Context Steering: A New Paradigm for Compression-based Embeddings by Synthesizing Relevant Information Features

Ana Granados; Francisco de Borja Rodr\'iguez; Guillermo Sarasa

arxiv: 2508.14780 · v2 · submitted 2025-08-20 · 💻 cs.LG · cs.IT· math.IT

Context Steering: A New Paradigm for Compression-based Embeddings by Synthesizing Relevant Information Features

Guillermo Sarasa , Ana Granados , Francisco de Borja Rodr\'iguez This is my paper

Pith reviewed 2026-05-18 21:46 UTC · model grok-4.3

classification 💻 cs.LG cs.ITmath.IT

keywords context steeringcompression distanceembeddingsinductive representationhierarchical clusteringnormalized compression distance

0 comments

The pith

Context steering turns compression dissimilarities into inductive embeddings by guiding feature shaping.

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

Compression-based dissimilarities capture similarity via redundancies in data but often produce structures that do not match a given task such as classification or clustering. Context steering corrects this by actively examining how each object shapes the relational context inside a hierarchical clustering built on those distances. The resulting embeddings isolate and amplify class-distinctive information rather than accepting whatever hierarchy emerges. The method is demonstrated on text and audio collections with both Normalized Compression Distance and Relative Compression Distance, yielding representations that can be applied directly to new points. A reader would care because the technique converts a traditionally transductive distance matrix into an inductive, task-aligned embedding without requiring hand-crafted features.

Core claim

By systematically analyzing how each object influences the relational context within a hierarchical clustering framework on compression distances, context steering generates custom-tailored embeddings that isolate and amplify class-distinctive information, thereby producing robust task-oriented representations that can be applied inductively to unseen data.

What carries the argument

Context steering, the process of guiding feature shaping by examining each object's influence on the relational context inside a clustering hierarchy built from compression dissimilarities.

If this is right

The learned embeddings can be applied directly to new data points without recomputing a full distance matrix.
Classification and clustering metrics improve on heterogeneous collections such as text and audio.
The approach works with both Normalized Compression Distance and Relative Compression Distance.
It shifts compression-based methods from purely transductive to inductive use.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The steering idea could be tested with other dissimilarity measures beyond compression distances.
It may support sequential updates when data arrives over time rather than in a single batch.
The resulting embeddings could serve as input features for downstream neural models.

Load-bearing premise

Systematically analyzing each object's influence within a hierarchical clustering of compression distances isolates class-distinctive information without bias or dependence on the particular clustering algorithm used.

What would settle it

A dataset in which the steered embeddings produce lower classification accuracy or poorer cluster quality than the raw compression distance matrix on held-out points would falsify the central claim.

Figures

Figures reproduced from arXiv: 2508.14780 by Ana Granados, Francisco de Borja Rodr\'iguez, Guillermo Sarasa.

**Figure 1.** Figure 1: Hierarchical dendrograms obtained from the Easy dataset experiment, each containing 138 samples from two classes (one indicated by a small double vertical line (“∥”), and the other by an empty space (“ ”)), for different compressors. The title of each plot reports the silhouette coefficient along with the compression method used (e.g., Bz2, Lzma, Zlib, Rlzap: non-standardized and standardized. Every compre… view at source ↗

**Figure 2.** Figure 2: Illustration of the two different methods of object selection applied in our method. Initially, the matrix is symmetric and square, where both rows and columns represent the same set of objects (x1, x2, x3, and x4), and each cell encodes the pairwise compression distance between them. Applying HCA over this matrix, will sort the objects into a single hierarchical tree, hence producing what we understand as… view at source ↗

**Figure 3.** Figure 3: Basic illustration of the two key phases in Hierarchical Clustering Analysis (HCA). For this example, a R2 space is defined by the features y1 and y2. The different points are defined by xi . In the left part of the figure, we have the matrix of observations that defines the position of every xi point in R2 . Then, to compute the matrix of distances that HCA will use, a certain distance function δ is used.… view at source ↗

**Figure 4.** Figure 4: Overview of the proposed method. In this example, rows and columns are labeled as ai , bi , or ci to reflect the class membership of the corresponding objects, differing slightly from the matrix index notation used in the main text. The process begins (top-left) with an initial distance matrix M, computed using either NCD or NRC values, as defined throughout the paper. In Step1.a, the modified Euclidean di… view at source ↗

**Figure 1.** Figure 1: The reason behind this is that, in the training step, the number of samples is one fifth smaller, given that the [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗

