Pith. sign in

REVIEW 3 major objections 5 minor 48 references

Tonnetz-guided graph wedges compress piano accompaniment into fewer, harmonically coherent regions that stay playable scores.

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 ↗

Adaptive graph wedgelets in a six-dimensional Tonnetz embedding compress piano subgraphs of vocal-pianistic scores into piecewise-constant, playable reconstructions that respect harmonic distance.

T0 review reviewed 2026-07-13 challenge →

load-bearing objection Clean, usable application of graph wedgelets to piano accompaniments via Tonnetz; evaluation is thin but the pipeline is real and referee-ready. the 3 major comments →

arxiv 2607.08806 v1 pith:NKOWS2PY submitted 2026-07-09 cs.SD cs.NAmath.NA

Tonnetz-Driven Graph Wedgelet for Harmonic Complexity Reduction in Music Scores

Emmanuel Caronna , Elisa Francomano , Silvia Licciardi This is my paper
classification cs.SD cs.NAmath.NA
keywords Symbolic musicTonnetzgraphwedge partitioning treecomputational musicologyharmonic complexity reductionscore compression
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 reading

Symbolic music scores contain a lot of harmonic redundancy in the accompaniment—broken chords, Alberti basses, ostinati—that repeatedly spell the same harmony. This paper treats the piano part as a graph and partitions it with binary wedge trees whose splits are driven by a six-dimensional Tonnetz embedding, so distance reflects fifths and thirds rather than raw chromatic steps. Each resulting region is replaced by a single mean pitch that is forced to stay inside the original notes of that region, producing a simplified MusicXML score that is still human-readable and playable. Experiments on seventy scores by three composers show how root-mean-square error in the Tonnetz space falls as the number of wedges grows, and how noteheads gradually return to their original pitches. The practical aim is automatic generation of simplified arrangements that remain recognisable as the source piece.

Core claim

A fully adaptive greedy binary-wedge partitioning tree that minimises L2 error inside a six-dimensional Tonnetz embedding of the piano subgraph yields piecewise-constant reconstructions that reduce harmonic complexity while remaining valid, playable MusicXML scores whose pitches stay inside the original pitch-class set of each wedge.

What carries the argument

Tonnetz-driven FA-greedy binary wedge partitioning trees: recursive splits of the piano graph that choose the next centre by minimising harmonic L2 error under the concatenated (k=3,4,7) angular embedding, then decode each wedge by its mean and nearest original pitch class.

Load-bearing premise

The Euclidean distance on that particular six-dimensional Tonnetz map, together with shortest-path connectivity on the piano graph, is assumed to be a faithful proxy for musical harmonic homogeneity when deciding where to split.

What would settle it

On the same corpus, replace the Tonnetz L2 metric with ordinary chromatic distance (or random splits) and check whether the resulting playable reconstructions produce systematically higher pitch-error rates or lose recognisable harmonic function at equal wedge budgets.

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

Share X Bluesky LinkedIn Reddit HN

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 / 5 minor

Summary. The paper proposes a structure-preserving compression scheme for the piano accompaniment subgraph of vocal-pianistic symbolic scores. It adapts binary wedge partitioning trees (BWP) with fully adaptive greedy (FA-greedy) L2-error minimization, but evaluates the error inside a six-dimensional Tonnetz embedding of pitch classes (angular projections with multipliers k=3,4,7). Wedges remain connected under shortest-path distance on the piano graph; each wedge is represented by its mean in Tonnetz space and then quantized back to a pitch class already present in that wedge. The result is a piecewise-constant, playable MusicXML score whose pitches stay inside the original local pitch-class set. Experiments on a 70-score corpus from three composers report RMSE versus compression ratio r in the same Tonnetz space, together with qualitative score visualizations.

Significance. If the Tonnetz-driven wedges are musically meaningful, the method supplies a clean, graph-native, harmony-aware compressor that produces human-readable and playable simplified scores—something existing symbolic compressors (LZ-style, wavelet-on-melody, tokenisation) do not simultaneously achieve. The algorithmic construction is fully specified (Algorithms 1–2), the reconstruction constraint is musically sensible, and the export pipeline to MusicXML is practical. These are genuine strengths for computational musicology and for downstream tasks that need reduced harmonic complexity while preserving graph structure. The contribution is therefore of clear interest, provided the harmonic-fidelity claim can be substantiated beyond self-referential reconstruction error.

