Pith. sign in

REVIEW 3 major objections 4 minor 53 references

A graph of macro drivers and FX prices, processed by logical message-passing, anticipates four USD/JPY market regimes with competitive out-of-sample accuracy while staying fully interpretable.

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 · grok-4.5

2026-07-10 22:45 UTC pith:XCUT5ZIY

load-bearing objection Solid GraphTM application to hourly USD/JPY regimes with a useful ablation, but the superiority claim is not apples-to-apples and Class 3 is too thin. the 3 major comments →

arxiv 2607.06719 v1 pith:XCUT5ZIY submitted 2026-07-07 cs.CE cs.LG

Macroeconomic Message Passing for Anticipating Foreign Exchange Regime Changes: A Deep Logical Learning Approach using Graph Tsetlin Machines

Christian Blakely , Melanie Gilmore This is my paper
classification cs.CE cs.LG
keywords Graph Tsetlin Machineforeign exchange regimesmessage passingmacroeconomic drivershypervector encodingUSD/JPYinterpretable machine learningregime anticipation
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.

The paper claims that foreign-exchange regimes can be anticipated more accurately, and more interpretably, if the market is represented as a directed multigraph whose nodes carry technical and macroeconomic variables and whose edges carry causal or correlative influence. Message-passing across that graph lets a Graph Tsetlin Machine build nested logical clauses that recognise complex sub-graph patterns, including the prolonged low-volatility “droughts” that defeat ordinary statistical models. On hourly USD/JPY data the full macro-topological model reaches roughly 71 percent overall out-of-sample accuracy, markedly higher than a local-only graph or convolutional Tsetlin baselines, and supplies a rolling risk score that a trend-following engine can use to throttle exposure. The practical stake is clear: a transparent, logic-based detector that reacts quickly to policy shocks and cross-market spillovers could improve regime-aware risk management without black-box opacity.

Core claim

Representing USD/JPY price, volatility, efficiency, bond yields and oil as hypervectorised nodes linked by typed edges, then training a Graph Tsetlin Machine with message-passing, produces deep conjunctive clauses that anticipate the next 72-hour majority-vote regime with 70.69 percent overall out-of-sample accuracy—outperforming reduced local graphs, convolutional Tsetlin machines and a wide suite of AutoML models—while remaining fully interpretable.

What carries the argument

Graph Tsetlin Machine message-passing: layer-zero clauses evaluate local node properties and emit sparse-binary messages along typed edges; deeper clause components inspect neighbour inboxes, building nested logical rules that recognise sub-graph patterns with far fewer clauses than a flat feature vector would require.

Load-bearing premise

The ground-truth label is defined as the majority-vote regime over the next 72 hours; if that horizon or the ATR/ER thresholds do not match the regime a trader actually experiences, every accuracy and risk number becomes mis-calibrated.

What would settle it

Re-label the identical hourly series with a materially different look-ahead window (for example 24 h or 120 h) or with different ATR/ER cut-offs and re-train; if the full GraphTM’s accuracy advantage over the reduced graph and CoTM baselines collapses, the central claim fails.

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

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

3 major / 4 minor

Summary. The paper proposes a Graph Tsetlin Machine (GraphTM) that encodes USD/JPY technical indicators and exogenous macro drivers (bond yields, oil, cross-pair ATR/ER) as a hypervectorized directed multigraph. Message-passing constructs deep conjunctive clauses that anticipate one of four 72-hour majority-vote regimes (stagnant, steady trend, choppy, volatile trend). On a purged chronological 60/40 split the full graph reports 70.69 % overall OOS accuracy (Class 0 80.5 %, Class 2 71.9 %), outperforming a local-only ablation (48 %) and several CoTM/HMM/GBM baselines while remaining fully symbolic.