**Figure 5.** Figure 5: Test F1 scores for various compressors and methods under the validation procedure. Each compressor uses the same number of features selected by our approach for that compressor, ensuring a fair comparison with alternative methods: kbest (anova), kbest (chi2), kbest (m.i.), knn, dummy, and random. Parameter selection for our approach was performed using a pool of multiple runs across different compressors a… view at source ↗

**Figure 6.** Figure 6: Medium dataset result distributions for each combination of classes during the validation process. The first and second rows display test F1 and silhouette scores, respectively. The modes named kbest refer to the feature-based alternatives to our approach, while dummy and random modes represent the baseline methods described previously. Knn represents a Nearest Neighbors computing the Euclidean distance ov… view at source ↗

**Figure 7.** Figure 7: Very hard dataset result distributions for each class combination using our approach and alternative methods. Each row represents a different file format (txt, wav, mp3, hex wap, and hex mp3), while each column corresponds to one of the five compressors evaluated (zlib, bz2, lzma, rlzap with external standardization, and rlzap with pipeline standardization). Due to incompatibilities already discussed in Se… view at source ↗

read the original abstract

Compression-based dissimilarities (CD) offer a flexible and domain-agnostic means of measuring similarity by identifying implicit information through redundancies between data objects. However, as similarity features are derived from the data, rather than defined as an input, it often proves difficult to align with the task at hand, particularly in complex clustering or classification settings. To address this issue, we introduce "context steering", a novel methodology that actively guides the feature-shaping process. Instead of passively accepting the emergent data structure (typically a hierarchy derived from clustering CDs), our approach "steers" the process by systematically analyzing how each object influences the relational context within a clustering framework. This process generates a custom-tailored embedding that isolates and amplifies class-distinctive information. We validate this supervised context-steering strategy using Normalized Compression Distance (NCD) and Relative Compression Distance (NRC) combined with hierarchical clustering, and evaluate the learned embeddings through both classification performance and cluster-quality metrics. Experiments on heterogeneous datasets-from text to real-world audio-show that the proposed approach yields robust task-oriented embeddings from compression dissimilarities, moving from traditional transductive uses of distance matrices to an inductive representation that can be applied to unseen data.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Context steering turns NCD into task-oriented inductive embeddings via hierarchy influence analysis, but the results look sensitive to the clustering subroutine chosen.

read the letter

The main point is that the authors take standard compression distances like NCD and NRC, run hierarchical clustering on them, then steer the embedding by measuring each object's effect on the relational structure to pull out class-relevant features. This produces representations that can be applied to new points instead of staying stuck in the original distance matrix. That inductive step is the actual novelty here, and it is a reasonable way to inject supervision without just fitting a classifier directly on the distances. The experiments on text and audio data report better classification accuracy and cluster quality scores than plain NCD baselines, which is concrete enough to show the idea has some traction in practice. The method stays grounded in the compression literature rather than inventing new distance functions from scratch. The soft spot is exactly the one the stress test flags: everything flows through a hierarchical clustering of the distance matrix, yet there is no check on whether single, complete, or Ward linkage (or even a flat clustering) produces the same steered features. If the extracted embedding shifts with that choice, the robustness claim does not hold up. The paper also does not spell out how the steering parameters are estimated on training data while keeping the test set fully out of the clustering step, which leaves a small opening for leakage. This is aimed at people already working with NCD or similar compression methods in domains like audio or text mining. A reader who wants a practical tweak to make those distances more task-aligned will get something usable from it. The idea is clear enough and the empirical support is present, so it deserves a serious referee rather than a desk reject.

Referee Report

2 major / 2 minor

Summary. The paper claims to introduce 'context steering' as a novel methodology for generating inductive, task-oriented embeddings from compression-based dissimilarities (NCD and NRC). Instead of passively using hierarchies from clustering these distances, the approach systematically analyzes each object's influence on the relational context to steer and amplify class-distinctive information features. This supervised strategy is validated on text and audio datasets via classification performance and cluster-quality metrics, enabling application to unseen data unlike traditional transductive distance-matrix uses.

Significance. If the central claim holds, the work could meaningfully advance compression-based ML by converting typically transductive dissimilarity measures into inductive representations that align with task goals in a domain-agnostic manner. The explicit steering via influence analysis in hierarchical clustering offers a fresh paradigm for synthesizing relevant features, with potential utility in heterogeneous settings like text and audio. Credit is due for the inductive framing and the attempt to move beyond passive data structures.

major comments (2)

