Pith. sign in

REVIEW 4 major objections 4 minor 58 references

Compositionality in Time Series: A Proof of Concept using Symbolic Dynamics and Compositional Data Augmentation

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The authors aim to establish that clinical time series are generated by composing a finite set of physiological states, and that data recombined from those states are interchangeable with real patient data in training and evaluation.

desk verdict The empirical pipeline is careful and the synthetic data result is real, but the central inference from risk ratios to compositionality is broken; the paper should be used for its method, not its conclusion. read the letter →

arxiv 2508.20656 v1 pith:NNFI5SYY submitted 2025-08-28 cs.LG

Michael Hagmann , Michael Staniek , Stefan Riezler This is my paper
classification cs.LG
keywords compositionalityclinicaltimeseriessymbolicdynamicscompositionaldataaugmentationdomainadaptationexpectedriskSOFAscore
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 the data-generation process behind multivariate clinical time series is compositional: patient trajectories can be modelled as ordered sequences of a finite set of latent physiological states, where each state emits a block of clinical measurements and the clinical meaning of the whole sequence is a function of its parts and their order. If true, the elementary states and composition rules can be reconstructed from data alone, and newly composed trajectories can serve as faithful proxies for real patient data in training and testing. The authors test this with two empirical criteria borrowed from domain-adaptation theory: one compares the expected forecasting risk of a model trained on synthesized versus original data, the other compares expected risks evaluated on synthesized versus original test data. On MIMIC-III and eICU, compositionally synthesized data produce near-zero risk differences and risk ratios close to one, while randomization-based augmentation does not, and enlarging the synthesized training set improves downstream SOFA-score prediction. The proof-of-concept conclusion is that clinical time series are compositionally generated, making compositional data synthesis a viable route around sparse and low-resource medical time series.

What carries the argument

The load-bearing machinery is the algebraic definition of compositional data generation, a homomorphism between a latent state structure and an observation structure, made operational by three concrete components. First, a symbol space built by clustering 3-hour measurement blocks, using k-means on random input centroids or on neural hidden-state representations, turns high-dimensional clinical time series into discrete symbol chains. Second, a distributional compositional data-synthesization algorithm finds interchangeable fragments by shared contexts, called environments, and recombines them in new templates through an insert operation. Third, Theorem 1, an equivalence between pointwise bo

What would settle it

Train several architecturally different forecasting models, such as an LSTM, a linear model, and a Transformer, on the original data and evaluate each on original and compositionally synthesized test sets. If some models show risk ratios far from 1 while the reported Transformer ratios stay near 1, then the risk-ratio test is measuring model-specific behavior rather than distributional identity. Alternatively, run a distribution-free two-sample test, such as maximum mean discrepancy, on the 3-hour blocks; a significant rejection with risk ratio near 1 would falsify the paper's inference.

Watch

Extended reading notes

Core claim

The central claim is that clinical time series are generated by a compositional process in the algebraic sense defined in the paper: a homomorphism links a latent structure of physiological states to observed measurement vectors, so that observing a sequence of states equals a composition of the observations of the individual states in their temporal order. Because neither the latent states nor the composition function are known, the paper reconstructs both from data: 3-hour blocks of multivariate measurements are clustered, either in input space or in the hidden-state space of a forecasting Transformer, into a finite symbol vocabulary; each patient stay becomes a symbol sequence; and a dist

Load-bearing premise

The load-bearing premise is that equal expected risk on synthesized and original data means the two data-generating distributions are identical; if that inference fails, the compositionality conclusion is just a statement about one forecast model.

Editorial extensions

If this is right

  • If the central claim is right, compositionally synthesized ICU time series can serve as surrogate training data, directly addressing sparse and low-resource clinical forecasting settings.
  • The near-unity risk ratios imply that compositionally synthesized test sets can stand in for original test sets in model evaluation, which helps when original test data are limited or privacy-restricted.
  • The sharp contrast with CutMix indicates that the benefit comes from respecting compositional structure rather than from sheer data volume.
  • The SOFA downstream results imply that scaling up compositionally synthesized data can improve clinical scoring beyond what original data alone allow.
  • The two empirical tests provide a transferable protocol: any time-series domain can be probed for compositionality by checking whether expected risks on original and synthesized data coincide.

