pith. sign in

arxiv: 1907.09285 · v1 · pith:7RUPHTVWnew · submitted 2019-07-15 · 💻 cs.AI · cs.LG· cs.NE

ParaFIS:A new online fuzzy inference system based on parallel drift anticipation

Clement Leroy , Eric Anquetil , Nathalie Girard This is my paper

Pith reviewed 2026-05-24 21:47 UTC · model grok-4.3

classification 💻 cs.AI cs.LGcs.NE
keywords evolving fuzzy systemsconcept driftdata stream classificationonline learningfuzzy inferenceanticipation modulenon-stationary environmentsincremental adaptation
0
0 comments X

The pith

ParaFIS pairs a generalized evolving fuzzy system with a parallel anticipation module to speed adaptation after abrupt concept drifts.

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

The paper introduces an architecture for incremental fuzzy inference systems that targets sudden shifts in data distribution during online classification. It pairs a generalized evolving fuzzy system with an anticipation module that prepares new rules in parallel when a brutal drift occurs. Tests use three UCI datasets modified by injecting artificial abrupt drifts. A fit model measures reactivity time to reach steady-state performance and the final score. Results show gains over the same system without anticipation and over a comparable state-of-the-art evolving fuzzy system.

Core claim

The architecture improves adaptation after brutal concept drifts by adding a parallel anticipation module to a generalized evolving fuzzy system. On three UCI datasets with artificial drifts, the module reduces the time to converge to steady-state and raises the end-of-stream score relative to the baseline system without anticipation and to a similar existing EFS.

What carries the argument

The parallel drift anticipation module, which operates alongside the evolving fuzzy system to forecast abrupt distribution changes and accelerate new rule generation.

If this is right

  • The system reaches steady-state performance faster after abrupt drifts.
  • Final classification scores are higher on streams containing sudden changes.
  • Performance exceeds both the anticipation-free version and a comparable state-of-the-art EFS.
  • The fit model supplies a concrete way to quantify reactivity time and end-of-stream score.

Where Pith is reading between the lines

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

  • The parallel anticipation idea could be tested with other incremental learners such as decision-tree streams or neural online classifiers.
  • Experiments on streams with naturally occurring rather than injected drifts would be a direct next measurement.
  • If the module generalizes, it might reduce the frequency of full model resets in long-running streaming applications.

Load-bearing premise

Artificial brutal drifts injected into UCI datasets stand in for real-world abrupt concept drifts and the fit model isolates the anticipation module's contribution without confounding effects from dataset choice or parameters.

What would settle it

Apply the system to a real-world data stream containing naturally occurring abrupt drifts and check whether reactivity time and end scores still improve over the version without the anticipation module.

Figures

Figures reproduced from arXiv: 1907.09285 by Clement Leroy, Eric Anquetil, Nathalie Girard.

Figure 1
Figure 1. Figure 1: Problems met in generalized EFS when brutal drift [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Impact of a drift on the score in time D. Discussion on problems in the generalized EFS Two scales of adaptation co-exist in EFS, the adaptation of rule parameters and the structure adaptation. Smooth drifts are tackled by introducing forgetting capacity in the parameter adaptation whereas brutal drifts are tackled by the creation of new rules. However, the system is degraded by the parameter adaptation wh… view at source ↗
Figure 5
Figure 5. Figure 5: Protocol P - Generation of data stream with brutal drifts Dataset Classes Features Samples Scriptwriters Letters 26 16 20000 20 LaViola 48 50 16891 34 PenDigits 10 16 10992 44 TABLE I: Information on the datasets used in the experiments on it. In this way, all the data are used to test the system and then, to train it while maintaining independence between each phase. Scores are then averaged over a certai… view at source ↗
Figure 6
Figure 6. Figure 6: Fit of the three phases score using Eq. (15) [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Prequential score (y-axis) with respect to data (x-axis) [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
read the original abstract

