Pith. sign in

REVIEW 2 major objections 5 minor 1 cited by

A Neuro-Symbolic Framework for Sequence Classification with Relational and Temporal Knowledge

T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read In knowledge-driven sequence classification that combines relational constraints with LTLf temporal specifications, purely neural pipelines fail to learn the sequence-level decision rule, and state-of-the-art neuro-symbolic components…

desk verdict A genuinely useful benchmark with an overstated neural-failure claim: Q1 never trains from sequence labels alone, so the paper's central negative result is broader than the evidence. read the letter →

arxiv 2505.05106 v1 pith:UZYPPKXQ submitted 2025-05-08 cs.AI

Luca Salvatore Lorello , Marco Lippi , Stefano Melacci This is my paper · ORCID
classification cs.AI
keywords neuro-symbolicAIsequenceclassificationLTLftemporallogicrelationalknowledgeautomata-basedreasoningknowledge-drivenlearningbenchmarkingframework
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that sequence classification driven by relational and temporal domain knowledge is a hard, under-explored setting in which neither purely neural nor naively stacked neuro-symbolic architectures succeed. It introduces LTLZinc, a benchmark generator that turns LTLf formulas and finite-domain constraints into labelled image sequences, and evaluates a modular pipeline with neural and symbolic components for constraint checking and next-state prediction. The authors claim that no neural-only pipeline learns the sequence rule from sequence labels, that symbolic components at both stages give the best accuracy trade-off, and that temporal reasoners extended to relational inputs can fail even with full background knowledge. A sympathetic reader would care because the setting matches safety-critical applications such as autonomous driving and industrial control, where decisions depend on relations at each step and on the order of events over time.

What carries the argument

The carrying object is the constraint-annotated deterministic finite automaton: LTLZinc compiles the LTLf temporal formula into a DFA whose transition guards are finite-domain constraint satisfaction problems, then generates positive and negative sequences by random walks along the automaton. Inference is decoupled into a relational stage, which checks per-timestep constraints, and a temporal stage, which predicts the automaton state from the constraint-validity trace. Symbolic temporal variants evaluate the automaton's transition formulas by algebraic model counting over probability or log-probability semirings, implemented either with fuzzy-logic automata or with formulas compiled to sd-DNNF, a knowledge compilation target that supports polynomial-time model counting.

What would settle it

Take one of the six task specifications and train a purely neural pipeline on a much larger generated dataset, with thousands of sequences and sequence-level labels only; if its sequence accuracy exceeds the best symbolic-symbolic configuration, the claim that neural-only pipelines fail to learn the decision rule is false. Alternatively, construct a real-world relational-temporal sequence dataset with ground-truth labels and show that a neural-only model matches or beats the symbolic-symbolic pipeline, which would challenge the generality of the benchmark's conclusions.

Watch

Extended reading notes

Core claim

On six generated tasks, the paper reports that purely neural pipelines, despite strong image classification, plateau near random or well below ceiling on sequence-level accuracy, so the sequence-level decision rule is not learned from sequence labels alone. When the pipeline is built from symbolic modules, the best configuration uses a symbolic constraint checker and a symbolic automaton-based next-state predictor, giving the best constraint-versus-temporal accuracy trade-off on every task. The paper also identifies two failure modes: temporal reasoners that excel on propositional inputs can fail when inputs are relational, even with perfect background knowledge and oracle labels, and first-order relational reasoners show training instabilities when recurrently stacked, with variance driven by architecture rather than initialization. Oracle experiments show that a confidently wrong classifier can beat a reluctantly correct one, meaning uncertainty can accumulate adversely over reasoning steps.

Load-bearing premise

The load-bearing premise is that the six generated tasks, built from handwritten-digit and fashion-item images with hand-written constraints and temporal formulas, are representative of the general relational-temporal sequence classification problem; if these tasks are too easy, too contrived, or too sensitive to the random-walk generator, the reported failure modes may not transfer to real applications.