Significance. If the empirical ranking holds under consistent evaluation, the work supplies a rare fully interpretable, logic-based alternative to black-box regime classifiers that can ingest typed macro relationships via message passing. Strengths already present are the leakage-protected walk-forward design with 72-hour purge/embargo, the explicit ablation of cross-market edges, the multi-architecture comparison (including H2O AutoML), the asymmetric trading-risk score, and the public code. These elements make the contribution falsifiable and useful for quantitative-finance audiences interested in symbolic ML.

major comments (3)
  1. [§4.3, Table 2] Table 2 / §4.3 claim of “superior predictive capabilities” rests on non-comparable statistics: GraphTM numbers are means ± std over 100 random hyper-parameter draws (Table 1), while every baseline (GBM, GraphNN, CoTM L=0–2, HMM, H2O suite) reports only the single best of 100 draws. Under the extreme imbalance (Class 0 ≈ 87 % of OOS samples) best-of-100 systematically inflates overall and minority-class accuracy; the ranking that places Full GraphTM above CoTM (L=2) and several AutoML models is therefore not secured.
  2. [Table 1, §4.3] Class 3 (volatile trend) contains only 385 OOS observations and yields 11.2 % mean accuracy. Because the paper’s central narrative includes anticipation of high-volatility structural shifts, the near-chance performance on this regime must be either (a) acknowledged as a hard limit of the current graph or (b) mitigated by re-balancing / cost-sensitive training before the “four-regime” claim can be maintained.
  3. [Appendix 3.1, Table 3] Appendix Table 3 shows several H2O models (DeepLearning-grid-2 72.80 %, GLM-11 71.99 %, GBM-grid-12 71.69 %) that exceed the reported GraphTM overall accuracy even under the authors’ own “best-of-100” protocol. The superiority statement in the abstract and §4.3 therefore needs quantitative qualification or a re-run under identical aggregation (mean or median across the same 100 seeds).
minor comments (4)
  1. [§3] Eq. (4) and the surrounding text never state the numerical threshold γ used to binarise the Efficiency Ratio; without it the four-class labelling is not reproducible.
  2. [Figures 2–12] Figures 2–7 and 10–12 lack axis units, colour-bar legends, and exact date ranges; several captions refer to “light blue” without a corresponding legend entry.
  3. [§§1–2] Typographical inconsistencies: “THe GraphTM”, “Y en”, mixed capitalisation of “usdjpy”/“USD/JPY”, and missing spaces after periods appear throughout §§1–2.
  4. [§4.4, Eq. (5)] The cost-matrix entries in Eq. (5) are presented as “example” values yet are used without sensitivity analysis; a one-sentence note that results are robust to moderate rescaling of λ_whipsaw would strengthen §4.4.

Circularity Check

0 steps flagged

No circularity: empirical supervised learning with properly future-looking labels and independent OOS evaluation; self-citations are methodological, not load-bearing reductions.

full rationale

The paper's central claims are empirical performance numbers (Table 1 mean accuracies, Table 2 comparative OOS matrix, ablation of full vs reduced graph) obtained by training GraphTM clauses on current hypervectorized multigraph features (local ATR/ER trajectories, bond yields, oil, cross-pair vol) to predict a 72-hour forward majority-vote regime label y_T = mode{R( au) | au ∈ [T+1, T+72]} (Eq. 4). Features never include the future prices or the label itself; the label is constructed solely from future ATR/ER of the target series and is purged/embargoed at train/test boundaries. No equation equates a reported accuracy or risk score to a fitted constant by construction. The asymmetric cost matrix C (Eq. 5) is an explicit application-specific prior for a downstream trading-risk indicator, not optimized against the accuracy tables. Self-citations (Blakely 2022/2024 on ConvTM, Granmo et al. 2026 on GraphTM) supply the learning architecture and hypervector algebra; they do not supply uniqueness theorems, forced ansatzes, or the numerical results. The derivation chain is therefore ordinary supervised learning plus ablation, fully self-contained against the reported OOS splits. Evaluation-protocol asymmetries (mean vs best-of-100) affect comparability but do not constitute circularity under the defined patterns.