This paper proposes a new architecture of incremen-tal fuzzy inference system (also called Evolving Fuzzy System-EFS). In the context of classifying data stream in non stationary environment, concept drifts problems must be addressed. Several studies have shown that EFS can deal with such environment thanks to their high structural flexibility. These EFS perform well with smooth drift (or incremental drift). The new architecture we propose is focused on improving the processing of brutal changes in the data distribution (often called brutal concept drift). More precisely, a generalized EFS is paired with a module of anticipation to improve the adaptation of new rules after a brutal drift. The proposed architecture is evaluated on three datasets from UCI repository where artificial brutal drifts have been applied. A fit model is also proposed to get a "reactivity time" needed to converge to the steady-state and the score at end. Both characteristics are compared between the same system with and without anticipation and with a similar EFS from state-of-the-art. The experiments demonstrates improvements in both cases.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes ParaFIS, a new architecture for evolving fuzzy systems (EFS) that pairs a generalized EFS with a parallel drift anticipation module to improve adaptation after brutal concept drifts in non-stationary data streams. Evaluation is performed on three UCI datasets with artificially injected brutal drifts; a fit model is introduced to quantify reactivity time to steady-state and end score. These metrics are compared against the same EFS without the anticipation module and against a similar state-of-the-art EFS, with the abstract claiming improvements in both reactivity and final performance.

Significance. If the central claim holds under more rigorous validation, the work could advance handling of abrupt drifts in online fuzzy systems, an area where existing EFS are noted to perform better on smooth drifts. The ablation-style comparison (with vs. without anticipation) and external baseline provide independent grounding, and the fit model for measuring reactivity time is a constructive methodological contribution for assessing adaptation speed in streaming settings.

major comments (2)
  1. [Abstract] Abstract: the claim that 'the experiments demonstrates improvements in both cases' is unsupported by any quantitative results, statistical tests, error bars, implementation details, or reported values for reactivity time and end score. This is load-bearing for the central claim of better adaptation attributable to the anticipation module.
  2. [Evaluation] Evaluation section: the fit model is used to quantify gains from the anticipation module on artificial brutal drifts injected into UCI datasets, yet no sensitivity checks on drift magnitude/timing or ablation that holds all other EFS components fixed while toggling only the anticipation path are reported. This leaves the attribution of reactivity and score improvements vulnerable to confounding from dataset choice or drift-injection artifacts.
minor comments (2)
  1. [Abstract] Typo: 'incremen-tal' should read 'incremental'.
  2. [Abstract] Grammar: 'demonstrates' should be 'demonstrate' to agree with plural subject 'experiments'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We provide point-by-point responses to the major comments and indicate planned revisions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that 'the experiments demonstrates improvements in both cases' is unsupported by any quantitative results, statistical tests, error bars, implementation details, or reported values for reactivity time and end score. This is load-bearing for the central claim of better adaptation attributable to the anticipation module.

    Authors: We agree that the abstract would benefit from including quantitative results to support the claim. The manuscript's evaluation section provides the detailed comparisons, including reactivity times and end scores for the system with and without the anticipation module, as well as against the state-of-the-art EFS. We will revise the abstract to report specific quantitative improvements observed in the experiments. revision: yes

  2. Referee: [Evaluation] Evaluation section: the fit model is used to quantify gains from the anticipation module on artificial brutal drifts injected into UCI datasets, yet no sensitivity checks on drift magnitude/timing or ablation that holds all other EFS components fixed while toggling only the anticipation path are reported. This leaves the attribution of reactivity and score improvements vulnerable to confounding from dataset choice or drift-injection artifacts.