Reading between the lines

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

  • Editorial inference: the practical augmentation recipe stands even if the strong compositionality claim does not; symbolizing clinical blocks and recombining context-matched fragments is a cheap way to multiply ICU data, and the paper's own downstream results show gains with dataset size.
  • Editorial inference: the learned symbol vocabulary implies a concrete cross-dataset prediction — if physiological states are the same across hospitals, k-means symbols learned on one ICU dataset should transfer to another with minor recalibration; the paper does not test transfer.
  • Editorial inference: a stricter test of compositionality would check whether rare but clinically important transitions, such as the onset of sepsis or organ failure, are reproduced in synthetic trajectories with the same conditional probabilities as in original data; aggregate risks could mask failures on rare events.
  • Editorial inference: the expected-risk-ratio protocol could be inverted to audit generative models of time series generally, offering a domain-adaptation-based evaluation alternative to discriminative scores and PCA visualizations.
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

4 major / 4 minor

Summary. The paper proposes a formal definition of compositionality for time series as a property of a data-generating process (Definition 1) and an empirical pipeline to detect it: multivariate clinical time series are segmented, embedded via a time-series-forecasting transformer encoder, clustered into symbols, and recombined by the distributional data-augmentation algorithm of Andreas (2020). On MIMIC-III and eICU, models trained on compositionally synthesized data achieve expected risk close to models trained on original data (Table 1), models evaluated on synthesized test data achieve risk ratios near 1 (Table 2), and training on larger synthesized sets improves SOFA-score prediction (Table 3). The authors claim these results establish that clinical time series are generated by a compositional process, using two domain-adaptation-based tests: equal expected risk is said to imply identical distributions, and Theorem 1 is said to convert risk-ratio bounds into density-ratio bounds. The empirical pipeline is executed carefully, but the inferential chain that connects the experiments to the paper's central claim is not sound.

Significance. If the central inference were valid, the paper would provide a meaningful proof-of-concept: a data-driven method for discovering compositional structure in clinical time series and a theoretical criterion for compositionality. The empirical part has real strengths: two public ICU datasets, multiple random seeds for both optimization and symbolization, a linear mixed-effects analysis for Table 1, an external SOFA benchmark, a comparison against CutMix as a non-compositional control, and a link to reproducible code. These assets make the empirical finding that compositionally synthesized data can serve as useful training and test data for clinical forecasting models credible. However, the paper's theoretical bridge from risk equality to distributional identity, and from distributional identity to compositional generation, is broken. The headline conclusion is therefore unsupported, and the claimed 'empirically testable criteria for compositionality' are not established.

major comments (4)
  1. [§3.2, Theorem 1 (Eq. 5); Appendix A.3] The proof of necessity mis-negates the universal statement. The proof assumes that the negation of '∀x: C f_Q(x) ≤ f_P(x)' is '∀x: C f_Q(x) > f_P(x)', but the correct negation is '∃x: C f_Q(x) > f_P(x)'. The contradiction argument therefore fails. Moreover, the asserted equivalence is false for arbitrary hypothesis classes and loss functions: a hypothesis class that cannot distinguish regions where the densities differ can have equal expected risks under P and Q even when the density ratio is unbounded. Since risk-ratio bounds only constrain the density ratio through the integrals realized by H, Theorem 1 as stated cannot serve as the theoretical foundation of Test 2.
  2. [§5.2.2, Table 2] Even if Theorem 1 were corrected, its statement requires the risk-ratio bounds to hold for every h in H. The experiment estimates the risk ratio for a single model h* selected as the best MSE model. Equality for one model does not constrain the density ratio; a single model can be insensitive to regions where the distributions differ. Thus the experiment does not implement the theorem's antecedent, and the reported ratios near 1 do not warrant the conclusion that the original and synthetic distributions are identical.
  3. [§3.2, Test 1 rationale] The paper states: 'If the estimated expected risks are identical, DA is considered successful and ... we can conclude that both distributions must be identical.' This is a non sequitur. The domain-adaptation results of Ben-David et al. provide sufficient conditions (small A-distance plus a low-error joint predictor) for low target risk; they do not state that equal expected risks of a particular learned model imply distributional identity. Two different marginals can yield equal expected risks for a fixed model, for example if the model's errors occur in different regions or if the model is constant. Therefore Test 1 cannot support the strong distribution-identity inference on which the paper's main claim rests.
  4. [§3.1, Definition 1; §6] Even if the two tests established equality of the original and synthetic distributions, the conclusion that 'the original data generating process must be compositional' does not follow. Definition 1 requires a specific algebraic structure with a homomorphism φ and composition functions. A distribution can be generated by a compositional process and also by many non-compositional processes. The CDS algorithm imposes a fragment/template recombination structure on the data; obtaining a similar marginal distribution under one forecasting model does not identify the latent generative mechanism. This is a logical gap that invalidates the paper's central scientific claim, not merely a caveat about interpretation.
minor comments (4)
  1. [§3.2, Theorem 1 and Eq. (5)-(6)] The loss notation is inconsistent: earlier definitions use ℓ(y, h(x)), while Theorem 1 writes ℓ(x, h(x)). Under covariate shift the conditional y|x is assumed shared, but the notation should be made precise to avoid ambiguity about what is being integrated.
  2. [§5.2.2, Table 2] The risk ratios in Table 2 are reported without standard errors or confidence intervals. Since the interpretation of Test 2 depends on the ratios being 'close to 1', some uncertainty quantification is needed.
  3. [Abstract and Appendix A.8] There are several typos: 'this methods' in the abstract; 'can he interpreted' in Appendix A.8; and an extra parenthesis after 'Pollard et al., 2018))' in Section 5.1.
  4. [Appendix A.7] Table 9 reports Hellinger distances for a single symbolization variant (input space) and does not explain how the 'desired properties for compositional data' are derived from the small unigram distance and larger n-gram distance. A short formal justification would improve clarity, though this is not central to the paper's main claim.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; the central inference is logically under-supported but not a self-referential reduction.