In the context steering methodology description, the analysis of object influence on relational context within hierarchical clustering does not demonstrate or test invariance to the choice of linkage criterion or clustering algorithm. Different linkages (single, complete, Ward) produce qualitatively different dendrograms and cophenetic structures from the same NCD/NRC matrix; without evidence that the steered embeddings isolate the same class signal regardless, the robustness claim for task-oriented inductive representations is undermined.
Experimental validation section: the manuscript asserts that experiments on text and audio datasets validate the approach via classification performance and cluster-quality metrics, yet provides no quantitative results, error bars, baseline comparisons, or implementation specifics (e.g., how the embedding is constructed for inductive use on unseen data). This leaves the central empirical support for the claims with limited evidential grounding.

minor comments (2)

Abstract: including at least one concrete performance number or metric improvement would strengthen the summary of results.
Notation and reproducibility: the exact mathematical steps for deriving the final embedding vector from the influence analysis should be formalized (e.g., as an equation) to aid implementation and verification.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive criticism. We believe the suggested revisions will improve the clarity and robustness of our presentation of the context steering methodology. Below, we provide point-by-point responses to the major comments.

read point-by-point responses

Referee: In the context steering methodology description, the analysis of object influence on relational context within hierarchical clustering does not demonstrate or test invariance to the choice of linkage criterion or clustering algorithm. Different linkages (single, complete, Ward) produce qualitatively different dendrograms and cophenetic structures from the same NCD/NRC matrix; without evidence that the steered embeddings isolate the same class signal regardless, the robustness claim for task-oriented inductive representations is undermined.

Authors: We agree that the manuscript does not currently include tests for invariance to different linkage criteria or clustering algorithms. The context steering approach focuses on analyzing each object's influence on the relational context to amplify class-distinctive information, which is intended to be robust. To address this valid concern and support the robustness claim, we will revise the methodology section to include experiments with multiple linkage methods (single, complete, and Ward) and demonstrate that the steered embeddings consistently isolate the class signal across these variations. revision: yes
Referee: Experimental validation section: the manuscript asserts that experiments on text and audio datasets validate the approach via classification performance and cluster-quality metrics, yet provides no quantitative results, error bars, baseline comparisons, or implementation specifics (e.g., how the embedding is constructed for inductive use on unseen data). This leaves the central empirical support for the claims with limited evidential grounding.

Authors: We acknowledge that the experimental validation section in the current manuscript provides only high-level assertions of performance improvements without detailed quantitative results, error bars, baseline comparisons, or full implementation specifics for inductive use. This is a valid point that limits the evidential support. We will revise this section to include comprehensive quantitative results from the text and audio experiments, including error bars from multiple runs, comparisons to baselines such as unsteered NCD/NRC clustering, and a clear description of the inductive embedding construction for unseen data by applying the learned steering to new compression dissimilarities. revision: yes

Circularity Check

0 steps flagged

No significant circularity; method is a constructive algorithmic proposal

full rationale

The paper defines context steering as an explicit algorithmic process: compute NCD/NRC compression distances, perform hierarchical clustering, analyze each object's influence on the relational context (dendrogram structure), and use that analysis under supervision to isolate class-distinctive features into an embedding. This construction is presented as a novel, defined procedure that is then validated empirically on text and audio datasets via classification and cluster-quality metrics. No equation or step reduces by construction to a fitted parameter renamed as a prediction, no self-citation chain is load-bearing for the central claim, and the inductive claim for unseen data follows directly from the supervised steering definition rather than assuming the result. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that influence analysis within hierarchical clustering on compression distances can isolate class-distinctive features; no explicit free parameters or invented physical entities are described in the abstract.

axioms (1)

domain assumption Hierarchical clustering applied to compression dissimilarities produces a relational context whose object influences reliably reflect class structure.
Invoked when describing how the steering process generates custom-tailored embeddings that isolate class-distinctive information.

pith-pipeline@v0.9.0 · 5755 in / 1347 out tokens · 55910 ms · 2026-05-18T21:46:15.933767+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We introduce 'context steering'... by systematically analyzing how each object influences the relational context within a clustering framework... using Normalized Compression Distance (NCD) and Relative Compression Distance (NRC) combined with hierarchical clustering
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

HCA... linkage criterion... Ward... silhouette coefficient... G*_k = arg max SilhouetteCoefficient(Tk; G)

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

[1]

Poe’s corpus of short stories

E.A. Poe’s corpus of short stories. https://www.kaggle.com/datasets/leangab/poe-short-stories-corpuscsv

work page
[2]

https://www.gutenberg.org/

Project Gutenberg. https://www.gutenberg.org/

work page
[3]

In Cluster Analysis, chapter 4, pages 71–110

Hierarchical Clustering. In Cluster Analysis, chapter 4, pages 71–110. John Wiley & Sons, Ltd, 2011

work page 2011
[4]