major comments (3)
  1. Section 4 and Eq. (13): the sole quantitative metric is RMSE computed inside the same six-dimensional Tonnetz space used for FA-greedy splitting (Fig. 3). No baseline (chromatic L2, single-axis Tonnetz, random connected partitions, non-adaptive BWP) is reported, nor any listening test, expert harmonic annotation, or re-evaluation of a downstream task (cadence detection, Roman-numeral analysis, voice separation) on the compressed scores. Without an external check, the claim that the Euclidean metric on Φ together with shortest-path connectivity yields harmonically homogeneous wedges remains untested; the RMSE curve alone cannot distinguish a musically faithful compressor from a self-consistent but musically arbitrary one.
  2. Section 3.2–3.3 and Algorithm 1: the free parameters of the embedding (the triple k1,k2,k3, currently fixed at 3,4,7) and of the initial centre v1 are never ablated. Because the entire partitioning criterion rests on the induced L2 metric, a short sensitivity study (or at least a comparison against the pure circle-of-fifths axis k=7) is required to show that the claimed harmonic advantage is robust rather than an artefact of one particular angular choice.
  3. Abstract and Section 5 assert that the scheme preserves “task-relevant information.” The experiments never re-run any of the analytical tasks cited in the introduction (cadence detection, voice separation, stylistic classification) on the compressed graphs. Either a concrete downstream experiment or a clear restriction of the claim to reconstruction fidelity is needed.
minor comments (5)
  1. Figure 1 caption and surrounding text: the example is taken from Mozart K. 620, yet the corpus description (Section 4) never names the three composers or the selection criteria; a short table or paragraph listing composers, number of scores per composer, and average N would improve reproducibility.
  2. Equation (10) and the paragraph that follows: the claim that “any permutation of k1,k2,k3 yields the same induced L2 metric” is true, but the text should also note that different unordered triples generally produce different metrics; the reader is left wondering why precisely {3,4,7} was chosen over other musically plausible sets.
  3. Section 3.4: the export pipeline overwrites MIDI pitches while leaving all other notational attributes untouched. A brief remark on how simultaneous notes that fall into different wedges are handled (or whether the current implementation forces them into the same wedge) would clarify edge cases.
  4. Appendix figures (Alberti bass, ostinato, chromatic circle) are pedagogically useful but are never referenced from the main text; a single cross-reference in Section 3.2 would integrate them.
  5. Typographical: “wedglet” appears once (Section 3.1) instead of “wedgelet”; “thepresentsection” and similar spacing artefacts occur in Section 2.

Circularity Check

1 steps flagged

Mild self-referential evaluation: RMSE is the same Tonnetz L2 the FA-greedy already minimises; the compression map itself is a genuine non-identity reconstruction with no definitional loop.

specific steps
  1. self definitional [Abstract; Sec. 3.2–3.3, Eqs. (5)–(6), (10), (13)]
    "The partitioning process employs a splitting criterion based on harmonic distance, resulting in regions that accurately reflect the intrinsic harmonic relationships among notes. … The L2-error in the partitioning is computed in a six-dimensional Tonnetz embedding of notes, whose euclidean metric captures harmonic rather than chromatic proximity. … RMSE = √(1/N ∑ ||f(v)-WMf(v)||^{2}_{2})"

    “Harmonic distance” is defined to be Euclidean distance on the Tonnetz embedding Φ (Eq. 10). FA-greedy selects the split that minimises precisely that L2 residual (Eqs. 5–6). Consequently the claim that the resulting wedges “accurately reflect the intrinsic harmonic relationships” is true by construction of the objective, not an independent empirical finding. The sole quantitative figure of merit (RMSE) is the same residual, so reported success is definitionally guaranteed once the metric is chosen.

full rationale

The paper adapts Erb’s FA-greedy binary wedge partitioning trees to the piano subgraph, replacing the usual signal space by a fixed six-dimensional Tonnetz embedding (k=3,4,7). The algorithm therefore produces wedges that are homogeneous in that metric by construction, and the reported RMSE (Eq. 13, Fig. 3) simply quantifies the residual of the same objective. This is ordinary for a lossy compressor and does not make the reconstructed MusicXML an identity map: the decoder still replaces each wedge by a single mean-based pitch class drawn only from the original notes of that wedge, yielding a coarser but playable score. Self-citations to the authors’ earlier heterogeneous score-graph paper supply only the input representation and are not load-bearing for the compression claim. No parameter is fitted to data and then re-used as a “prediction,” no uniqueness theorem is imported, and no known empirical pattern is merely renamed. The musical interpretation of the Tonnetz metric remains an untested modelling assumption (correctness risk, not circularity). Hence the derivation chain is essentially self-contained; the only circularity is the mild, expected self-reference of reporting approximation error inside the optimisation metric.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 1 invented entities