Axiom & Free-Parameter Ledger

7 free parameters · 4 axioms · 2 invented entities

The central empirical claim rests on a large set of modelling choices (regime definitions, look-ahead window, graph topology, hypervector embedding ranges, clause hyper-parameters, cost-matrix entries) that are either hand-chosen or randomly searched rather than derived. No new physical or mathematical entities are postulated; the invented objects are engineering constructs whose only evidence is the reported OOS tables.

free parameters (7)
  • 72-hour majority-vote horizon
    Chosen after “preliminary exploratory analysis”; directly defines every training label y_T.
  • ATR look-back N=10 hours and ER window
    Fixed without sensitivity study; controls the instantaneous regime R(τ).
  • regime threshold γ for ER
    Mentioned but never given a numeric value; partitions the four classes.
  • clause count C ∈ [100,400], specificity s ∈ [2,20], max literals L ∈ [20,100]
    Randomly searched; reported means depend on the sampled distribution.
  • hypervector dimension D=3200, sparsity K=320, embedding bins (40 divisions)
    Hand-set; affect Hamming geometry of all node and edge symbols.
  • asymmetric cost matrix C(i,j) entries (e.g. λ_whipsaw=10)
    Application-specific prior that defines the R_score; not learned from data.
  • voting threshold T as random fraction of clauses
    Stochastic hyper-parameter that modulates Type-I/II feedback.
axioms (4)
  • domain assumption The four-quadrant partition of (normalised ATR, ER) exhaustively and usefully describes FX regimes relevant to trading.
    §3 defines Classes 0–3 solely by these two scalars; all subsequent claims inherit this taxonomy.
  • domain assumption Pre-defined directed edges (yield→price, ATR→price, cross-vol→ATR) correctly encode the economically relevant information flow.
    Graph topology in Fig. 1 and §3 is fixed a priori; message-passing only works if those edges are informative.
  • domain assumption Sparse binary hypervectors with linear embedding preserve the order and similarity structure of continuous market scalars sufficiently for logical learning.
    Appendix 2; the entire feature representation rests on this encoding fidelity.
  • domain assumption A 60/40 chronological split with 72-hour purge eliminates all leakage for the chosen label horizon.
    §4.1; if residual serial dependence remains, OOS numbers are optimistic.
invented entities (2)
  • USD/JPY macroeconomic multigraph (nodes + typed edges of Fig. 1) no independent evidence
    purpose: Supplies the relational structure on which GraphTM message passing operates.
    Topology is hand-designed for this paper; no independent validation outside the reported tables.
  • Trading Risk Score R_score derived from the 4×4 cost matrix C no independent evidence
    purpose: Converts the confusion matrix into an asymmetric capital-protection metric for live monitoring.
    Defined in §4.4; values of C are application-specific and not externally falsifiable.

pith-pipeline@v1.1.0-grok45 · 21263 in / 3339 out tokens · 43358 ms · 2026-07-10T22:45:03.325888+00:00 · methodology

0 comments
read the original abstract

This paper introduces a graph-theoretic approach for predicting market regimes in foreign exchange (FX) currency prices. Specifically, the proposed model incorporates exogenous macroeconomic variables to update localized node features via message-passing operations. Utilizing the Graph Tsetlin Machine (GraphTM) framework, we empirically demonstrate the efficacy of this approach in anticipating market regimes for the US Dollar and Japanese Yen currency pair (USD/JPY). By representing multivariate macroeconomic drivers and technical indicators as hypervectorized directed multigraphs, the GraphTM leverages structured message passing to construct deep, interpretable logical clauses capable of recognizing complex sub-graph patterns.

Figures

Figures reproduced from arXiv: 2607.06719 by Christian Blakely, Melanie Gilmore.