Editorial extensions

If this is right

  • Combining symbolic modules at both the constraint and temporal stages yields the best accuracy trade-off on all six tasks, so the paper's evidence favors decoupled symbolic-symbolic pipelines over neural-only or mixed stacks.
  • Neural-only methods cannot recover the sequence-level rule from sequence labels alone in this setting; the sequence label is predicted at or near chance for several tasks even when image classification accuracy exceeds 0.9.
  • Temporal reasoners known to work on propositional traces are not robust to relational inputs: they can fail even when given full background knowledge, so relational-temporal integration needs dedicated treatment rather than simple propositionalization.
  • Relational reasoners exhibit training instability in recurrent settings, and convergence depends on architectural choices rather than weight initialization, so multi-run reporting is essential for reliable conclusions.
  • The oracle experiments imply that wrong-but-confident upstream predictions can outperform correct-but-uncertain ones, making calibration and uncertainty handling first-order concerns in multi-stage neuro-symbolic reasoning.

Reading between the lines

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

  • If these failure modes are general, practical neuro-symbolic systems for safety-critical temporal decision-making should keep symbolic constraint solvers and symbolic automata in the loop, using neural components only at perception, and should avoid end-to-end gradient flow through stacked soft reasoners.
  • A testable extension would vary the temporal horizon and the arity of constraints in generated tasks; the paper's results predict a sharp cliff where neural next-state prediction falls to chance as the DFA grows, while symbolic-stack variance increases.
  • The confidently-wrong-beats-reluctantly-correct result suggests that training objectives penalizing overconfidence on the sequence label, or calibration-aware losses, could improve downstream temporal accuracy.
  • LTLZinc's incremental generation mode could be used to stage curriculum learning over formulas of increasing temporal depth, potentially mitigating the training instability the paper observes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper introduces LTLZinc, a benchmarking framework for sequence classification under relational and temporal background knowledge, where relational constraints are expressed as finite-domain constraints and temporal specifications as LTLf formulas. The authors define a multi-stage neuro-symbolic pipeline (IC, CC, NSP, SC), instantiate each stage with neural or symbolic components, and evaluate the resulting combinations on six synthetic tasks built from MNIST and Fashion-MNIST. The main empirical findings are that neural-only pipelines struggle on these tasks, that combining symbolic modules for both constraint classification and next-state prediction gives the best trade-off, and that stacking state-of-the-art neuro-symbolic components exposes training instabilities and failure modes. The paper also studies the effect of upstream label noise using oracle predictors.

Significance. The LTLZinc framework, the released generator, datasets, and code, and the extensive experimentation (2,232 runs) are concrete contributions that could serve as a useful testbed for neuro-symbolic temporal reasoning. The observation that symbolic modules are not automatically robust when stacked—temporal reasoners can fail even with full background knowledge, and relational reasoners show training instabilities—is a valuable reference for future research. However, the central negative claim about purely neural pipelines is not established by the experiments as reported, and the headline comparison in Table 1 is weakened by a validation-based selection across many hyperparameter configurations. These issues need to be addressed before the paper's main conclusions can be accepted.