The central claim rests on a small set of free design choices (compression ratio schedule, embedding axes, initial centre) and on standard graph-signal and music-theoretic assumptions. No new physical entities are postulated; the only invented construct is the specific six-dimensional embedding used as the error space.

free parameters (3)
  • compression ratio r
    User-chosen scalar that sets the number of wedges M = max(2, floor(r N)); the entire rate-distortion curve is parameterized by it.
  • Tonnetz angular multipliers (k1,k2,k3)
    Fixed to the conventional triple (3,4,7) without ablation; any other triple would induce a different metric and different partitions.
  • initial centre v1
    Chosen as the temporally first piano vertex; the paper notes alternatives exist but does not explore them.
axioms (4)
  • domain assumption Shortest-path distance on the undirected piano subgraph yields musically contiguous wedges.
    Invoked via Erb’s Prop. IV.2 and used throughout Section 3.1–3.3; assumes the graph topology already encodes musical contiguity.
  • domain assumption Euclidean distance in the six-dimensional (k=3,4,7) Tonnetz embedding is a faithful proxy for harmonic proximity.
    Core of Section 3.2; standard neo-Riemannian claim, but treated as given rather than validated on the corpus.
  • ad hoc to paper Piecewise-constant approximation followed by nearest-pitch-class projection preserves a valid score in the original instrumentation and scale.
    Decoder step (Eq. 12 and Algorithm 2); ensures playability by construction but is an engineering choice, not a theorem.
  • standard math FA-greedy selection of the worst L2 region and best split is an adequate adaptive strategy.
    Taken directly from Erb 2023; no new optimality proof is supplied.
invented entities (1)
  • six-dimensional Tonnetz error space for graph-wedgelet partitioning no independent evidence
    purpose: Supplies the harmonic metric inside which L2 error is minimized when growing the binary wedge tree.
    The individual circular projections are classical; their concatenation as the sole error space for adaptive graph partitioning of music scores is introduced here.

reviewed 2026-07-13 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Tonnetz-Driven Graph Wedgelet for Harmonic Complexity Reduction in Music Scores." pith.science (2026). https://pith.science/paper/NKOWS2PY

@misc{pith2026260708806,
  author       = {Pith},
  title        = {Pith review of: Tonnetz-Driven Graph Wedgelet for Harmonic Complexity Reduction in Music Scores},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NKOWS2PY}},
  note         = {Machine review of arXiv:2607.08806}
}
Share X Bluesky LinkedIn Reddit HN
abstract

Heterogeneous graph built on notes, lyric syllables, and accompaniment events is a natural representation of symbolic music score, providing a substrate for both philological analysis and computational tasks. Music features are therefore well-captured by graph geometry and its properties. This representation has proved effective for analytical tasks as cadence detection, voice separation, and stylistic classification. In the present work, the reduction of harmonic complexity of a music score on graph, by preserving task-relevant information, relation between notes, and graph structure is investigated. A compression scheme for the piano subgraph of vocal-pianistic scores, built on binary wedge partitioning trees, is proposed. The wedges are generated through a fully adaptive greedy algorithm that recursively minimizes the $L^2$-error within a six-dimensional Tonnetz embedding of musical notes. The partitioning process employs a splitting criterion based on harmonic distance, resulting in regions that accurately reflect the intrinsic harmonic relationships among notes. The reconstructed music scores obtained through piecewise-constant functions and the mean values of the notes inside each wedge are used as a new simplified scores human-readable and playable. Some experiments on a corpus of symbolic music scores of three different composers are performed to assess the proposed approach.

Figures

Figures reproduced from arXiv: 2607.08806 by Elisa Francomano, Emmanuel Caronna, Silvia Licciardi.