Xeno-canto - Bird sounds from around the world

Vellinga , W. Xeno-canto - Bird sounds from around the world. Xeno-canto Foundation for Nature Sounds. Occurrence Dataset, 2024

work page 2024
[5]

Ali, Prakash Chourasia, and Murray Patterson

Sarwan Ali, Tamkanat E. Ali, Prakash Chourasia, and Murray Patterson. A Universal Non-parametric Approach for Improved Molecular Sequence Analysis. In De-Nian Yang, Xing Xie, Vincent S. Tseng, Jian Pei, Jen-Wei Huang, and Jerry Chun-Wei Lin, editors, Advances in Knowledge Discovery and Data Mining, pages 194–206, Singapore, 2024. Springer Nature

work page 2024
[6]

On Normalized Compression Distance and Large Malware, September 2015

Rebecca Schuller Borbely. On Normalized Compression Distance and Large Malware, September 2015

work page 2015
[7]

Carvalho, Susana Br ˜as, Jacqueline Ferreira, Sandra C

Jo˜ao M. Carvalho, Susana Br ˜as, Jacqueline Ferreira, Sandra C. Soares, and Armando J. Pinho. Impact of the Acquisition Time on ECG Compression-Based Biometric Identification Systems. In Lu ´ıs A. Alexandre, Jos´e Salvador S´anchez, and Jo˜ao M. F. Rodrigues, editors, Pattern Recognition and Image Analysis, pages 169–176, Cham, 2017. Springer Internation...

work page 2017
[8]

Extende d-alphab et finite-context models

Jo˜ao M Carvalho, Susana Br ´as, Diogo Pratas, Jacqueline Ferreira, Sandra C Soares, and Armando J Pinho. Extende d-alphab et finite-context models. Pattern Recognition Letters, 112:49–55, 2018

work page 2018
[9]

Towards FHR Biometric Identification: A Comparison between Compression and Entropy Based Approaches

Luisa Castro, Andreia Teixeira, Susana Br´as, Marcelo Santos, and Cristina Costa-Santos. Towards FHR Biometric Identification: A Comparison between Compression and Entropy Based Approaches. In 2018 IEEE 31st International Symposium on Computer-Based Medical Systems (CBMS), pages 440–441, June 2018

work page 2018
[10]

Common Pitfalls Using the Normalized Compression Distance: What to Watch Out for in a Compressor

Manuel Cebri´an, Manuel Alfonseca, and Alfonso Ortega. Common Pitfalls Using the Normalized Compression Distance: What to Watch Out for in a Compressor. Communications in Information & Systems, 5(4):367–384, January 2005

work page 2005
[11]

Vit´anyi

Rudi Cilibrasi and Paul M.B. Vit´anyi. Clustering by compression. IEEE Transactions on Information Theory, 51(4):1523–1545, April 2005

work page 2005
[12]

Cilibrasi and Paul M

Rudi L. Cilibrasi and Paul M. B. Vitanyi. Fast phylogeny of SARS-CoV-2 by compression. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 24(439), April 2022

work page 2022
[13]

On the use of normalized compression distances for image similarity detection

Dinu Coltuc, Mihai Datcu, and Daniela Coltuc. On the use of normalized compression distances for image similarity detection. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 20(99), February 2018

work page 2018
[14]

Cox, Andrea Farruggia, Travis Gagie, Simon J

Anthony J. Cox, Andrea Farruggia, Travis Gagie, Simon J. Puglisi, and Jouni Sir´en. RLZAP: Relative lempel-Ziv with adaptive pointers. In International Symposium on String Processing and Information Retrieval, volume 9954 LNCS, pages 1–14. Springer Verlag, 2016

work page 2016
[15]

Fast relative Lempel-Ziv self-index for similar sequences

Huy Hoang Do, Jesper Jansson, Kunihiko Sadakane, and Wing Kin Sung. Fast relative Lempel-Ziv self-index for similar sequences. Theoretical Computer Science, 532:14–30, 2014

work page 2014
[16]

H´ector Ferrada, Travis Gagie, Simon Gog, and Simon J. Puglisi. Relative Lempel-Ziv with Constant-Time Random Access. In Edleno Moura and Maxime Crochemore, editors, String Processing and Information Retrieval, pages 13–17, Cham, 2014. Springer International Publishing

work page 2014
[17]

Normalized graph compression distance – A novel graph matching framework

Anthony Gillioz and Kaspar Riesen. Normalized graph compression distance – A novel graph matching framework. Pattern Recognition Letters, 190:97–104, April 2025

work page 2025
[18]

