Pith. sign in

REVIEW 3 major objections 6 minor 27 references

A New Framework For Spatial Modeling And Synthesis of Genome Sequence

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper argues that a genome sequence can be compressed into an ARIMA-GARCH statistical model, re-synthesized from that model, and paired with a 'counteracting' PDF built by flipping the sign of off-diagonal covariance entries.

desk verdict Routine ARIMA-GARCH fitting to Huffman-coded DNA; the 'counteracting PDF' is an artifact of the arbitrary numeric code, not a property of the genome. read the letter →

arxiv 1908.03342 v1 pith:WS53RRAL submitted 2019-08-09 q-bio.OT

Salman Mohamadi , Farhang Yeganegi , Hamidreza Amindavar This is my paper · ORCID
classification q-bio.OT
keywords genomesequencesynthesisARIMA-GARCHmodelingHuffmancodingHodrick-PrescottfilterGaussianmixturemodelstatisticalcounteractionHIVheteroskedasticity
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that a genome sequence—a string of A, T, G, and C letters—can be treated as a digital time series and reduced to a compact statistical model. Huffman coding maps the alphabet to decimals; a Hodrick-Prescott filter splits the decimals into a trend and a cyclic part; and an ARIMA-GARCH model fitted to the trend produces a formula that can synthesize new sequences with the same statistical features as the original. The paper then estimates the two-dimensional probability density of a sequence with a Gaussian mixture model and constructs a 'counteracting' density by reversing the signs of the off-diagonal covariance entries. If the framework works, it would allow generation of arbitrary-size ensembles for PDF estimation and propose sequences that are statistically, and possibly biologically, opposed to a target such as HIV.

What carries the argument

The load-bearing machinery is the ARIMA-GARCH synthesis equation plus the sign-flipped Gaussian mixture covariance. ARIMA (autoregressive integrated moving average) models linear dependence after differencing; GARCH (generalized autoregressive conditional heteroskedasticity) models the time-varying variance of the residuals. Together they produce a recurrence that generates a synthetic trend, which is added to the HP-filtered cyclic component and Huffman-decoded into base pairs. The counteracting-sequence construction is Eq. (6): from a Gaussian mixture fit to the two-dimensional marginal $p(x_1,x_2)$, a second mixture is formed with identical means, weights, and diagonal covariance entries, but off-diagonal entries $\Sigma'_{ij}=-\Sigma_{ij}$ for $i\neq j$; this is the object claimed to describe sequences negatively correlated with, and by extension counteracting, the original.

What would settle it

A direct test would be to synthesize candidate sequences from the sign-flipped Gaussian mixture PDF and measure whether they inhibit the biological activity of the original sequence, for instance HIV replication in cell culture; if they do not, the statistical-to-biological counteraction claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that the information in a genome sequence can be captured by a three-stage statistical pipeline. Huffman coding turns the four-letter alphabet into a decimal signal; the Hodrick-Prescott filter separates that signal into a heteroskedastic trend and a homoskedastic cyclic component; and an ARIMA-GARCH model fitted to the trend gives a compact mathematical recipe—for most of the 18 genes studied, ARIMA(2,1,2)-GARCH(1,1)—that can generate new decimal sequences statistically like the original. Adding the original cyclic component and Huffman-decoding yields a synthesized gene sequence of the same length. The paper further claims that if the original sequence's two-dimensional marginal PDF is estimated as a Gaussian mixture, then the PDF whose off-diagonal covariance entries have their signs flipped represents sequences with the strongest possible negative statistical correlation to the original; the conclusion states that this negative statistical correlation 'would suggest biological functional counteraction.'

Load-bearing premise

The paper's counteraction claim rests on the assumption that negative statistical correlation in a two-dimensional marginal of the Huffman-coded sequence corresponds to biological functional counteraction, a link asserted but not demonstrated.

Editorial extensions

If this is right

  • A gene sequence can be compressed into a handful of ARIMA-GARCH coefficients, so any of the 18 studied genes can be re-synthesized from its formula with the same linear and nonlinear statistical features.
  • The synthesis procedure provides an arbitrarily large ensemble of sequences from one gene, which the paper identifies as useful for empirical PDF estimation where only one copy of the sequence exists.
  • For every sequence whose two-dimensional marginal is fitted by a GMM, a companion 'counteracting' PDF is defined by Eq. (6), giving a search target for sequences statistically negatively correlated with the original.
  • The paper's conclusion extends this statistical counteraction to biological functional counteraction, meaning the framework is proposed as a route toward sequences that oppose a target genome, such as HIV.

Reading between the lines

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

  • The paper does not test whether known biological counteragents of HIV, such as neutralizing antibodies or siRNAs, actually show the sign-flipped covariance signature in their Huffman-encoded marginals; checking this would directly probe the statistical-to-biological link.
  • A testable extension is to generate candidate sequences by sampling from the sign-flipped Gaussian mixture PDF and screening them for the strongest negative correlation, then measuring any biological inhibition in cell culture.
  • The same pipeline could be applied to protein or RNA sequences by replacing the four-letter DNA alphabet with the appropriate larger alphabet, since the HP filter and ARIMA-GARCH steps do not depend on the alphabet size.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The manuscript proposes a statistical framework for modeling and synthesizing gene sequences. It encodes nucleotide sequences as decimal sequences via Huffman coding, decomposes them with an HP filter into trend and cyclic components, fits ARIMA-GARCH models to the trend, and then synthesizes new sequences by simulating from the fitted model and reattaching the original cyclic component. In a separate step, the paper estimates a two-dimensional marginal PDF of the encoded sequence using a Gaussian mixture model and constructs a 'counteracting' PDF by flipping the signs of off-diagonal covariance entries (Eq. 6). The authors apply the procedure to 18 human genes and an HIV sequence, and conclude that the negatively correlated PDF 'would suggest biological functional counteraction' (Section 5).