full rationale

The paper does not fit parameters to test labels in the central TSF tests, and the SOFA downstream evaluation is an external benchmark with real clinical labels. The compositional synthesizer is built from original subsequences, so distributional similarity is partly expected by construction, but the TSF-risk tests are not identical to the distributional objective used to build the synthesizer, and CutMix provides a non-compositional control. The main problems are logical validity, not circularity: inferring identical distributions from equal expected risks of a single model misreads Ben-David et al.'s sufficient conditions, and the proof of Theorem 1 in Appendix A.3 is invalid because it negates 'for all x: C f_Q(x) <= f_P(x)' as 'for all x: C f_Q(x) > f_P(x)' instead of 'there exists x with C f_Q(x) > f_P(x)'. These are correctness risks, not circular reductions. Definition 1 is also broad enough that any data-generating process could be called compositional by choosing Z=X, phi=id, and CZ=CX=f, which would make the empirical tests unnecessary; the paper does not address this, but it does not reduce a fitted quantity to a prediction. The only self-citations (Riezler & Hagmann 2024 for mixed-effects models; Staniek et al. 2024 for 24-hour binning) are methodological and not load-bearing for the compositionality claim. Hence no significant circularity; score 2 reflects minor non-load-bearing self-citation.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The free parameters are modeling choices in the symbolization and synthesis pipeline. The axioms include the distributional assumptions required for the domain adaptation tests and the stronger assumption that the recombination algorithm reflects the true generative structure. The one invented entity is the interpretive mapping of clusters to physiological states, which lacks external evidence.

free parameters (5)
  • Environment window size w = not reported in paper; set in code
    Controls how contexts are compared for fragment interchangeability in Algorithm 1; directly shapes which sequences are synthesized.
  • Number of symbols (#Syms) = 160 (best of 40, 80, 160)
    Vocabulary size of the symbol space; selected on validation data.
  • Block length Delta = 3 hours
    Subsequence length used for symbolization; hand-picked as compromise between imputation sparsity and temporal detail.
  • Encoder hidden size for symbolization = 50
    Dimensionality of the representations clustered to form symbols.
  • SOFA label construction rules = missing values set to 0; recalculated MAP, PaO2/FiO2
    These rules define the labels used in the downstream experiment; they are choices, not externally given labels.
assumptions (5)
  • domain assumption Covariate shift between synthetic and original distributions
    Invoked in Theorem 1 and in the two tests; synthetic sequences are stitched from different patients, so P(Y|X) likely differs from Q(Y|X).
  • domain assumption Existence of a low-error joint predictor h* on both synthetic and original distributions
    Needed for the Ben-David et al. sufficient conditions that the paper relies on for Test 1 (Section 3.2).
  • domain assumption Common support of synthetic and original distributions
    Definition 3 assumes common support; novel recombinations may produce points outside the original support or missing combinations.
  • ad hoc to paper The fragment/template/environment recombination algorithm (Andreas 2020) captures the composition structure of the clinical data generation process
    The central inference assumes that the data-driven recombination rules correspond to the true latent composition rules, which is stated but not tested.
  • domain assumption Clusters correspond to meaningful physiological states
    Interpretation of clusters as tachycardia, hypertension, and similar states rests on qualitative inspection (Appendix A.8), with no clinical validation; the paper itself notes a lack of clinical interpretation as a limitation.
invented entities (1)
  • Physiological states as symbolic cluster labels
    purpose: Provide the latent-state vocabulary that the compositional synthesis is said to recombine
    Clusters are labeled post hoc as tachycardia, hypertension, tachypnea in Appendix A.8; no independent clinical or predictive validation is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Compositionality in Time Series: A Proof of Concept using Symbolic Dynamics and Compositional Data Augmentation." pith.science (2026). https://pith.science/paper/NNFI5SYY