major comments (2)
  1. [Section 6 (Q1) and Appendix B.5] The Q1 experiments provide dense supervision at every stage (λ_IC = λ_CC = λ_NSP = λ_SC = 1.0), so the conclusion that "no task can be effectively learned with sequence-level labels alone" is not supported by the evidence. No condition removes the intermediate IC, CC, and NSP supervision; therefore the poor sequence-level accuracy could be an artifact of the modular architecture, the auxiliary loss weighting, or the optimization, rather than a general limitation of neural sequence classifiers trained from sequence labels. Since the motivation in Section 1 and the conclusions in Section 7 rely on this neural failure, an end-to-end baseline trained only on the sequence-level loss (for example, a GRU or transformer over image-derived features) should be added before this claim can be made.
  2. [Table 1 and Section 6.1] The headline results select, for each task, the best configuration out of 96 hyperparameter combinations based on validation-set average accuracy across 3 seeds. This model-selection procedure can inflate the reported test accuracies, and the number of configurations is not balanced across the four compared families (for instance, Neural-Neural has more CC×NSP choices than Symbolic-Symbolic). The claim that the Symbolic-Symbolic family "achieves the best trade-off for every task considered" should be supported by reporting the full distribution of results across hyperparameters (as in Fig. 2) or by using a nested validation protocol; otherwise the apparent dominance may be an artifact of selection. Please provide mean and standard deviation across the selected models or apply a proper multiple-comparison correction.
minor comments (5)
  1. [Throughout (e.g., Section 5, Section 6, Appendix)] Many cross-references are unresolved in the compiled manuscript ("Appendix ??", "Figures ??, ??, ??"), making it difficult to locate the supporting material; please fix these references before publication.
  2. [Table 1] The markers "∗", "†", and the bracketed sample run are explained only partially in the caption; add a complete legend that also explains what the combined markers (e.g., "†*") mean.
  3. [Section 6.2] The claim that the Scallop module performs "near-exact inference" is imprecise: with the default top-1 proof provenance, Scallop performs approximate inference. Please rephrase or justify, since the interpretation of the oracle experiments depends on this point.
  4. [Appendix B.4-B.6] The total experiment count (2,232) is not obviously reconciled with the per-experiment counts (1,728 main, 36 Q1, 468 Q3); please make the arithmetic explicit to aid reproducibility.
  5. [Section 6.1] The sentence "Exploiting a symbolic component both for CC and NSP allows to achieve the best trade-off for every task considered" is broader than the evidence; consider changing "every task considered" to "all six tasks considered" to avoid overgeneralization.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the symbolic modules receive the same user-specified knowledge used to generate benchmark labels, but this is the intended knowledge-injection setup, not a fitted prediction renamed as a result.

full rationale

The paper's derivation chain is an empirical comparison, not a derivation of a result from fitted inputs. LTLZinc generates sequences and labels from the user-supplied quadruple (X,Y,C,F), and the symbolic CC/NSP modules implement the same C and F (Datalog program and DFA, respectively). This is deliberate knowledge injection: the binary label is defined as S |= T with T = C ∪ F, so a symbolic module that executes T is expected to align with the generator. This does not make the comparison circular because the empirical content lies in whether the stacked modules preserve performance under noisy perceptual inputs (IC accuracy around 0.9), calibration choices, and optimization stability, not in a fitted constant being renamed as a prediction. No parameter is fitted to the claimed outcome: temperature calibration and pre-training are standard hyper-parameters; the semantic loss is compared against binary cross-entropy. The cited implementations (Umili et al. 2023; Manginas et al. 2024) are external prior work, not self-citations, and no uniqueness theorem is imported from the authors. One caveat is correctness, not circularity: the Q1 conclusion that 'no task can be effectively learned with sequence-level labels alone' is not established by the Q1 experiment, since Section 6 states supervision is provided at every level with lambda_IC = lambda_CC = lambda_NSP = lambda_SC = 1.0. That omission of an end-to-end sequence-label-only baseline is an inference gap, but it does not reduce any claim to its inputs by construction.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

This paper introduces no fitted constants and no new theoretical entities. The central claims rest on standard results (LTLf-to-DFA equivalence), on the correctness of the MiniZinc and Scallop encodings, and on the representativeness of the generated tasks. The released code provides grounds to check the tooling assumptions.

assumptions (4)
  • standard math LTLf formulas are equivalent to deterministic finite automata over finite traces.
    Invoked in Section 2 and used throughout the LTLZinc generator and the NSP modules; the paper relies on flfloat for the conversion and on the DFA equivalence theorem from De Giacomo and Vardi (2013).
  • domain assumption MiniZinc global constraints correctly encode the relational properties used to generate and label the datasets.
    Section 4 states the generator solves MiniZinc programs for each transition guard; the ground-truth labels and the sequence acceptance labels depend on the solver returning correct constraint solutions.
  • domain assumption The neuro-symbolic reasoning engines (Scallop, fuzzy automata, sd-DNNF automata) are faithful implementations of the cited methods.
    Section 5 and Appendix B describe the modules as following Umili et al. (2023) and Manginas et al. (2024), with one stated modification for sd-DNNF; any discrepancy in these implementations would affect the empirical comparisons and the reported instabilities.
  • domain assumption The six generated tasks are representative of the relational-temporal sequence classification setting.
    Section 6 evaluates only tasks assembled from MNIST and Fashion-MNIST with hand-designed constraints and formulas; the generality of the conclusions about this setting depends on this representativeness.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Neuro-Symbolic Framework for Sequence Classification with Relational and Temporal Knowledge." pith.science (2026). https://pith.science/paper/UZYPPKXQ