Figure 1
Figure 1. Figure 1: Illustrates the full graph topology adopted in this study, including the node types, input variables, and edge relationships used to facilitate information transmission across the network. We use normalized ATR and ER values of audjpy, eurusd along with bond and Oil prices that can influence usdjpy supply and demand [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The different market data transformation on an hourly scale along 500 market trading hours. Starting from top to bottom, log price of the usdjpy, the ATR features, the macro features, and the regime class prediction [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: usdjpy spot price sequence paired with colored structural background regime tracking [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Changing from trending to mean-reversion prediction. Light blue represents low volatility trending [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Exogenous Global Macro helping navigate between trending and high-vol mean reverting [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The four different regimes our model will learn to anticipate given current market conditions. All four regimes can be characterized by two features: normalized ATR and ER [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: show examples of the four different regime labels constructed by the 3-day look-ahead window for yT [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The reduced graph representation for testing feature importance of the additional macro market indicators. All external nodes representing global macro liquidity (US and Japanese bond yield levels, complementary currency volatility profiles, and commodities (WTI_Oil) were pruned from the graph network, alongside their respective edge mappings. 4.3 Empirical Predictive Performance and Class Analysis The hyp… view at source ↗
Figure 9
Figure 9. Figure 9: Comparing the Class Regime 1 and 2 OOS accuracy across multiple models including ConvTM with two different lags and AutoML from H2O [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The entire two year period between February 2020 and June 2022. Notice the large amount of deviations in the rolling risk metric during the 2020 period followed by more structural formations with trends [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: A closeup of the COVID period when regime changes were frequent. During the initial phase of the pandemic in 2020, the asset market exhibited unprecedented volatility, driven by rapid, exogenous macroeconomic shocks and frequent central bank interventions. This systemic instability is directly captured by the risk metric plot, which displays extreme variance and remains consistently elevated toward the fi… view at source ↗
Figure 12
Figure 12. Figure 12: Post COVID period in 2021 and 2022 where regimes were more predictable and thus lower risk profile. by the GraphTM aligned closely with prevailing market conditions during this period. The minimal overlap in the rolling confusion matrix during this phase could indi￾cate that when the market enters a stable, prolonged regime, the localized cost-loss framework appears to identify the reduction in structural… view at source ↗
Figure 13
Figure 13. Figure 13: Example of encoding rich information from one node to parent node Appendix 3. Graph Learning Algorithm A GraphTM is built upon the Convolutional CoTM (see Glimsdal and Granmo 2021; Granmo, Glimsdal et al 2019), with the following hyperparemeters: depth d, number of clauses n, specificity s, and voting margin T. The GraphTM evaluates clauses and nodes in parallel as follows when given an input graph G: 1. … 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

53 extracted references · 53 canonical work pages · 9 internal anchors

  1. [1]

    2026 , eprint=

    The Tsetlin Machine Goes Deep: Logical Learning and Reasoning With Graphs , author=. 2026 , eprint=

    work page 2026

  2. [2]

    LeDell, Erin and Candel, Arno , booktitle=. H2O. 2020 , url=

    work page 2020

  3. [3]

    2024 , eprint=

    High resolution microprice estimates from limit orderbook data using hyperdimensional vector Tsetlin Machines , author=. 2024 , eprint=

    work page 2024

  4. [4]

    Blakely , booktitle=

    Christian D. Blakely , booktitle=. Tsetlin LOB: Realtime Regime Learning and Interpretable Prediction in Financial Limit Orderbooks using Convolutional Tsetlin Machines , year=. doi:10.1109/ISTM54910.2022.00012 , month=

    work page doi:10.1109/istm54910.2022.00012 2022

  5. [5]

    and Papapostolou, Nikos C

    Lee, Taehyun and Moutzouris, Ioannis C. and Papapostolou, Nikos C. and Fatouh, Mahmoud , title =. 2023 , url =

    work page 2023

  6. [6]

    , title =

    Hamilton, James D. , title =. Econometrica , year =

    work page

  7. [7]

    1997 , volume =

    Krolzig, Hans-Martin , title =. 1997 , volume =

    work page 1997

  8. [8]