@misc{pith2026250820656,
  author       = {Pith},
  title        = {Pith review of: Compositionality in Time Series: A Proof of Concept using Symbolic Dynamics and Compositional Data Augmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NNFI5SYY}},
  note         = {Machine review of arXiv:2508.20656}
}
read the original abstract

This work investigates whether time series of natural phenomena can be understood as being generated by sequences of latent states which are ordered in systematic and regular ways. We focus on clinical time series and ask whether clinical measurements can be interpreted as being generated by meaningful physiological states whose succession follows systematic principles. Uncovering the underlying compositional structure will allow us to create synthetic data to alleviate the notorious problem of sparse and low-resource data settings in clinical time series forecasting, and deepen our understanding of clinical data. We start by conceptualizing compositionality for time series as a property of the data generation process, and then study data-driven procedures that can reconstruct the elementary states and composition rules of this process. We evaluate the success of this methods using two empirical tests originating from a domain adaptation perspective. Both tests infer the similarity of the original time series distribution and the synthetic time series distribution from the similarity of expected risk of time series forecasting models trained and tested on original and synthesized data in specific ways. Our experimental results show that the test set performance achieved by training on compositionally synthesized data is comparable to training on original clinical time series data, and that evaluation of models on compositionally synthesized test data shows similar results to evaluating on original test data, outperforming randomization-based data augmentation. An additional downstream evaluation of the prediction task of sequential organ failure assessment (SOFA) scores shows significant performance gains when model training is entirely based on compositionally synthesized data compared to training on original data.

Figures

Figures reproduced from arXiv: 2508.20656 by the authors.

Figure 1
Figure 1. Conceptualization of compositional data generation process for time series of clinical variables: [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Empirical deconstruction of multivariate time series for compositional data synthesization: First, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Compositional data synthesization procedure for symbol sequences adapted from Andreas (2020). [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Physiological states learned by time series symbolization based on clustering of neural representa [PITH_FULL_IMAGE:figures/full_fig_p026_4.png]
Figure 5
Figure 5. Figure 5: Physiological states learned by time series symbolization based on clustering in input space. [PITH_FULL_IMAGE:figures/full_fig_p027_5.png]
Figure 6
Figure 6. Figure 6: PCA visualisation for MIMIC-III dataset. PCA is applied to the input portion of the dense [PITH_FULL_IMAGE:figures/full_fig_p029_6.png]
Figure 7
Figure 7. Figure 7: PCA visualisation on masked portions of dense representations for MIMIC-III dataset. Here, CDS [PITH_FULL_IMAGE:figures/full_fig_p030_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

58 extracted references · 38 canonical work pages

  1. [1]

    u rek, Afra Feyza Aky \

    Ekin Aky \"u rek, Afra Feyza Aky \"u rek, and Jacob Andreas. Learning to recombine and resample data for compositional generalization. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=PS3IMnScugk

    work page 2021

  2. [2]

    Good-enough compositional data augmentation

    Jacob Andreas. Good-enough compositional data augmentation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL) , Online, 2020. URL 10.18653/v1/2020.acl-main.676

    work page doi:10.18653/v1/2020.acl-main.676 2020

  3. [3]

    Bolker, and Steven C

    Douglas Bates, Martin M \"a chler, Benjamin M. Bolker, and Steven C. Walker. Fitting linear mixed-effects models using lme4 . Journal of Statistical Software, 67 0 (1): 0 1--48, 2015. URL https://doi.org/10.18637/jss.v067.i01

    doi:10.18637/jss.v067.i01 2015

  4. [4]