@misc{pith2026250505106,
  author       = {Pith},
  title        = {Pith review of: A Neuro-Symbolic Framework for Sequence Classification with Relational and Temporal Knowledge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UZYPPKXQ}},
  note         = {Machine review of arXiv:2505.05106}
}
read the original abstract

One of the goals of neuro-symbolic artificial intelligence is to exploit background knowledge to improve the performance of learning tasks. However, most of the existing frameworks focus on the simplified scenario where knowledge does not change over time and does not cover the temporal dimension. In this work we consider the much more challenging problem of knowledge-driven sequence classification where different portions of knowledge must be employed at different timesteps, and temporal relations are available. Our experimental evaluation compares multi-stage neuro-symbolic and neural-only architectures, and it is conducted on a newly-introduced benchmarking framework. Results demonstrate the challenging nature of this novel setting, and also highlight under-explored shortcomings of neuro-symbolic methods, representing a precious reference for future research.

Figures

Figures reproduced from arXiv: 2505.05106 by the authors.

Figure 1
Figure 1. Stages of our architecture. We implemented CC and NSP in mul￾tiple ways, that we listed by means of the green (symbolic) and reddish (neural) dashed-borders sub-blocks. Task Config. CC-NSP Best Model CC NSP (epoch) Avg Acc. ↑ IC acc. ↑ CC Acc. ↑ NSP Acc. ↑ SC Acc. ↑ Task 1 Neural-Neural MLP-L MLP-L (24) 0.84 0.88 0.85 ± 0.01 0.72 0.90 Task 1 Neural-Symbolic MLP-L sd-DNNF-LP (24)* 0.85 0.86 0.85 0.81 0.90 Task 1 Symb… view at source ↗
Figure 2
Figure 2. Q2. CC-NSP accuracy trade-off for different families of architectures (i.e., Neural/Symbolic CC, Neural/Symbolic NSP, where Neural NSP is indicated either by MLP or GRU, to compare them as well). The horizontal dashed line indicates the performance of a deterministic baseline of NSP always choosing the successor state most represented in the training set. labels alone. We experienced slow convergence and sub￾optimal… view at source ↗
Figure 3
Figure 3. Q3. Accuracies for Task 4 with oracular predictors. Oracle types (perfect, flip, confidence) are described in Section 6. with the only exception of task 5, where it performs slightly worse than the Symbolic-Neural family. When observing constraint accuracy alone, a downstream Symbolic NSP mod￾ule often negatively affects performance, with task 4 pre￾senting the highest delta between Symbolic-Symbolic and Symbolic-Ne… view at source ↗
Figures from the paper (5 more)
Figure 1
Figure 1. Figure 1: Neural-only experiments (Q1). Task Automaton module Avg acc. ↑ Label acc. ↑ Constraint acc. ↑ Successor acc. ↑ Sequence Acc. ↑ MP Successor MP Sequence Task 1 MLP-L 0.89 0.91 ± 0.01 0.90 ± 0.01 0.83 ± 0.01 0.90 0.72 0.90 Task 1 GRU-L 0.88 0.91 0.90 ± 0.01 0.81 ± 0.01 0…
Figure 2
Figure 2. Figure 2: Average accuracy when replacing IC or IC/CC modules with an oracular predictor ( [PITH_FULL_IMAGE:figures/full_fig_p015_2.png]
Figure 3
Figure 3. Figure 3: CC accuracy when replacing IC or IC/CC modules with an oracular predictor ( [PITH_FULL_IMAGE:figures/full_fig_p016_3.png]
Figure 4
Figure 4. Figure 4: NSP accuracy when replacing IC or IC/CC modules with an oracular predictor ( [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 5
Figure 5. Figure 5: Accuracies for Task 4 with oracular predictors. These are the same data points as [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. LTLZinc: a Benchmarking Framework for Continual Learning and Neuro-Symbolic Temporal Reasoning

    cs.AI 2025-07 conditional novelty 7.0 of 10

    LTLZinc generates image-based temporal reasoning and continual learning benchmarks from LTLf formulas over MiniZinc constraints, and experiments show existing methods often fail.