Influence of music representation on compression-based clustering

Antonio Gonz´alez-Pardo, Ana Granados, David Camacho, and Francisco de Borja Rodr´ıguez. Influence of music representation on compression-based clustering. In IEEE Congress on Evolutionary Computation, pages 1–8, July 2010

work page 2010
[19]

Is the contextual information relevant in text clustering by compression? EXPERT SYSTEMS WITH APPLICATIONS, 39(10):8537–8546, August 2012

Ana Granados, David Camacho, and Francisco Borja Rodriguez. Is the contextual information relevant in text clustering by compression? EXPERT SYSTEMS WITH APPLICATIONS, 39(10):8537–8546, August 2012

work page 2012
[20]

Reducing the Loss of Information through Annealing Text Distortion

Ana Granados, Manuel Cebrian, David Camacho, and Francisco de Borja Rodriguez. Reducing the Loss of Information through Annealing Text Distortion. IEEE Transactions on Knowledge and Data Engineering , 23(7):1090–1102, July 2011

work page 2011
[21]

Neural Normalized Compression Distance and the Disconnect Between Compression and Classification, October 2024

John Hurwitz, Charles Nicholas, and Edward Raff. Neural Normalized Compression Distance and the Disconnect Between Compression and Classification, October 2024

work page 2024
[22]

Zhiying Jiang, Matthew Y . R. Yang, Mikhail Tsirlin, Raphael Tang, and Jimmy Lin. Less is More: Parameter-Free Text Classification with Gzip, December 2022. 22 arXiv Template A PREPRINT

work page 2022
[23]

Puglisi, and Justin Zobel

Shanika Kuruppu, Simon J. Puglisi, and Justin Zobel. Relative Lempel-Ziv Compression of Genomes for Large- Scale Storage and Retrieval. In Edgar Chavez and Stefano Lonardi, editors, String Processing and Information Retrieval, pages 201–206, Berlin, Heidelberg, 2010. Springer

work page 2010
[24]

Ming Li, Xin Chen, Xin Li, Bin Ma, and P.M.B. Vitanyi. The similarity metric.IEEE Transactions on Information Theory, 50(12):3250–3264, December 2004

work page 2004
[25]

Malware family classification via efficient Huffman features

Stephen O’Shaughnessy and Frank Breitinger. Malware family classification via efficient Huffman features. FORENSIC SCIENCE INTERNATIONAL-DIGITAL INVESTIGATION, 37(301192), September 2021

work page 2021
[26]

Functional balance at rest of hemispheric homologs assessed via normalized compression distance

Annalisa Pascarella, Vittoria Bruni, Karolina Armonaite, Camillo Porcaro, Livio Conti, Federico Cecconi, Luca Paulon, Domenico Vitulano, and Franca Tecchio. Functional balance at rest of hemispheric homologs assessed via normalized compression distance. Frontiers in Neuroscience, 17, January 2024

work page 2024
[27]

Pinho, Diogo Pratas, and Paulo J.S.G

Armando J. Pinho, Diogo Pratas, and Paulo J.S.G. Ferreira. Authorship Attribution Using Relative Compression. Data Compression Conference Proceedings, pages 329–338, December 2016

work page 2016
[28]

Silva, and Armando J

Diogo Pratas, Raquel M. Silva, and Armando J. Pinho. Comparison of compression-based measures with application to the evolution of primate genomes. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 20(393), June 2018

work page 2018
[29]

Ramos, Jo˜ao M

Mariana S. Ramos, Jo˜ao M. Carvalho, Armando J. Pinho, and Susana Br´as. On the Impact of the Data Acquisition Protocol on ECG Biometric Identification. Sensors 2021, Vol. 21, Page 4645, 21(14):4645, July 2021

work page 2021
[30]

Resende, Rolando Martins, and Luis Antunes

Joao S. Resende, Rolando Martins, and Luis Antunes. A survey on using kolmogorov complexity in cybersecurity. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 21(1196), December 2019

work page 2019
[31]

Rodriguez

Guillermo Sarasa, Ana Granados, and Francisco B. Rodriguez. An approach of algorithmic clustering based on string compression to identify bird songs species in xeno-canto database. In K Szczypiorski, editor, 2017 3rd International Conference on Frontiers of Signal Processing (Icfsp), pages 101–104, 345 E 47TH ST, NEW YORK, NY 10017 USA, 2017. IEEE / IEEE

work page 2017
[32]

Rodriguez

Guillermo Sarasa, Ana Granados, and Francisco B. Rodriguez. Automatic Treatment of Bird Audios by Means of String Compression Applied to Sound Clustering in Xeno-Canto Database. In V ˇera K˚urkov´a, Yannis Manolopou- los, Barbara Hammer, Lazaros Iliadis, and Ilias Maglogiannis, editors, Artificial Neural Networks and Machine Learning – ICANN 2018, pages 6...