    Domain adaptation - can quantity compensate for quality? Annals of Mathematics and Artificial Intelligence, 70: 0 185--202, 2014

    Shai Ben-David and Ruth Urner. Domain adaptation - can quantity compensate for quality? Annals of Mathematics and Artificial Intelligence, 70: 0 185--202, 2014. URL https://doi.org/10.1007/s10472-013-9371-9

    work page doi:10.1007/s10472-013-9371-9 2014

  5. [5]

    Analysis of representations for domain adaptation

    Shai Ben-David, John Blitzer, Koby Crammer, and Fernando Pereira. Analysis of representations for domain adaptation. In Proceedings of the 20th Annual Conference on Neural Information Processing Systems (NIPS'06) , Vancouver, Canada, 2006. URL https://proceedings.neurips.cc/paper_files/paper/2006/file/b1b0432ceafb0ce714426e9114852ac7-Paper.pdf

    work page 2006

  6. [6]

    A theory of learning from different domains

    Shai Ben-David, John Blitzer, Koby Crammer, Alex Kulesza, Fernando Pereira, and Jennifer Wortman Vaughan. A theory of learning from different domains. Machine learning, 79: 0 151--175, 2010 a . URL https://link.springer.com/article/10.1007/s10994-009-5152-4

    doi:10.1007/s10994-009-5152-4 2010

  7. [7]

    Impossibility theorems for domain adaptation

    Shai Ben-David, Tyler Lu, Teresa Luu, and David Pal. Impossibility theorems for domain adaptation. In Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics (AISTATS) , pp.\ 129--136, Chia Laguna Resort, Sardinia, Italy, 2010 b . URL https://proceedings.mlr.press/v9/david10a.html

    work page 2010

  8. [8]

    Representation learning: A review and new perspectives

    Yoshua Bengio, Aaron Courville, and Pascal Vincent. Representation learning: A review and new perspectives. IEEE Trans. Pattern Anal. Mach. Intell., 35 0 (8): 0 1798–1828, 2013. URL https://doi.org/10.1109/TPAMI.2013.50

    doi:10.1109/tpami.2013.50 2013

Show all 58 references
  1. [9]

    A survey of mix-based data augmentation: Taxonomy, methods, applications, and explainability

    Chengtai Cao, Fan Zhou, Yurou Dai, Jianping Wang, and Kunpeng Zhang. A survey of mix-based data augmentation: Taxonomy, methods, applications, and explainability. ACM Comput. Surv., 57 0 (2), 2024. URL https://doi.org/10.1145/3696206

    2024 doi

  2. [10]

    Syntactic Structures

    Noam Chomsky. Syntactic Structures. De Gruyter Mouton, 1957

    1957

  3. [11]

    On certain formal properties of grammars

    Noam Chomsky. On certain formal properties of grammars. Information and Control, 2 0 (2): 0 137--167, 1959. URL https://doi.org/10.1016/S0019-9958(59)90362-6

    1959 doi

  4. [12]

    Mixed Models: Theory and Applications with R

    Eugene Demidenko. Mixed Models: Theory and Applications with R . Wiley, 2013. URL https://doi.org/10.1002/0471728438

    2013 doi

  5. [13]

    Hyland, and Gunnar Rätsch

    Cristóbal Esteban, Stephanie L. Hyland, and Gunnar Rätsch. Real-valued (medical) time series generation with recurrent conditional GANs . arXiv, abs/1706.02633, 2017. URL https://arxiv.org/abs/1706.02633

    2017 arXiv

  6. [14]

    John R. Firth. A synopsis of linguistic theory, 1930-55. In Studies in Linguistic Analysis, pp.\ 1--31. Blackwell, 1957

    1930

  7. [15]

    Reddy, and Vignesh Subbian

    Hamid Ghaderi, Brandon Foreman, Amin Nayebi, Sindhu Tipirneni, Chandan K. Reddy, and Vignesh Subbian. A self-supervised learning-based approach to clustering multivariate time-series data with missing values (SLAC-Time) : An application to TBI phenotyping. Journal of Biomedica...