    2021 , url =

    Botte, Alex and Bao, Doris , title =. 2021 , url =

    work page 2021

  9. [9]

    Regime-dependent dynamics of nuclear and coal energy in high-income economies: evidence from a Markov switching analysis , volume =

    Erdoğdu, Mevsim and Baycan, İsmail , year =. Regime-dependent dynamics of nuclear and coal energy in high-income economies: evidence from a Markov switching analysis , volume =. Frontiers in Environmental Science , doi =

    work page

  10. [10]

    and Welling, Max , title =

    Kipf, Thomas N. and Welling, Max , title =. Proceedings of the 5th International Conference on Learning Representations (ICLR) , year =

    work page

  11. [11]

    2022 , publisher=

    Jiao, Lei and Zhang, Xuan and Granmo, Ole-Christoffer and Abeyrathna, Kuruge Darshana , journal=. 2022 , publisher=

    work page 2022

  12. [12]

    Simpler Context-Dependent Logical Forms via Model Projections

    Simpler context-dependent logical forms via model projections , author=. arXiv preprint arXiv:1606.05378 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  13. [13]

    From Language to Programs: Bridging Reinforcement Learning and Maximum Marginal Likelihood

    From language to programs: Bridging reinforcement learning and maximum marginal likelihood , author=. arXiv preprint arXiv:1704.07926 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  14. [14]

    Nature Machine Intelligence , author =

    Multimodal learning with graphs , volume =. Nature Machine Intelligence , author =. 2023 , note =. doi:10.1038/s42256-023-00624-6 , language =

    work page doi:10.1038/s42256-023-00624-6 2023

  15. [15]

    Expert Systems , author =

    Using. Expert Systems , author =. 2023 , note =. doi:10.1111/exsy.12873 , language =

    work page doi:10.1111/exsy.12873 2023

  16. [16]

    Enhancing Interpretable Clauses Semantically using Pretrained Word Representation

    Yadav, Rohan Kumar and Jiao, Lei and Granmo, Ole-Christoffer and Goodwin, Morten. Enhancing Interpretable Clauses Semantically using Pretrained Word Representation. Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP. 2021

    work page 2021

  17. [17]

    2023 , eprint=

    Cognitive modeling and learning with sparse binary hypervectors , author=. 2023 , eprint=

    work page 2023

  18. [18]

    and Clarke, Paul F

    Halenka, Vojtech and Kadhim, Ahmed K. and Clarke, Paul F. A. and Bhattarai, Bimal and Saha, Rupsa and Granmo, Ole-Christoffer and Jiao, Lei and Andersen, Per-Arne , booktitle=. Exploring Effects of Hyperdimensional Vectors for Tsetlin Machines , year=

    work page

  19. [19]

    IEEE transactions on Knowledge and Data Engineering , volume=

    Representation and processing of structures with binary sparse distributed codes , author=. IEEE transactions on Knowledge and Data Engineering , volume=. 2001 , publisher=

    work page 2001

  20. [20]

    Ramalakshmi and A

    Hemalatha Gunasekaran and K. Ramalakshmi and A. Rex Macedo Arokiaraj and S. Deepa Kanmani and Chandran Venkatesan and C. Suresh Gnana Dhas , title =. Computational and Mathematical Methods in Medicine , volume =. 2021 , doi =

    work page 2021

  21. [21]

    The Convolutional Tsetlin Machine

    The convolutional Tsetlin machine , author=. arXiv preprint arXiv:1905.09688 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 1905

  22. [22]

    The Tsetlin Machine -- A Game Theoretic Bandit Driven Approach to Optimal Pattern Recognition with Propositional Logic

    The Tsetlin Machine--A Game Theoretic Bandit Driven Approach to Optimal Pattern Recognition with Propositional Logic , author=. arXiv preprint arXiv:1804.01508 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  23. [23]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Geometric deep learning on graphs and manifolds using mixture model cnns , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  24. [24]