work page 2018
[33]

Rodriguez

Guillermo Sarasa, Ana Granados, and Francisco B. Rodriguez. Algorithmic clustering based on string compression to extract P300 structure in EEG signals. Computer Methods and Programs in Biomedicine, 176:225–235, July 2019

work page 2019
[34]

Rodriguez

Guillermo Sarasa, Aaron Montero, Ana Granados, and Francisco B. Rodriguez. Compression-Based Clustering of Video Human Activity Using an ASCII Encoding. In Vˇera K˚urkov´a, Yannis Manolopoulos, Barbara Hammer, Lazaros Iliadis, and Ilias Maglogiannis, editors, Artificial Neural Networks and Machine Learning – ICANN 2018, pages 66–75, Cham, 2018. Springer I...

work page 2018
[35]

Vilmibm/lovecraftcorpus, October 2024

Nate Smith. Vilmibm/lovecraftcorpus, October 2024

work page 2024
[36]

Visual analysis of research paper collections using normalized relative compression

Pere-Pau Vazquez. Visual analysis of research paper collections using normalized relative compression. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 21(6), June 2019

work page 2019
[37]

Joe H. Ward. Hierarchical Grouping to Optimize an Objective Function. Journal of the American Statistical Association, 58(301):236–244, 1963

work page 1963
[38]

A Measure of Relative Entropy Between Individual Sequences with Application to Universal Classification

Jacob Ziv and Neri Merhav. A Measure of Relative Entropy Between Individual Sequences with Application to Universal Classification. IEEE Transactions on Information Theory, 39(4):1270–1279, 1993. 9 Notation We summarize here the notation used throughout the paper for quick reference. Summary of Symbols. • C(·): Generic compression algorithm. For two input...

work page 1993

[1] [1]

Poe’s corpus of short stories

E.A. Poe’s corpus of short stories. https://www.kaggle.com/datasets/leangab/poe-short-stories-corpuscsv

work page

[2] [2]

https://www.gutenberg.org/

Project Gutenberg. https://www.gutenberg.org/

work page

[3] [3]

In Cluster Analysis, chapter 4, pages 71–110

Hierarchical Clustering. In Cluster Analysis, chapter 4, pages 71–110. John Wiley & Sons, Ltd, 2011

work page 2011

[4] [4]

Xeno-canto - Bird sounds from around the world

Vellinga , W. Xeno-canto - Bird sounds from around the world. Xeno-canto Foundation for Nature Sounds. Occurrence Dataset, 2024

work page 2024

[5] [5]

Ali, Prakash Chourasia, and Murray Patterson

Sarwan Ali, Tamkanat E. Ali, Prakash Chourasia, and Murray Patterson. A Universal Non-parametric Approach for Improved Molecular Sequence Analysis. In De-Nian Yang, Xing Xie, Vincent S. Tseng, Jian Pei, Jen-Wei Huang, and Jerry Chun-Wei Lin, editors, Advances in Knowledge Discovery and Data Mining, pages 194–206, Singapore, 2024. Springer Nature

work page 2024

[6] [6]

On Normalized Compression Distance and Large Malware, September 2015

Rebecca Schuller Borbely. On Normalized Compression Distance and Large Malware, September 2015

work page 2015

[7] [7]

Carvalho, Susana Br ˜as, Jacqueline Ferreira, Sandra C

Jo˜ao M. Carvalho, Susana Br ˜as, Jacqueline Ferreira, Sandra C. Soares, and Armando J. Pinho. Impact of the Acquisition Time on ECG Compression-Based Biometric Identification Systems. In Lu ´ıs A. Alexandre, Jos´e Salvador S´anchez, and Jo˜ao M. F. Rodrigues, editors, Pattern Recognition and Image Analysis, pages 169–176, Cham, 2017. Springer Internation...

work page 2017

[8] [8]

Extende d-alphab et finite-context models

Jo˜ao M Carvalho, Susana Br ´as, Diogo Pratas, Jacqueline Ferreira, Sandra C Soares, and Armando J Pinho. Extende d-alphab et finite-context models. Pattern Recognition Letters, 112:49–55, 2018

work page 2018

[9] [9]

Towards FHR Biometric Identification: A Comparison between Compression and Entropy Based Approaches

Luisa Castro, Andreia Teixeira, Susana Br´as, Marcelo Santos, and Cristina Costa-Santos. Towards FHR Biometric Identification: A Comparison between Compression and Entropy Based Approaches. In 2018 IEEE 31st International Symposium on Computer-Based Medical Systems (CBMS), pages 440–441, June 2018

work page 2018

[10] [10]