    2023

  8. [16]

    Empirical study of mix-based data augmentation methods in physiological time series data

    Peikun Guo, Huiyuan Yang, and Akane Sano. Empirical study of mix-based data augmentation methods in physiological time series data. In The 11th IEEE International Conference on Healthcare Informatics (IEEE ICHI) , Houston, TX, USA , 2023. URL https://doi.org/10.48550/arXiv.2309.09970

    2023 doi arXiv

  9. [17]

    Zellig S. Harris. Distributional structure. WORD, 10 0 (2-3): 0 146--162, 1954. URL https://doi.org/10.1080/00437956.1954.11659520

    1954

  10. [18]

    Higgins, Lo \"i c Matthey, A

    I. Higgins, Lo \"i c Matthey, A. Pal, C. Burgess, Xavier Glorot, M. Botvinick, S. Mohamed, and Alexander Lerchner. beta-VAE : Learning basic visual concepts with a constrained variational framework. In Proceedings of the 5th International Conference on Learning Representations...

    2017

  11. [19]

    Set functions for time series

    Max Horn, Michael Moor, Christian Bock, Bastian Rieck, and Karsten Borgwardt. Set functions for time series. In Proceedings of the 37th International Conference on Machine Learning (ICML) , Online, 2020. URL https://proceedings.mlr.press/v119/horn20a.html

    2020

  12. [20]

    Towards equipping transformer with the ability of systematic compositionality

    Chen Huang, Peixin Qin, Wenqiang Lei, and Jiancheng Lv. Towards equipping transformer with the ability of systematic compositionality. In Proceedings of the AAAI Conference on Artificial Intelligence, 2024. URL 10.1609/aaai.v38i16.29788

    2024 doi

  13. [21]

    Theo M.V. Janssen. Compositionality: its historic context. In The Oxford handbook of compositionality, pp.\ 19--46. Oxford University Press, 2012. URL https://eprints.illc.uva.nl/1239/2/LP-1996-03.text.pdf

    2012

  14. [22]

    Johnson, Tom J

    Alistair E.W. Johnson, Tom J. Pollard, Lu Shen, Li wei H. Lehman, Mengling Feng, Mohammad Ghassemi, Benjamin Moody, Peter Szolovits, Leo Anthony Celi, and Roger G. Mark. MIMIC-III , a freely accessible critical care database. Scientific Data, 3 0 (1): 0 160035, 2016. URL https...

    2016 doi

  15. [23]

    Measuring compositional generalization: A comprehensive method on realistic data

    Daniel Keysers, Nathanael Sch \"a rli, Nathan Scales, Hylke Buisman, Daniel Furrer, Sergii Kashubin, Nikola Momchev, Danila Sinopalnikov, Lukasz Stafiniak, Tibor Tihon, Dmitry Tsarkov, Xiao Wang, Marc van Zee, and Olivier Bousquet. Measuring compositional generalization: A com...

    2020

  16. [24]

    COGS : A compositional generalization challenge based on semantic interpretation

    Najoung Kim and Tal Linzen. COGS : A compositional generalization challenge based on semantic interpretation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) , Online, 2020. URL 10.18653/v1/2020.emnlp-main.731

    2020 doi

  17. [25]

    Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks

    Brenden Lake and Marco Baroni. Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks. In Proceedings of the 35th International Conference on Machine Larning (ICML) , Stockholm, Sweden, 2018. URL http://proceedings.mlr.pres...

    2018

  18. [26]

    Lake, Tomer D

    Brenden M. Lake, Tomer D. Ullman, Joshua B. Tenenbaum, and Samuel J. Gershman. Building machines that learn and think like people. Behavioral and Brain Sciences, 40: 0 e253, 2017. URL https://doi.org/10.1017/S0140525X16001837

    2017 doi

  19. [27]

    An Introduction to Symbolic Dynamics and Coding

    Douglas Lind and Brian Marcus. An Introduction to Symbolic Dynamics and Coding. Cambridge University Press, 1995

    1995

  20. [28]

    Stuart P. Lloyd. Least square quantization in PCM . IEEE Transactions on Information Theory , 28 0 (2): 0 129--137, 1982. URL https://doi.org/10.1109/TIT.1982.1056489

    1982

  21. [29]

    Challenging common assumptions in the unsupervised learning of disentangled representations

    Francesco Locatello, Stefan Bauer, Mario Lucic, Gunnar Raetsch, Sylvain Gelly, Bernhard Sch \"o lkopf, and Olivier Bachem. Challenging common assumptions in the unsupervised learning of disentangled representations. In Proceedings of the 36th International Conference on Machin...

    2019

  22. [30]

    Learning representations for time series clustering

    Qianli Ma, Jiawei Zheng, Sen Li, and Gary W Cottrell. Learning representations for time series clustering. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alch\' e -Buc, E. Fox, and R. Garnett (eds.), Advances in Neural Information Processing Systems (NeurIPS) , 2019. URL h...

