Journal article · 2026 · Open access

Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression

Ezequiel L´opez-Rubio, Mario Pascual-Gonz´alez, Karl Thurnhofer‐Hemsi

IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, 1–12 · Published 12 November 2026 · DOI 10.1109/tpami.2026.3752551

Open full textCite this paper

In brief

“Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression” is a 2026 journal article by Ezequiel L´opez-Rubio and 2 co-authors, published in IEEE Transactions on Pattern Analysis and Machine Intelligence. OpenAlex counts 0 citations as of 18 November 2026. It is open access under a CC BY licence: the full text is free to read at IEEE Transactions on Pattern Analysis and Machine Intelligence.

Citations
0
Authors
3

Abstract

Symbolic Regression (SR) aims to discover a closed-form mathematical expression that fits an observed dataset by representing those expressions as Directed Acyclic Graphs (DAGs). A DAG with k internal nodes admits Θ(k!) equivalent node orderings of the same expression, so structurally identical candidates repeatedly reach the fitness evaluator of the solver model, consuming search budget. We introduce ISALSR (Instruction Set and Language for Symbolic Regression), an encoding that maps every labeled DAG to a canonical string over a two-tier alphabet of structural and operation-labeled tokens. Canonical strings are reached via a greedy algorithm guided by 1-Weisfeiler-Leman (1-WL) subtree hashing. We prove that the canonical string is a complete labeled-DAG invariant under a reachability condition: two expressions share their string if and only if their DAGs are isomorphic. ISALSR integrates at the evaluation boundary and needs no change to the host search algorithm. It has been validated on a systematic enumerator (UDFS) and an evolutionary search (Bingo) over a 70-problem suite assembled from nine published benchmarks, against both the host solvers' native representation and a naive hash of a fixed-order DAG serialization. Canonicalization eliminates a mean of 38.1% of evaluations on UDFS and 43.7% on Bingo, on every one of the 70 problems under both hosts. A paired test across problems puts its reduction factor above that of the naive hash at a Cohen's d of 2.54 and 7.05, both at p < 10-12. Canonicalization consumes 0.04% of run time on UDFS and 16.1% on Bingo. Deduplication is therefore close to free where a fitness evaluation costs orders of magnitude more than a canonical string, and expensive where the two are comparable. ISALSR is the first instruction-based graph encoding to carry operation labels, edge direction and a non-commutative operator, and because the canonical form is itself a string, it also gives sequence models a target that is invariant to node renumbering. Our code is publicly available at https://github.com/MarioPasc/IsalSR.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

L´opez-Rubio, E., Pascual-Gonz´alez, M., & Thurnhofer‐Hemsi, K. (2026). Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression. IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, 1–12. https://doi.org/10.1109/tpami.2026.3752551

MLA 9

L´opez-Rubio, Ezequiel, et al. “Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression.” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PP, 2026, pp. 1–12. https://doi.org/10.1109/tpami.2026.3752551.

Chicago

L´opez-Rubio, Ezequiel, Mario Pascual-Gonz´alez, and Karl Thurnhofer‐Hemsi. 2026. “Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression.” IEEE Transactions on Pattern Analysis and Machine Intelligence PP: 1–12. https://doi.org/10.1109/tpami.2026.3752551.

Harvard

L´opez-Rubio, E., Pascual-Gonz´alez, M. and Thurnhofer‐Hemsi, K. (2026) ‘Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression’, IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, pp. 1–12. doi:10.1109/tpami.2026.3752551.

IEEE

E. L´opez-Rubio, M. Pascual-Gonz´alez, and K. Thurnhofer‐Hemsi, “Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PP, pp. 1–12, 2026, doi: 10.1109/tpami.2026.3752551.

BibTeX

@article{lopezrubio2026representation,
  title     = {{Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression}},
  author    = {L´opez-Rubio, Ezequiel and Pascual-Gonz´alez, Mario and Thurnhofer‐Hemsi, Karl},
  journal   = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year      = {2026},
  volume    = {PP},
  pages     = {1--12},
  doi       = {10.1109/tpami.2026.3752551},
  url       = {https://doi.org/10.1109/tpami.2026.3752551}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Evolutionary Algorithms and Applications

  • symbolic regression
  • directed acyclic graphs
  • graph isomorphism
  • evaluation cost reduction
  • evolutionary search
  • UDF
  • bingo

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

L´opez-Rubio, E., Pascual-Gonz´alez, M., & Thurnhofer‐Hemsi, K. (2026). Representation of Directed Acyclic Graphs by Sequences of Instructions for Symbolic Regression. IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, 1–12. https://doi.org/10.1109/tpami.2026.3752551

How many times has this paper been cited?

0 times, according to OpenAlex on 18 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at IEEE Transactions on Pattern Analysis and Machine Intelligence: sign in and open the full text from this page.

Who wrote this paper?

Ezequiel L´opez-Rubio, Mario Pascual-Gonz´alez, Karl Thurnhofer‐Hemsi.

What is the DOI of this paper?

10.1109/tpami.2026.3752551, permanently at https://doi.org/10.1109/tpami.2026.3752551. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 18 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.