the paper explicitly states this is a proxy population, not direct farming labels.
  • domain assumption The isolation-forest polarity inversion is robust rather than an artifact of the skewed benign sample
    §4.3/§6: asserted to be robust across seeds and forest sizes, but the magnitude is conceded to depend on the benign sampling frame.

pith-pipeline@v1.3.0-daily-deepseek · 12693 in / 14392 out tokens · 129176 ms · 2026-08-01T00:08:28.254313+00:00 · methodology

0 comments
read the original abstract

Incentive programs are central to user acquisition in decentralized finance, but many reward systems rely on raw volume, transaction count, and wallet count, making them vulnerable to bots and sybil operations. We present ZAPs, a reward attribution framework that combines economic contribution scoring with adversarial robustness. A composite activity score uses protocol-specific percentile normalization to limit whale dominance while preserving differentiation among users. A two-layer weighting mechanism combines protocol share within sector and sector share within the ecosystem, which reduces the profitability of farming small protocols. We show that the maximum reward obtainable from any protocol is bounded by that protocol's global volume share. ZAPs also introduces a four-layer defense stack consisting of transaction-level integrity checks, a parallel anomaly ensemble, post-distribution behavioral memory, and graph-based sybil clustering. The anomaly ensemble combines a one-class reconstruction model with an isolation forest and applies graduated rather than binary penalties. On 1,073 labeled malicious wallets covering 124,638 transactions, the ensemble achieves 0.923 +/- 0.013 ROC-AUC, compared with 0.891 +/- 0.016 for the reconstruction model alone, when the isolation forest is trained on benign wallets. Training it on the pooled population reverses its polarity and removes the ensemble gain. Controlled simulations reduce adversarial reward capture by 30-90 percent while legitimate-user scenarios change by 1-8 percent. Live campaigns recorded a 56 percent reduction in sybil allocation, a 49 percent increase in quality-wallet participation, and a 50 percent reduction in sell pressure.

Figures

Figures reproduced from arXiv: 2607.27859 by Ajay Bhat, Akshay SP, Ashutosh Sahoo, Dhanashekar Kandaswamy, Girish G N, Gurukiran S, Parag Paul.

Figure 1
Figure 1. Figure 1: ZAPs attribution pipeline. The structural path (composite score, cross-domain weighting) [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: One-class reconstruction detector. The latent width exceeds the input width, so discrim [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Funding-provenance clustering. Wallets sharing a first funder form candidate families; [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Graduated versus binary penalty response. Under binary exclusion a one-tier misclassifi [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Ensemble ROC-AUC across the fusion weight [PITH_FULL_IMAGE:figures/full_fig_p014_5.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

22 extracted references · 4 linked inside Pith

  1. [1]

    Malamud and M

    S. Malamud and M. Rostek. Decentralized exchange.American Economic Review, 107(11):3320–3362, 2017

    2017

  2. [2]

    Y. C. Lo and F. Medda. Uniswap and the emergence of the decentralized exchange.Journal of Financial Market Infrastructures, 10(2):1–25, 2021

    2021

  3. [3]

    wild west

    A. Aspris, S. Foley, J. Svec, and L. Wang. Decentralized exchanges: The “wild west” of cryp- tocurrency trading.International Review of Financial Analysis, 77:101845, 2021

    2021

  4. [4]

    Heimbach, E

    L. Heimbach, E. Schertenleib, and R. Wattenhofer. Risks and returns of Uniswap v3 liquidity providers. InProc. 4th ACM Conference on Advances in Financial Technologies, pages 89–101, 2022

    2022

  5. [5]

    A. A. Aigner and G. Dhaliwal. Uniswap: Impermanent loss and risk profile of a liquidity provider.arXiv preprint arXiv:2106.14404, 2021

    Pith/arXiv arXiv 2021

  6. [6]