Reference graph

Works this paper leans on

42 extracted references · 28 canonical work pages · cited by 1 Pith paper

  1. [1]

    Inference of reversible languages

    Dana Angluin. Inference of reversible languages. Journal of the ACM (JACM) , 29(3):741--765, 1982

    work page 1982

  2. [2]

    Logic tensor networks

    Samy Badreddine, Artur d'Avila Garcez, Luciano Serafini, and Michael Spranger. Logic tensor networks. Artificial Intelligence , 303:103649, 2022

    2022

  3. [3]

    Neural-symbolic learning and reasoning: A survey and interpretation 1

    Tarek R Besold, Artur d’Avila Garcez, Sebastian Bader, Howard Bowman, Pedro Domingos, Pascal Hitzler, Kai-Uwe K \"u hnberger, Luis C Lamb, Priscila Machado Vieira Lima, Leo de Penning, et al. Neural-symbolic learning and reasoning: A survey and interpretation 1. In Neuro-Symbolic Artificial Intelligence: The State of the Art , pages 1--51. IOS press, 2021

    work page 2021

  4. [4]

    Learning interpretable models expressed in linear temporal logic

    Alberto Camacho and Sheila A McIlraith. Learning interpretable models expressed in linear temporal logic. In Proceedings of the International Conference on Automated Planning and Scheduling , volume 29, pages 621--630, 2019

    work page 2019

  5. [5]

    Device classification for industrial control systems using predicted traffic features

    Indrasis Chakraborty, Brian M Kelley, and Brian Gallagher. Device classification for industrial control systems using predicted traffic features. Frontiers in Computer Science , 4:777089, 2022

    work page 2022

  6. [6]

    On probabilistic inference by weighted model counting

    Mark Chavira and Adnan Darwiche. On probabilistic inference by weighted model counting. Artificial Intelligence , 172(6-7):772--799, 2008

    2008

  7. [7]

    A knowledge compilation map

    Adnan Darwiche and Pierre Marquis. A knowledge compilation map. Journal of Artificial Intelligence Research , 17:229--264, 2002

    2002

  8. [8]

    Linear temporal logic and linear dynamic logic on finite traces

    Giuseppe De Giacomo and Moshe Y Vardi. Linear temporal logic and linear dynamic logic on finite traces. In Ijcai , volume 13, pages 854--860, 2013

    work page 2013

Show all 42 references
  1. [9]

    Statistical relational artificial intelligence: Logic, probability, and computation

    Luc De Raedt, Kristian Kersting, Sriraam Natarajan, and David Poole. Statistical relational artificial intelligence: Logic, probability, and computation. Synthesis lectures on artificial intelligence and machine learning , 10(2):1--189, 2016

    2016

  2. [10]

    The power of symbolic automata and transducers

    Loris D’Antoni and Margus Veanes. The power of symbolic automata and transducers. In Computer Aided Verification: 29th International Conference, CAV 2017, Heidelberg, Germany, July 24-28, 2017, Proceedings, Part I 30 , pages 47--67. Springer, 2017

    2017

  3. [11]

    Learning regular expressions from noisy sequences

    Ugo Galassi and Attilio Giordana. Learning regular expressions from noisy sequences. In International Symposium on Abstraction, Reformulation, and Approximation , pages 92--106. Springer, 2005

    2005

  4. [12]

    Foundations of support constraint machines

    Giorgio Gnecco, Marco Gori, Stefano Melacci, and Marcello Sanguineti. Foundations of support constraint machines. Neural computation , 27(2):388--480, 2015

    2015

  5. [13]

    Provenance semirings

    Todd J Green, Grigoris Karvounarakis, and Val Tannen. Provenance semirings. In Proceedings of the twenty-sixth ACM SIGMOD-SIGACT-SIGART symposium on Principles of database systems , pages 31--40, 2007

    2007

  6. [14]