    Authors: The comparison between the system with and without the anticipation module, while holding other components fixed, constitutes the requested ablation study and supports attribution to the anticipation module. However, sensitivity checks on drift magnitude and timing were not performed. We will add these analyses in the revision to address potential confounding factors. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical ablation and external baseline provide independent grounding

full rationale

The paper describes an incremental EFS architecture paired with an anticipation module, evaluated via direct comparison on UCI datasets with injected artificial drifts. It reports reactivity time and end-score metrics obtained from a fit model, but these are post-experiment quantifications rather than derivations that reduce to the inputs by construction. The central claims rest on ablation (with vs. without anticipation) plus comparison to an external state-of-the-art EFS; no equations, uniqueness theorems, self-citations, or ansatzes are invoked in the supplied text. The evaluation design therefore remains self-contained against external benchmarks and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies no explicit free parameters, axioms, or invented entities; the fit model for reactivity time is mentioned but not detailed enough to identify fitted values or assumptions.

pith-pipeline@v0.9.0 · 5710 in / 1156 out tokens · 23117 ms · 2026-05-24T21:47:54.467009+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

  1. [1]

    Bouillon and E

    M. Bouillon and E. Anquetil. Online active supervision of an evolving classifier for customized-gesture-command learning. Neurocomputing, 262:77–89, Nov 2017

    work page 2017

  2. [2]

    Lughofer, M

    E. Lughofer, M. Pratama, and I. Skrjanc. Incremental rule splitting in generalized evolving fuzzy systems for autonomous drift compensation. IEEE Transactions on Fuzzy Systems , 26(4):1854–1865, Aug 2018

    work page 2018

  3. [3]

    Angelov and R

    P. Angelov and R. Yager. Simplified fuzzy rule-based systems using non- parametric antecedents and relative data density. 2011 IEEE Workshop on Evolving and Adaptive Intelligent Systems (EAIS), pages 62–69, 2011

    work page 2011

  4. [4]

    A. Liu, G. Zhang, and J. Lu. Fuzzy time windowing for gradual concept drift adaptation. 2017 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pages 1–6, Jul 2017

    work page 2017

  5. [5]

    Bouillon, E

    M. Bouillon, E. Anquetil, and A. A. Almaksour. Decremental learning of evolving fuzzy inference systems using a sliding window. InProceedings of the 2012 11th International Conference on Machine Learning and Applications - Volume 01 , ICMLA ’12, pages 598–601, Washington, DC, USA, 2012. IEEE Computer Society

    work page 2012

  6. [6]

    Y . Song, G. Zhang, H. Lu, and J. Lu. A self-adaptive fuzzy network for prediction in non-stationary environments. 2018 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) , pages 1–8, 2018

    work page 2018

  7. [7]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio. Generative adversarial nets. In Advances in Neural Information Processing Systems 27 , pages 2672–

    work page

  8. [8]

    Curran Associates, Inc., 2014

    work page 2014

  9. [9]

    A. A. Almaksour and E. Anquetil. Improving premise structure in evolv- ing takagi–sugeno neuro-fuzzy classifiers. Evolving Systems, 2(1):25–33, Mar 2011

    work page 2011

  10. [10]

    Lemos, W

    A. Lemos, W. Caminhas, and F. Gomide. Multivariable gaussian evolving fuzzy modeling system. IEEE Transactions on Fuzzy Systems , 19(1):91–104, Feb 2011

    work page 2011

  11. [11]

    P. P. Angelov and D. P. Filev. An approach to online identification of takagi-sugeno fuzzy models. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics) , 34(1):484–498, Feb 2004

    work page 2004

  12. [12]

    J. Gama, R. Sebasti ˜ao, and P. P. Rodrigues. On evaluating stream learning algorithms. Machine Learning, 90(3):317–346, Mar 2013

    work page 2013

  13. [13]

    Alimoglu

    F. Alimoglu. Combining Multiple Classifiers for Pen-Based Handwritten Digit Recognition. PhD thesis, Bogazici Univeristy, 1996

    work page 1996

  14. [14]

    P. W. Frey and D. J. Slate. Letter recognition using Holland-style adaptive classifiers. Mach. Learn., 6(2):161–182, Mar 1991

    work page 1991

  15. [15]

    J. J. LaViola and R. C. Zeleznik. A practical approach for writer- dependent symbol recognition using a writer-independent symbol rec- ognizers, November 2007

    work page 2007

  16. [16]

    Dheeru and E

    D. Dheeru and E. Karra Taniskidou. UCI machine learning repository, 2017

    work page 2017