    Gene , volume=

    Entrez nucleotide and entrez protein FAQs , author=. Gene , volume=

    work page

  25. [25]

    Proceedings of the IEEE , volume=

    Gradient-based learning applied to document recognition , author=. Proceedings of the IEEE , volume=. 1998 , publisher=

    work page 1998

  26. [26]

    Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

    Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms , author=. arXiv preprint arXiv:1708.07747 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  27. [27]

    2009 , journal=

    Learning Multiple Layers of Features from Tiny Images , author=. 2009 , journal=

    work page 2009

  28. [28]

    Journal of Intelligent Information Systems , volume=

    A relational tsetlin machine with applications to natural language understanding , author=. Journal of Intelligent Information Systems , volume=. 2022 , publisher=

    work page 2022

  29. [29]

    2023 , eprint=

    TMComposites: Plug-and-Play Collaboration Between Specialized Tsetlin Machines , author=. 2023 , eprint=

    work page 2023

  30. [30]

    Interpretable Rule-based Architecture for GNSS Jamming Signal Classification , volume =

    Jeeru, Sindhusha and Jiao, Lei and Andersen, Per-Arne and Granmo, Ole-Christoffer , year =. Interpretable Rule-based Architecture for GNSS Jamming Signal Classification , volume =. IEEE Sensors Journal , doi =

    work page

  31. [31]

    and Granmo, Ole-Christoffer , booktitle=

    Grønningsæter, Ylva and Smørvik, Halvor S. and Granmo, Ole-Christoffer , booktitle=. An Optimized Toolbox for Advanced Image Processing with Tsetlin Machine Composites , year=

    work page

  32. [32]

    2015 , publisher=

    Keras , author=. 2015 , publisher=

    work page 2015

  33. [33]

    Proceedings of human language technology conference and conference on empirical methods in natural language processing , pages=

    Recognizing contextual polarity in phrase-level sentiment analysis , author=. Proceedings of human language technology conference and conference on empirical methods in natural language processing , pages=

    work page

  34. [34]

    Amazon Sales Dataset , howpublished=

    KARKAVELRAJA J , year =. Amazon Sales Dataset , howpublished=

    work page

  35. [35]

    2023 , url =

    hfawaz , title =. 2023 , url =

    work page 2023

  36. [36]

    Coalesced Multi-Output Tsetlin Machines with Clause Sharing

    Coalesced multi-output tsetlin machines with clause sharing , author=. arXiv preprint arXiv:2108.07594 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  37. [37]

    REDRESS: Generating Compressed Models for Edge Inference Using Tsetlin Machines , year=

    Maheshwari, Sidharth and Rahman, Tousif and Shafik, Rishad and Yakovlev, Alex and Rafiev, Ashur and Jiao, Lei and Granmo, Ole-Christoffer , journal=. REDRESS: Generating Compressed Models for Edge Inference Using Tsetlin Machines , year=

    work page

  38. [38]

    Proceedings of the 30th International Conference on Neural Information Processing Systems , year =

    Defferrard, Michaël and Bresson, Xavier and Vandergheynst, Pierre , title =. Proceedings of the 30th International Conference on Neural Information Processing Systems , year =

    work page

  39. [39]

    Tsetlin Machine Embedding: Representing Words Using Logical Expressions

    Bhattarai, Bimal and Granmo, Ole-Christoffer and Jiao, Lei and Yadav, Rohan and Sharma, Jivitesh. Tsetlin Machine Embedding: Representing Words Using Logical Expressions. Findings of the Association for Computational Linguistics: EACL 2024. 2024

    work page 2024

  40. [40]

    Darshana

    K. Darshana. Philosophical Transactions of the Royal Society A , title="

    work page

  41. [41]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Drop Clause: Enhancing Performance, Robustness and Pattern Recognition Capabilities of the Tsetlin Machine. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2023 , month=. doi:10.1609/aaai.v37i11.26588 , number=

    work page doi:10.1609/aaai.v37i11.26588 2023

  42. [42]