    Udupi, A

    H. Udupi, A. Sahoo, A. S. P., G. S., P. Paul, and P. C. Martens. zScore: A universal de- centralised reputation system for the blockchain economy.arXiv preprint arXiv:2503.05718, 2025

    Pith/arXiv arXiv 2025

  7. [7]

    Kandaswamy, A

    D. Kandaswamy, A. Sahoo, A. SP, G. S, P. Paul, and G. G. N. Deep reputation scor- ing in DeFi: zScore-based wallet ranking from liquidity and trading signals.arXiv preprint arXiv:2507.20494, 2025

    Pith/arXiv arXiv 2025

  8. [8]

    D. Lin, J. Wu, Q. Fu, Z. Zheng, and T. Chen. RiskProp: Account risk rating on Ethereum via de-anonymous score and network propagation.IEEE Transactions on Dependable and Secure Computing, 2024

    2024

  9. [9]

    Nguyen, T.-D

    M.-T. Nguyen, T.-D. Trinh, and V.-B. Pham. A reputation scoring framework for lending pro- tocols using the PageRank algorithm. InInternational Symposium on Information and Com- munication Technology, pages 478–494. Springer, 2024. 18

    2024

  10. [10]

    S. S. Jain, H. Zhou, and G. Gu. Wire: Web3 integrated reputation engine. In2024 IEEE 44th International Conference on Distributed Computing Systems (ICDCS), pages 1388–1399. IEEE, 2024

    2024

  11. [11]

    N. G. Packin and Y. Lev-Aretz. Decentralized credit scoring: Black box 3.0.American Business Law Journal, 61(2):91–111, 2024

    2024

  12. [12]

    J. R. Douceur. The Sybil attack. InInternational Workshop on Peer-to-Peer Systems (IPTPS), pages 251–260. Springer, 2002

    2002

  13. [13]

    H. Yu, M. Kaminsky, P. B. Gibbons, and A. Flaxman. SybilGuard: Defending against Sybil attacks via social networks. InProc. ACM SIGCOMM, pages 267–278, 2006

    2006

  14. [14]

    F. Victor. Address clustering heuristics for Ethereum. InFinancial Cryptography and Data Security (FC), pages 617–633. Springer, 2020

    2020

  15. [15]

    Buterin, Z

    V. Buterin, Z. Hitzig, and E. G. Weyl. A flexible design for funding public goods.Management Science, 65(11):5171–5187, 2019

    2019

  16. [16]

    D. W. E. Allen, C. Berg, S. Davidson, and J. Potts. Why airdrop cryptocurrency tokens? Journal of Business Research, 163:113945, 2023

    2023

  17. [17]

    F. T. Liu, K. M. Ting, and Z.-H. Zhou. Isolation Forest. In2008 Eighth IEEE International Conference on Data Mining (ICDM), pages 413–422. IEEE, 2008

    2008

  18. [18]

    Sakurada and T

    M. Sakurada and T. Yairi. Anomaly detection using autoencoders with nonlinear dimension- ality reduction. InProc. MLSDA 2014 2nd Workshop on Machine Learning for Sensory Data Analysis, pages 4–11, 2014

    2014

  19. [19]

    Zhou and R

    C. Zhou and R. C. Paffenroth. Anomaly detection with robust deep autoencoders. InProc. 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 665–674, 2017

    2017

  20. [20]

    Chalapathy and S

    R. Chalapathy and S. Chawla. Deep learning for anomaly detection: A survey.arXiv preprint arXiv:1901.03407, 2019

    Pith/arXiv arXiv 1901

  21. [21]

    Dastile, T

    X. Dastile, T. Celik, and M. Potsane. Statistical and machine learning models in credit scoring: A systematic literature survey.Applied Soft Computing, 91:106263, 2020

    2020

  22. [22]

    Bücker, G

    M. Bücker, G. Szepannek, A. Gosiewska, and P. Biecek. Transparency, auditability, and ex- plainability of machine learning models in credit scoring.Journal of the Operational Research Society, 73(1):70–90, 2022. 19

    2022