    A comprehensive explanation framework for biomedical time series classification

    Praharsh Ivaturi, Matteo Gadaleta, Amitabh C Pandey, Michael Pazzani, Steven R Steinhubl, and Giorgio Quer. A comprehensive explanation framework for biomedical time series classification. IEEE journal of biomedical and health informatics , 25(7):2398--2408, 2021

    2021

  7. [15]

    Algebraic model counting

    Angelika Kimmig, Guy Van den Broeck, and Luc De Raedt. Algebraic model counting. Journal of Applied Logic , 22:46--62, 2017

    2017

  8. [16]

    The mnist database of handwritten digits

    Yann LeCun. The mnist database of handwritten digits. http://yann. lecun. com/exdb/mnist/ , 1998

    1998

  9. [17]

    Scallop: A language for neurosymbolic programming

    Ziyang Li, Jiani Huang, and Mayur Naik. Scallop: A language for neurosymbolic programming. Proceedings of the ACM on Programming Languages , 7(PLDI):1463--1487, 2023

    2023

  10. [18]

    Checking ltl satisfiability via end-to-end learning

    Weilin Luo, Hai Wan, Delong Zhang, Jianfeng Du, and Hengdi Su. Checking ltl satisfiability via end-to-end learning. In Proceedings of the 37th IEEE/ACM International Conference on Automated Software Engineering , pages 1--13, 2022

    2022

  11. [19]

    Learning to check ltl satisfiability and to generate traces via differentiable trace checking

    Weilin Luo, Pingjia Liang, Junming Qiu, Polong Chen, Hai Wan, Jianfeng Du, and Weiyuan Fang. Learning to check ltl satisfiability and to generate traces via differentiable trace checking. In Proceedings of the 33rd ACM SIGSOFT International Symposium on Software Testing and An...

    2024

  12. [20]

    On the hardness of probabilistic neurosymbolic learning

    Jaron Maene, Vincent Derkinderen, and Luc De Raedt. On the hardness of probabilistic neurosymbolic learning. arXiv preprint arXiv:2406.04472 , 2024

    2024 arXiv

  13. [21]

    Nesya: Neurosymbolic automata

    Nikolaos Manginas, George Paliouras, and Luc De Raedt. Nesya: Neurosymbolic automata. arXiv preprint arXiv:2412.07331 , 2024

    2024 arXiv

  14. [22]

    Deepproblog: Neural probabilistic logic programming

    Robin Manhaeve, Sebastijan Dumancic, Angelika Kimmig, Thomas Demeester, and Luc De Raedt. Deepproblog: Neural probabilistic logic programming. Advances in neural information processing systems , 31, 2018

    2018

  15. [23]

    Approximate inference for neural probabilistic logic programming

    Robin Manhaeve, Giuseppe Marra, and Luc De Raedt. Approximate inference for neural probabilistic logic programming. In KR , pages 475--486, 2021

    2021

  16. [24]

    From statistical relational to neurosymbolic artificial intelligence: A survey

    Giuseppe Marra, Sebastijan Duman c i \'c , Robin Manhaeve, and Luc De Raedt. From statistical relational to neurosymbolic artificial intelligence: A survey. Artificial Intelligence , page 104062, 2024

    2024

  17. [25]

    Octal: Graph representation learning for ltl model checking

    Prasita Mukherjee, Haoteng Yin, Susheel Suresh, and Tiark Rompf. Octal: Graph representation learning for ltl model checking. arXiv preprint arXiv:2207.11649 , 2022

    2022 arXiv

  18. [26]

    Minizinc: Towards a standard cp modelling language

    Nicholas Nethercote, Peter J Stuckey, Ralph Becket, Sebastian Brand, Gregory J Duck, and Guido Tack. Minizinc: Towards a standard cp modelling language. In International Conference on Principles and Practice of Constraint Programming , pages 529--543. Springer, 2007

    2007

  19. [27]

    Reading digits in natural images with unsupervised feature learning

    Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Baolin Wu, Andrew Y Ng, et al. Reading digits in natural images with unsupervised feature learning. In NIPS workshop on deep learning and unsupervised feature learning , volume 2011, page 4. Granada, 2011

    2011

  20. [28]

    The temporal logic of programs

    Amir Pnueli. The temporal logic of programs. In 18th annual symposium on foundations of computer science (sfcs 1977) , pages 46--57. ieee, 1977

    1977

  21. [29]

    A scenario-based assessment approach for automated driving by using time series classification of human-driving behaviour

    Christian Roesener, Felix Fahrenkrog, Axel Uhlig, and Lutz Eckstein. A scenario-based assessment approach for automated driving by using time series classification of human-driving behaviour. In 2016 IEEE 19th international conference on intelligent transportation systems (ITS...