Figure 1
Figure 1. Figure 1: W. A. Mozart, “Der Hölle Rache kocht in meinem Herzen” ( [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Binary wedge partitioning of the piano subgraph, shown on the score. For the selected example, r = 0.03 produces M = 4 final leaves. Starting from the root (the whole accompaniment), each node is split into two children by the FA-greedy criterion. Different notes assigned to different leaves of the tree are highlighted in distinct colors. Algorithm 1 Tonnetz-driven FA-greedy BWP encoding of the piano signa… view at source ↗
Figure 3
Figure 3. Figure 3: RMSE and std of the wedgelet approximation in the Tonnetz embedding versus the [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: One measure of an accompaniment reconstructed at compression ratios [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Octave equivalence: pitches separated by one or more octaves ( [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Piano-roll representation: columns index time steps, rows index MIDI pitches. Each [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: An Alberti bass figure: the notes of a single triad are not struck together but broken into [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: An ostinato accompaniment: a fixed pattern repeated with little or no change in its [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The twelve pitch classes on the chromatic circle. Chromatic distance (semitone separa [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

48 extracted references · 2 canonical work pages

  1. [1]

    Müller,Fundamentals of Music Processing: Audio, Analysis, Algorithms, Applications, Springer, Cham, 2015

    M. Müller,Fundamentals of Music Processing: Audio, Analysis, Algorithms, Applications, Springer, Cham, 2015

    2015

  2. [2]

    Goodfellow, Y

    I. Goodfellow, Y. Bengio, and A. Courville,Deep Learning, MIT Press, Cambridge, MA, 2016

    2016

  3. [3]

    Purwins, B

    H. Purwins, B. Li, T. Virtanen, J. Schlüter, S.-Y. Chang, and T. Sainath, Deep learning for audio signal processing,IEEE Journal of Selected Topics in Signal Processing, vol. 13, no. 2, pp. 206–219, 2019

    2019

  4. [4]

    Zhang, E

    H. Zhang, E. Karystinaios, S. Dixon, G. Widmer, and C. E. Cancino-Chacón, Symbolic music representations for classification tasks: a systematic evaluation, inProc. 24th Int. Soc. for Music Information Retrieval Conf. (ISMIR), 2023, pp. 848–858

    2023

  5. [5]

    Karystinaios, F

    E. Karystinaios, F. Foscarin, and G. Widmer, GraphMuse: a library for symbolic music graph processing, inProc. 25th Int. Soc. for Music Information Retrieval Conf. (ISMIR), 2024

    2024

  6. [6]

    Briot, G

    J.-P. Briot, G. Hadjeres, and F.-D. Pachet, Deep Learning Techniques for Music Generation, Springer, Cham, 2020

    2020

  7. [7]

    Good, MusicXML for notation and analysis, inThe Virtual Score: Representation, Re- trieval, Restoration, W

    M. Good, MusicXML for notation and analysis, inThe Virtual Score: Representation, Re- trieval, Restoration, W. B. Hewlett and E. Selfridge-Field, Eds. MIT Press, 2001, pp. 113–124

    2001

  8. [8]

    MIDI Manufacturers Association,The Complete MIDI 1.0 Detailed Specification, MIDI Man- ufacturers Association, Los Angeles, CA, 1996

    1996

  9. [9]

    Q. Kong, K. Choi, and Y. Wang, Large-scale MIDI-based composer classification, arXiv:2010.14805, 2020

    Pith/arXiv arXiv 2010

  10. [10]

    S. Kim, H. Lee, S. Park, J. Lee, and J. Choi, Deep composer classification using symbolic representation, arXiv:2010.00823, 2020

    Pith/arXiv arXiv 2010

  11. [11]

    C.-Z. A. Huanget al., Music Transformer: generating music with long-term structure, inInt. Conf. on Learning Representations (ICLR), 2019

    2019

  12. [12]

    Fradet, J.-P

    N. Fradet, J.-P. Briot, F. Chhel, A. El Fallah Seghrouchni, and N. Gutowski, MidiTok: a Python package for MIDI file tokenization, Late-Breaking Demo of ISMIR, 2021

    2021

  13. [13]

    Karystinaios and G

    E. Karystinaios and G. Widmer, Cadence detection in symbolic classical music using graph neural networks, inProc. 23rd Int. Soc. for Music Information Retrieval Conf. (ISMIR), 2022

    2022

  14. [14]

    Karystinaios and G

    E. Karystinaios and G. Widmer, Roman numeral analysis with graph neural networks: onset- wise predictions from note-wise features, inProc. 24th Int. Soc. for Music Information Re- trieval Conf. (ISMIR), 2023

    2023

  15. [15]