Common Pitfalls Using the Normalized Compression Distance: What to Watch Out for in a Compressor

Manuel Cebri´an, Manuel Alfonseca, and Alfonso Ortega. Common Pitfalls Using the Normalized Compression Distance: What to Watch Out for in a Compressor. Communications in Information & Systems, 5(4):367–384, January 2005

work page 2005

[11] [11]

Vit´anyi

Rudi Cilibrasi and Paul M.B. Vit´anyi. Clustering by compression. IEEE Transactions on Information Theory, 51(4):1523–1545, April 2005

work page 2005

[12] [12]

Cilibrasi and Paul M

Rudi L. Cilibrasi and Paul M. B. Vitanyi. Fast phylogeny of SARS-CoV-2 by compression. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 24(439), April 2022

work page 2022

[13] [13]

On the use of normalized compression distances for image similarity detection

Dinu Coltuc, Mihai Datcu, and Daniela Coltuc. On the use of normalized compression distances for image similarity detection. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 20(99), February 2018

work page 2018

[14] [14]

Cox, Andrea Farruggia, Travis Gagie, Simon J

Anthony J. Cox, Andrea Farruggia, Travis Gagie, Simon J. Puglisi, and Jouni Sir´en. RLZAP: Relative lempel-Ziv with adaptive pointers. In International Symposium on String Processing and Information Retrieval, volume 9954 LNCS, pages 1–14. Springer Verlag, 2016

work page 2016

[15] [15]

Fast relative Lempel-Ziv self-index for similar sequences

Huy Hoang Do, Jesper Jansson, Kunihiko Sadakane, and Wing Kin Sung. Fast relative Lempel-Ziv self-index for similar sequences. Theoretical Computer Science, 532:14–30, 2014

work page 2014

[16] [16]

H´ector Ferrada, Travis Gagie, Simon Gog, and Simon J. Puglisi. Relative Lempel-Ziv with Constant-Time Random Access. In Edleno Moura and Maxime Crochemore, editors, String Processing and Information Retrieval, pages 13–17, Cham, 2014. Springer International Publishing

work page 2014

[17] [17]

Normalized graph compression distance – A novel graph matching framework

Anthony Gillioz and Kaspar Riesen. Normalized graph compression distance – A novel graph matching framework. Pattern Recognition Letters, 190:97–104, April 2025

work page 2025

[18] [18]

Influence of music representation on compression-based clustering

Antonio Gonz´alez-Pardo, Ana Granados, David Camacho, and Francisco de Borja Rodr´ıguez. Influence of music representation on compression-based clustering. In IEEE Congress on Evolutionary Computation, pages 1–8, July 2010

work page 2010

[19] [19]

Is the contextual information relevant in text clustering by compression? EXPERT SYSTEMS WITH APPLICATIONS, 39(10):8537–8546, August 2012

Ana Granados, David Camacho, and Francisco Borja Rodriguez. Is the contextual information relevant in text clustering by compression? EXPERT SYSTEMS WITH APPLICATIONS, 39(10):8537–8546, August 2012

work page 2012

[20] [20]

Reducing the Loss of Information through Annealing Text Distortion

Ana Granados, Manuel Cebrian, David Camacho, and Francisco de Borja Rodriguez. Reducing the Loss of Information through Annealing Text Distortion. IEEE Transactions on Knowledge and Data Engineering , 23(7):1090–1102, July 2011

work page 2011

[21] [21]

Neural Normalized Compression Distance and the Disconnect Between Compression and Classification, October 2024

John Hurwitz, Charles Nicholas, and Edward Raff. Neural Normalized Compression Distance and the Disconnect Between Compression and Classification, October 2024

work page 2024

[22] [22]

Zhiying Jiang, Matthew Y . R. Yang, Mikhail Tsirlin, Raphael Tang, and Jimmy Lin. Less is More: Parameter-Free Text Classification with Gzip, December 2022. 22 arXiv Template A PREPRINT

work page 2022

[23] [23]

Puglisi, and Justin Zobel

Shanika Kuruppu, Simon J. Puglisi, and Justin Zobel. Relative Lempel-Ziv Compression of Genomes for Large- Scale Storage and Retrieval. In Edgar Chavez and Stefano Lonardi, editors, String Processing and Information Retrieval, pages 201–206, Berlin, Heidelberg, 2010. Springer

work page 2010

[24] [24]

Ming Li, Xin Chen, Xin Li, Bin Ma, and P.M.B. Vitanyi. The similarity metric.IEEE Transactions on Information Theory, 50(12):3250–3264, December 2004

work page 2004

[25] [25]