    2016

  22. [30]

    Ltl satisfiability checking

    Kristin Y Rozier and Moshe Y Vardi. Ltl satisfiability checking. In International SPIN Workshop on Model Checking of Software , pages 149--167. Springer, 2007

    2007

  23. [31]

    Deepdfa: Automata learning through neural probabilistic relaxations

    Elena Umili and Roberto Capobianco. Deepdfa: Automata learning through neural probabilistic relaxations. In ECAI 2024 , pages 1051--1058. Ios Press, 2024

    2024

  24. [32]

    Grounding ltlf specifications in image sequences

    Elena Umili, Roberto Capobianco, and Giuseppe De Giacomo. Grounding ltlf specifications in image sequences. In Proceedings of the International Conference on Principles of Knowledge Representation and Reasoning , volume 19, pages 668--678, 2023

    2023

  25. [33]

    Analyzing differentiable fuzzy logic operators

    Emile van Krieken, Erman Acar, and Frank van Harmelen. Analyzing differentiable fuzzy logic operators. Artificial Intelligence , 302:103602, 2022

    2022

  26. [34]

    On the independence assumption in neurosymbolic learning

    Emile van Krieken, Pasquale Minervini, Edoardo M Ponti, and Antonio Vergari. On the independence assumption in neurosymbolic learning. arXiv preprint arXiv:2404.08458 , 2024

    2024 arXiv

  27. [35]

    Symbolic automata constraint solving

    Margus Veanes, Nikolaj Bj rner, and Leonardo De Moura. Symbolic automata constraint solving. In Logic for Programming, Artificial Intelligence, and Reasoning: 17th International Conference, LPAR-17, Yogyakarta, Indonesia, October 10-15, 2010. Proceedings 17 , pages 640--654. S...

    2010

  28. [36]

    Applications of symbolic finite automata

    Margus Veanes. Applications of symbolic finite automata. In Implementation and Application of Automata: 18th International Conference, CIAA 2013, Halifax, NS, Canada, July 16-19, 2013. Proceedings 18 , pages 16--23. Springer, 2013

    2013

  29. [37]

    Learning finite linear temporal logic specifications with a specialized neural operator

    Homer Walke, Daniel Ritter, Carl Trimbach, and Michael Littman. Learning finite linear temporal logic specifications with a specialized neural operator. arXiv preprint arXiv:2111.04147 , 2021

    2021 arXiv

  30. [38]

    A comprehensive survey of continual learning: theory, method and application

    Liyuan Wang, Xingxing Zhang, Hang Su, and Jun Zhu. A comprehensive survey of continual learning: theory, method and application. IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024

    2024

  31. [39]

    Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms, 2017

    Han Xiao, Kashif Rasul, and Roland Vollgraf. Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms, 2017

    2017

  32. [40]

    Embedding symbolic temporal knowledge into deep sequential models

    Yaqi Xie, Fan Zhou, and Harold Soh. Embedding symbolic temporal knowledge into deep sequential models. In 2021 IEEE International Conference on Robotics and Automation (ICRA) , pages 4267--4273. IEEE, 2021

    2021

  33. [41]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence '...

  34. [42]

    Sympy: symbolic computing in python

    Aaron Meurer, Christopher P Smith, Mateusz Paprocki, Ond r ej C ert \' k, Sergey B Kirpichev, Matthew Rocklin, AMiT Kumar, Sergiu Ivanov, Jason K Moore, Sartaj Singh, et al. Sympy: symbolic computing in python. PeerJ Computer Science , 3:e103, 2017

    2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.