    Karystinaios, F

    E. Karystinaios, F. Foscarin, and G. Widmer, Musical voice separation as link prediction: modeling a musical perception task as a multi-trajectory tracking problem, inProc. Int. Joint Conf. on Artificial Intelligence (IJCAI), 2023

    2023

  16. [16]

    Jeong, T

    D. Jeong, T. Kwon, Y. Kim, and J. Nam, Graph neural network for music score data and modeling expressive piano performance, inProc. Int. Conf. on Machine Learning (ICML), 2019, pp. 3060–3070. 15

    2019

  17. [17]

    Licciardi, D

    S. Licciardi, D. Macchione, E. Caronna, and E. Francomano, Advanced Scientific Methodology Plays Rossini,Advances in Computational Science and Engineering, 2026, 8: 74-94. doi: 10.3934/acse.2026008

    work page doi:10.3934/acse.2026008 2026

  18. [18]

    Cilibrasi and P

    R. Cilibrasi and P. M. B. Vitányi, Clustering by compression,IEEE Trans. Inf. Theory, vol. 51, no. 4, pp. 1523–1545, 2005

    2005

  19. [19]

    C. Ding, A. Gorle, S. Bhattacharya, D. Hasteer, N. Sagan, and T. Weissman, LZMidi: compression-based symbolic music generation, arXiv:2503.17654, 2025

    Pith/arXiv arXiv 2025

  20. [20]

    G. Ala, M. L. Di Silvestre, E. Francomano and A. Tortorici, Wavelet-based efficient simulation ofelectromagnetictransientsinalightningprotectionsystem, IEEETransactionsonMagnetics 39(3), 2006, pp. 1257–1260. 2006

    2006

  21. [21]

    C. K. Chui,An Introduction to Wavelets, Academic Press, San Diego, CA, 1992

    1992

  22. [22]

    Strang and T

    G. Strang and T. Nguyen,Wavelets and Filter Banks, Wellesley–Cambridge Press, Wellesley, MA, 1996

    1996

  23. [23]

    Velarde, T

    G. Velarde, T. Weyde, and D. Meredith, An approach to melodic segmentation and classifica- tion based on filtering with the Haar wavelet,Journal of New Music Research, 2013

    2013

  24. [24]

    McLeod and K

    A. McLeod and K. Yoshii, The MIDI Degradation Toolkit: symbolic music augmentation and correction, Late-Breaking Demo of ISMIR, arXiv:2010.00059, 2020

    Pith/arXiv arXiv 2010

  25. [25]

    D. I. Shuman, S. K. Narang, P. Frossard, A. Ortega, and P. Vandergheynst, The emerging field of signal processing on graphs,IEEE Signal Processing Magazine, vol. 30, no. 3, pp. 83–98, 2013

    2013

  26. [26]

    Ortega, P

    A. Ortega, P. Frossard, J. Kovačević, J. M. F. Moura, and P. Vandergheynst, Graph signal processing: overview, challenges and applications,Proceedings of the IEEE, vol. 106, no. 5, pp. 808–828, 2018

    2018

  27. [27]

    Erb, Graph wedgelets: adaptive data compression on graphs based on binary wedge partitioning trees and geometric wavelets,IEEE Trans

    W. Erb, Graph wedgelets: adaptive data compression on graphs based on binary wedge partitioning trees and geometric wavelets,IEEE Trans. Signal Inf. Process. Netw., vol. 9, pp. 24–34, 2023

    2023

  28. [28]

    W. Erb, Split-and-merge segmentation of biomedical images using graph wedgelet decomposi- tions, inComputational Science and Its Applications – ICCSA 2025 Workshops, Lecture Notes in Computer Science, vol. 15899, Springer, 2026, pp. 252–263

    2025

  29. [29]

    Harte, M

    C. Harte, M. Sandler, and M. Gasser, Detecting harmonic change in musical audio, inProc. ACM Workshop on Audio and Music Computing Multimedia, pp. 21–26, 2006

    2006

  30. [30]

    Cohn, Introduction to neo-Riemannian theory: a survey and a historical perspective, Journal of Music Theory, vol

    R. Cohn, Introduction to neo-Riemannian theory: a survey and a historical perspective, Journal of Music Theory, vol. 42, no. 2, pp. 167–180, 1998

    1998

  31. [31]

    Karystinaios, J

    E. Karystinaios, J. Hentschel, M. Neuwirth, and G. Widmer, AnalysisGNN: unified music analysis with graph neural networks, arXiv:2509.06654, 2025