    2019

  23. [31]

    Universal grammar

    Richard Montague. Universal grammar. Theoria, 36: 0 373--398, 1970

    1970

  24. [32]

    Lost in latent space: Examining failures of disentangled models at combinatorial generalisation

    Milton Montero, Jeffrey Bowers, Rui Ponte Costa, Casimir Ludwig, and Gaurav Malhotra. Lost in latent space: Examining failures of disentangled models at combinatorial generalisation. In Advances in Neural Information Processing Systems (NeurIPS) , 2022. URL https://proceedings...

    2022

  25. [33]

    The role of disentanglement in generalisation

    Milton Llera Montero, Casimir JH Ludwig, Rui Ponte Costa, Gaurav Malhotra, and Jeffrey Bowers. The role of disentanglement in generalisation. In International Conference on Learning Representations (ICLR) , 2021. URL https://openreview.net/forum?id=qbH974jKUVy

    2021

  26. [34]

    Barbara H. Partee. Compositionality. In Varieties of formal semantics, volume 3, pp.\ 281--311. 1984. URL https://web.mit.edu/jda/www/teaching/6.884/readings/partee_1984.pdf

    1984

  27. [35]

    Revisiting the compositional generalization abilities of neural sequence models

    Arkil Patel, Satwik Bhattamishra, Phil Blunsom, and Navin Goyal. Revisiting the compositional generalization abilities of neural sequence models. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL) , Dublin, Ireland, 2022. URL https...

    2022

  28. [36]

    Patel, E

    P. Patel, E. Keogh, J. Lin, and S. Lonardi. Mining motifs in massive time series databases. In Proceedings of the 2002 IEEE International Conference on Data Mining (ICDM) , Maebashi City, Japan, 2002. URL https://doi.org/10.1109/ICDM.2002.1183925

    2002 arXiv

  29. [37]

    Scikit-learn: Machine learning in python

    Fabian Pedregosa, Ga \"e l Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, et al. Scikit-learn: Machine learning in python. Journal of machine learning research (JMLR) , 12: 0 282...

    2011

  30. [38]

    Towards Generating Real-World Time Series Data

    Hengzhi Pei, Kan Ren, Yuqing Yang, Chang Liu, Tao Qin, and Dongsheng Li. Towards Generating Real-World Time Series Data . In 2021 IEEE International Conference on Data Mining (ICDM) , Los Alamitos, CA, USA, 2021. URL https://doi.ieeecomputersociety.org/10.1109/ICDM51629.2021.00058

    2021

  31. [39]

    Pollard, Alistair E

    Tom J. Pollard, Alistair E. W. Johnson, Jesse D. Raffa, Leo A. Celi, Roger G. Mark, and Omar Badawi. The eICU collaborative research database, a freely available multi-center database for critical care research. Scientific Data, 5 0 (180178), 2018. URL https://doi.org/10.1038/...

    2018 doi

  32. [40]

    Improving compositional generalization with latent structure and data augmentation

    Linlu Qiu, Peter Shaw, Panupong Pasupat, Pawel Nowak, Tal Linzen, Fei Sha, and Kristina Toutanova. Improving compositional generalization with latent structure and data augmentation. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Com...

    2022 doi

  33. [41]

    Validity, Reliability, and Significance: Empirical Methods for NLP and Data Science

    Stefan Riezler and Michael Hagmann. Validity, Reliability, and Significance: Empirical Methods for NLP and Data Science . Springer, second edition, 2024. URL https://doi.org/10.1007/978-3-031-57065-0

    2024 doi

  34. [42]

    Compositional generalization by factorizing alignment and translation

    Jacob Russin, Jason Jo, Randall O ' Reilly, and Yoshua Bengio. Compositional generalization by factorizing alignment and translation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop, Online, 2020. URL 10.186...

    2020 doi

  35. [43]

    Transformer grammars: Augmenting transformer language models with syntactic inductive biases at scale

    Laurent Sartran, Samuel Barrett, Adhiguna Kuncoro, Milo s Stanojevi \'c , Phil Blunsom, and Chris Dyer. Transformer grammars: Augmenting transformer language models with syntactic inductive biases at scale. Transactions of the Association for Computational Linguistics, 10: 0 1...