Malware family classification via efficient Huffman features

Stephen O’Shaughnessy and Frank Breitinger. Malware family classification via efficient Huffman features. FORENSIC SCIENCE INTERNATIONAL-DIGITAL INVESTIGATION, 37(301192), September 2021

work page 2021

[26] [26]

Functional balance at rest of hemispheric homologs assessed via normalized compression distance

Annalisa Pascarella, Vittoria Bruni, Karolina Armonaite, Camillo Porcaro, Livio Conti, Federico Cecconi, Luca Paulon, Domenico Vitulano, and Franca Tecchio. Functional balance at rest of hemispheric homologs assessed via normalized compression distance. Frontiers in Neuroscience, 17, January 2024

work page 2024

[27] [27]

Pinho, Diogo Pratas, and Paulo J.S.G

Armando J. Pinho, Diogo Pratas, and Paulo J.S.G. Ferreira. Authorship Attribution Using Relative Compression. Data Compression Conference Proceedings, pages 329–338, December 2016

work page 2016

[28] [28]

Silva, and Armando J

Diogo Pratas, Raquel M. Silva, and Armando J. Pinho. Comparison of compression-based measures with application to the evolution of primate genomes. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 20(393), June 2018

work page 2018

[29] [29]

Ramos, Jo˜ao M

Mariana S. Ramos, Jo˜ao M. Carvalho, Armando J. Pinho, and Susana Br´as. On the Impact of the Data Acquisition Protocol on ECG Biometric Identification. Sensors 2021, Vol. 21, Page 4645, 21(14):4645, July 2021

work page 2021

[30] [30]

Resende, Rolando Martins, and Luis Antunes

Joao S. Resende, Rolando Martins, and Luis Antunes. A survey on using kolmogorov complexity in cybersecurity. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 21(1196), December 2019

work page 2019

[31] [31]

Rodriguez

Guillermo Sarasa, Ana Granados, and Francisco B. Rodriguez. An approach of algorithmic clustering based on string compression to identify bird songs species in xeno-canto database. In K Szczypiorski, editor, 2017 3rd International Conference on Frontiers of Signal Processing (Icfsp), pages 101–104, 345 E 47TH ST, NEW YORK, NY 10017 USA, 2017. IEEE / IEEE

work page 2017

[32] [32]

Rodriguez

Guillermo Sarasa, Ana Granados, and Francisco B. Rodriguez. Automatic Treatment of Bird Audios by Means of String Compression Applied to Sound Clustering in Xeno-Canto Database. In V ˇera K˚urkov´a, Yannis Manolopou- los, Barbara Hammer, Lazaros Iliadis, and Ilias Maglogiannis, editors, Artificial Neural Networks and Machine Learning – ICANN 2018, pages 6...

work page 2018

[33] [33]

Rodriguez

Guillermo Sarasa, Ana Granados, and Francisco B. Rodriguez. Algorithmic clustering based on string compression to extract P300 structure in EEG signals. Computer Methods and Programs in Biomedicine, 176:225–235, July 2019

work page 2019

[34] [34]

Rodriguez

Guillermo Sarasa, Aaron Montero, Ana Granados, and Francisco B. Rodriguez. Compression-Based Clustering of Video Human Activity Using an ASCII Encoding. In Vˇera K˚urkov´a, Yannis Manolopoulos, Barbara Hammer, Lazaros Iliadis, and Ilias Maglogiannis, editors, Artificial Neural Networks and Machine Learning – ICANN 2018, pages 66–75, Cham, 2018. Springer I...

work page 2018

[35] [35]

Vilmibm/lovecraftcorpus, October 2024

Nate Smith. Vilmibm/lovecraftcorpus, October 2024

work page 2024

[36] [36]

Visual analysis of research paper collections using normalized relative compression

Pere-Pau Vazquez. Visual analysis of research paper collections using normalized relative compression. Entropy. An International and Interdisciplinary Journal of Entropy and Information Studies, 21(6), June 2019

work page 2019

[37] [37]

Joe H. Ward. Hierarchical Grouping to Optimize an Objective Function. Journal of the American Statistical Association, 58(301):236–244, 1963

work page 1963

[38] [38]

A Measure of Relative Entropy Between Individual Sequences with Application to Universal Classification

Jacob Ziv and Neri Merhav. A Measure of Relative Entropy Between Individual Sequences with Application to Universal Classification. IEEE Transactions on Information Theory, 39(4):1270–1279, 1993. 9 Notation We summarize here the notation used throughout the paper for quick reference. Summary of Symbols. • C(·): Generic compression algorithm. For two input...

work page 1993