    Pith/arXiv arXiv 2025

  32. [32]

    W. L. Hamilton, R. Ying, and J. Leskovec, Inductive representation learning on large graphs, inAdvances in Neural Information Processing Systems (NeurIPS), 2017, pp. 1024–1034. 16

    2017

  33. [33]

    A. C. M. da Silva, D. F. Silva, and R. M. Marcacini, Heterogeneous graph neural network for music emotion recognition, inProc. 23rd Int. Soc. for Music Information Retrieval Conf. (ISMIR), 2022

    2022

  34. [34]

    A. C. M. da Silva, D. F. Silva, and R. M. Marcacini, Artist similarity based on heterogeneous graph neural networks,IEEE/ACM Trans. on Audio, Speech, and Language Processing, 2024

    2024

  35. [35]

    Ziv and A

    J. Ziv and A. Lempel, A universal algorithm for sequential data compression,IEEE Transac- tions on Information Theory, vol. 23, no. 3, pp. 337–343, 1977

    1977

  36. [36]

    Cilibrasi, P

    R. Cilibrasi, P. M. B. Vitányi, and R. de Wolf, Algorithmic clustering of music based on string compression,Computer Music Journal, vol. 28, no. 4, pp. 49–67, 2004

    2004

  37. [37]

    Euler,Tentamen Novae Theoriae Musicae, Saint Petersburg Academy of Sciences, 1739

    L. Euler,Tentamen Novae Theoriae Musicae, Saint Petersburg Academy of Sciences, 1739

  38. [38]

    von Oettingen,Harmoniesystem in dualer Entwickelung, W

    A. von Oettingen,Harmoniesystem in dualer Entwickelung, W. Gläser, Dorpat and Leipzig, 1866

  39. [39]

    Riemann, Ideen zu einer Lehre von den Tonvorstellungen,Jahrbuch der Musikbibliothek Peters, vol

    H. Riemann, Ideen zu einer Lehre von den Tonvorstellungen,Jahrbuch der Musikbibliothek Peters, vol. 21–22, pp. 1–26, 1914

    1914

  40. [40]

    Tymoczko,A Geometry of Music: Harmony and Counterpoint in the Extended Common Practice, Oxford University Press, 2011

    D. Tymoczko,A Geometry of Music: Harmony and Counterpoint in the Extended Common Practice, Oxford University Press, 2011

    2011

  41. [41]

    Chuan and D

    C.-H. Chuan and D. Herremans, Modeling temporal tonal relations in polyphonic music through deep networks with a novel image-based representation, inProc. 32nd AAAI Conf. on Artificial Intelligence, 2018

    2018

  42. [42]

    Kehoeet al., Exploring musical structure using Tonnetz lattice geometry and LSTMs, in Int

    E. Kehoeet al., Exploring musical structure using Tonnetz lattice geometry and LSTMs, in Int. Conf. on Computational Science (ICCS), Springer, 2020

    2020

  43. [43]

    K. Choi, G. Fazekas, and M. Sandler, Text-based LSTM networks for automatic music compo- sition, inProc. 1st Conf. on Computer Simulation of Musical Creativity (CSMC),Huddersfield, UK, 2016

    2016

  44. [44]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber, Long short-term memory,Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  45. [45]

    Licciardi, A

    S. Licciardi, A. Di Fatta, G. Ala, P. Romano, A. Imburgia, and E. Francomano, Artificial learning algorithms for partial discharge classification,IEEE Transactions on Magnetics, 2025, doi:10.1109/TMAG.2025.3641965

    work page doi:10.1109/tmag.2025.3641965 2025

  46. [46]

    G. Ala, P. Catrini, M. G. Ippolito, M. La Villetta, S. Licciardi, and R. Musca, Deep learning for smart grid and energy context, inProc. Asia Meeting on Environment and Electrical Engineering (EEE-AM), Hanoi, Vietnam, 2023, pp. 1–6

    2023

  47. [47]

    M. S. Cuthbert and C. Ariza,music21: a toolkit for computer-aided musicology and symbolic music data, inProc. 11th Int. Soc. for Music Information Retrieval Conf. (ISMIR), 2010, pp. 637–642

    2010

  48. [48]

    Piston and M

    W. Piston and M. DeVoto,Harmony, 5th ed., W. W. Norton, New York, 1987. 17

    1987

This paper was first reviewed by grok-4.5 on July 13, 2026.