    Tsetlin Machine for Solving Contextual Bandit Problems , url =

    Seraj, Raihan and Sharma, Jivitesh and Granmo, Ole-Christoffer , booktitle =. Tsetlin Machine for Solving Contextual Bandit Problems , url =

    work page

  43. [43]

    Language Resources and Evaluation , author =

    Annotating. Language Resources and Evaluation , author =. 2005 , pages =. doi:10.1007/s10579-005-7880-9 , language =

    work page doi:10.1007/s10579-005-7880-9 2005

  44. [44]

    and Daly, Raymond E

    Maas, Andrew L. and Daly, Raymond E. and Pham, Peter T. and Huang, Dan and Ng, Andrew Y. and Potts, Christopher , title =. Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies - Volume 1 , pages =. 2011 , isbn =

    work page 2011

  45. [45]

    Proceedings of the 29th International Conference on Neural Information Processing Systems - Volume 1 , pages =

    Zhang, Xiang and Zhao, Junbo and LeCun, Yann , title =. Proceedings of the 29th International Conference on Neural Information Processing Systems - Volume 1 , pages =. 2015 , publisher =

    work page 2015

  46. [46]

    NCBI Taxonomy: enhanced access via NCBI datasets , volume =

    Cox, Eric and Tsuchiya, Mirian and Ciufo, Stacy and Torcivia, John and Falk, Robert and Anderson, W and Holmes, J and Hem, Vichet and Breen, Laurie and Davis, Emily and Ketter, Anne and Zhang, Peifen and Soussov, Vladimir and Schoch, Conrad and O'Leary, Nuala , year =. NCBI Taxonomy: enhanced access via NCBI datasets , volume =. Nucleic acids research , doi =

    work page

  47. [47]

    ACM Transactions on Information Systems , author =

    Our. ACM Transactions on Information Systems , author =. 2024 , pages =. doi:10.1145/3675163 , abstract =

    work page doi:10.1145/3675163 2024

  48. [48]

    Anelli, Vito Walter and Bellogin, Alejandro and Ferrara, Antonio and Malitesta, Daniele and Merra, Felice Antonio and Pomo, Claudio and Donini, Francesco Maria and Di Noia, Tommaso , month = jul, year =. Elliot:. Proceedings of the 44th. doi:10.1145/3404835.3463245 , language =

    work page doi:10.1145/3404835.3463245

  49. [49]

    and Hoffman, Matthew D

    Liang, Dawen and Krishnan, Rahul G. and Hoffman, Matthew D. and Jebara, Tony , year =. Variational. Proceedings of the 2018. doi:10.1145/3178876.3186150 , language =

    work page doi:10.1145/3178876.3186150 2018

  50. [50]

    Proceedings of the 1994

    Resnick, Paul and Iacovou, Neophytos and Suchak, Mitesh and Bergstrom, Peter and Riedl, John , year =. Proceedings of the 1994. doi:10.1145/192844.192905 , language =

    work page doi:10.1145/192844.192905 1994

  51. [51]

    He, Xiangnan and Liao, Lizi and Zhang, Hanwang and Nie, Liqiang and Hu, Xia and Chua, Tat-Seng , year =. Neural. doi:10.48550/ARXIV.1708.05031 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1708.05031

  52. [52]

    Explicit expanders of every degree and size

    Amazon.com recommendations: item-to-item collaborative filtering , volume =. IEEE Internet Computing , author =. 2003 , pages =. doi:10.1109/MIC.2003.1167344 , language =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1109/mic.2003.1167344 2003

  53. [53]

    BPR: Bayesian Personalized Ranking from Implicit Feedback

    Rendle, Steffen and Freudenthaler, Christoph and Gantner, Zeno and Schmidt-Thieme, Lars , year =. doi:10.48550/ARXIV.1205.2618 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1205.2618