    2022 doi

  36. [44]

    Motiflets: Simple and accurate detection of motifs in time series

    Patrick Sch\" a fer and Ulf Leser. Motiflets: Simple and accurate detection of motifs in time series. Proc. VLDB Endow., 16 0 (4): 0 725–737, 2022. URL https://doi.org/10.14778/3574245.3574257

    2022

  37. [45]

    u gelgen, Frederik Tr \

    Lukas Schott, Julius Von K \"u gelgen, Frederik Tr \"a uble, Peter Vincent Gehler, Chris Russell, Matthias Bethge, Bernhard Sch \"o lkopf, Francesco Locatello, and Wieland Brendel. Visual representation learning does not generalize strongly within the same domain. In Internati...

    2022

  38. [46]

    Early prediction of causes (not effects) in healthcare by long-term clinical time series forecasting

    Michael Staniek, Marius Fracarolli, Michael Hagmann, and Stefan Riezler. Early prediction of causes (not effects) in healthcare by long-term clinical time series forecasting. Proceedings of Machine Learning Research: Machine Learning for Healthcare, 252: 0 1--29, 2024. URL htt...

    2024 arXiv

  39. [47]

    Compositionality

    Zolt \'a n Gendler Szab \'o . Compositionality. Stanford Encyclopedia of Philosophy, 2020. URL https://plato.stanford.edu/entries/compositionality/

    2020

  40. [48]

    Sindhu Tipirneni and Chandan K. Reddy. Self-supervised transformer for sparse and irregularly sampled multivariate clinical time-series. ACM Transactions on Knowledge Discovery from Data , 16 0 (6), 2022. doi:10.1145/3516367. URL https://doi.org/10.1145/3516367

    2022 doi

  41. [49]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in Neural Information Processing Systems (NIPS) , Long Beach, CA , 2017. URL https://proceedings.neurips.cc/pape...

    2017

  42. [50]

    Vincent, R

    J.L. Vincent, R. Moreno, J. Takala, S. Willatts, A. De Mendon c a, H. Bruining, C. Reinhart, P. Suter, and L. Thijs. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. Intensive Care Medicine, 22 0 (7): 0 707--710, 1996. URL https:/...

    1996 doi

  43. [51]

    Time series data augmentation for deep learning: A survey

    Qingsong Wen, Liang Sun, Fan Yang, Xiaomin Song, Jingkun Gao, Xue Wang, and Huan Xu. Time series data augmentation for deep learning: A survey. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21 , 2021. URL https://doi.org/10.24...

    2021 doi

  44. [52]

    Compositional generalization from first principles

    Thadd \"a us Wiedemer, Prasanna Mayilvahanan, Matthias Bethge, and Wieland Brendel. Compositional generalization from first principles. In Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS) , New Orleans, LA, USA , 2023. URL https://openreview.net/for...

    2023

  45. [53]

    Introduction to symbolic dynamics

    Susan Williams. Introduction to symbolic dynamics. In Proceedings of the Symposia in Applied Mathematics, volume 60, 1-12 2004

    2004

  46. [54]

    Compositional generalization in unsupervised compositional representation learning: a study on disentanglement and emergent language

    Zhenlin Xu, Marc Niethammer, and Colin Raffel. Compositional generalization in unsupervised compositional representation learning: a study on disentanglement and emergent language. In Proceedings of the 36th International Conference on Neural Information Processing Systems (Ne...

    2022

  47. [55]

    Robust augmentation for multivariate time series classification

    Hong Yang and Travis Desell. Robust augmentation for multivariate time series classification. arXiv, abs/2201.11739, 2022. URL https://doi.org/10.48550/arXiv.2201.11739

    2022 doi arXiv

  48. [56]

    Time-series generative adversarial networks

    Jinsung Yoon, Daniel Jarrett, and Mihaela van der Schaar. Time-series generative adversarial networks. In Advances in Neural Information Processing Systems (NeurIPS) , 2019. URL https://proceedings.neurips.cc/paper_files/paper/2019/file/c9efe5f26cd17ba6216bbe2a7d26d490-Paper.pdf

    2019

  49. [57]

    Cutmix: Regularization strategy to train strong classifiers with localizable features

    Sangdoo Yun, Dongyoon Han, Sanghyuk Chun, Seong Joon Oh, Youngjoon Yoo, and Junsuk Choe. Cutmix: Regularization strategy to train strong classifiers with localizable features. In Proceedings of ICCV , Seoul, Korea (South), 2019. URL 10.1109/ICCV.2019.00612

    2019

  50. [58]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.