Significance. If the framework were validated, it could be a useful tool for generating synthetic genomic sequences with prescribed statistical properties, and the PDF approach might have heuristic value for dimensionality reduction in sequence analysis. The paper is transparent about its model-selection steps, presents AICc-based order choices, and describes a concrete algorithmic pipeline. However, the central load-bearing claims are not supported: the synthesis step is not validated against the original sequence statistics, and the counteracting-PDF construction is shown to depend on the arbitrary Huffman numeric labeling, making it a code artifact rather than a genome property. The biological conclusion in Section 5 is therefore unjustified at the formal level, not merely empirically under-supported.

major comments (3)
  1. [Section 4.3, Eq. (6)] The construction of the counteracting PDF by flipping off-diagonal covariance signs is not invariant under the Huffman labeling. The paper states in Section 2.1.1 that different mappings produce the same result, but the covariance Cov(phi(X_i), phi(X_{i+1})) used in the GMM changes sign and magnitude under arbitrary permutations of the numeric labels [0,1,2,3] assigned to A, T, G, C. Since the biological sequence is unchanged by such relabeling, the PDF p(x') defined by Eq. (6) is a property of the numerical code, not of the genome. This directly undermines the Section 5 conclusion that 'negative statistical correlation would suggest biological functional counteraction.'
  2. [Section 4.2] The equation displayed for gene CAV1 includes the term h_n on the left and labels it as 'synthesized sequence of Trend component,' but in Eqs. (2) and (3), h_n is the conditional variance of the GARCH process. This is a fundamental notation error: the synthesis procedure is never defined as a generative model for a new trend sequence. Moreover, no results are shown that compare the statistical properties of the synthesized sequences (autocorrelation, marginal distribution, spectral content) with the original gene sequences, so the claim in Section 4.1 that the synthesized sequences 'possess the linear and nonlinear characteristics of the original gene' is unvalidated.
  3. [Section 5 and Section 4.3] The leap from a statistically negatively correlated PDF to biological functional counteraction is unsupported. The manuscript provides no biological mechanism, no experimental assay, and no quantitative link between the sign of an off-diagonal covariance in a fitted two-dimensional GMM and any functional property of a genome. The phrase 'would suggest' in the conclusion acknowledges the speculative nature, but the paper presents this as a result. Given that the construction in Eq. (6) is code-dependent, this conclusion cannot be considered a valid inference from the presented analysis.
minor comments (6)
  1. [Section 2.1.1] The statement that 'different mappings, result in the same ways' is contradicted by the covariance-sign-flip construction in Section 4.3; the authors should either restrict the counteraction claim to the chosen encoding or prove invariance.
  2. [Abstract and Section 1] The term 'spatial modeling' is used, but the methods are standard time-series techniques; the connection to spatial modeling is never explained.
  3. [Figure 2] There are typographical errors in the axis labels ('veritacl' for 'vertical'), and the figure caption does not describe how normalized variance was computed.
  4. [Section 4.3] The sentence 'For instance to estimate PDF of a virus such as HIV with the length 9181 can be represented in such dimension' is incomplete and grammatically unclear.
  5. [Section 4.2] The sample of 120 synthesized base pairs is presented without any comparison to the original gene, so it does not substantiate the synthesis claim.
  6. [References] There are reference formatting inconsistencies, such as '[17, 18] [15,14]' in Section 2.2.2, and the numbering does not always match the citation order.

Circularity Check

2 steps flagged · score 6.0 of 10

The 'counteracting sequence' PDF is defined by flipping fitted covariance signs (Eq. 6), so its negative correlation is true by construction; the biological counteraction conclusion reduces to a definitional artifact.

  1. self definitional [Section 4.3, Eq. (6); Section 5 Conclusion]
    "Σii′ = Σii, Σij′ =−Σij, i ⁄=j, (6) that is, the diagonal entries of both covariance matrices are the same but their off-diagonal entries are opposite in sign to rationalize the utmost negative statistical correlation tieing x andx′. ... This negative statistical correlation would suggest biological functional counteraction."

    The PDF p(x′) said to represent sequences that 'counteract statistically' the original is not derived or tested; it is defined by Eq. (6) as the fitted GMM with every off-diagonal covariance sign negated. Any negative correlation between x and x′ is therefore imposed by construction, not discovered. The conclusion that this negative statistical correlation 'would suggest biological functional counteraction' then rests entirely on that definitional flip.

  2. fitted input called prediction [Section 4.1, final paragraph; Section 4.2]
    "Fitting these best models, allow us to synthesize artificial gene sequence for any given original gene, that possess the linear and nonlinear characteristics of the original gene."

    The 'synthesized' sequences are produced from the ARIMA-GARCH coefficients fitted to the very same original gene, and the synthesis procedure reattaches the original gene's cyclic component. Saying the synthesized sequence possesses the linear and nonlinear characteristics of the original gene is therefore a restatement of what was used to construct it, not an independent check. No held-out gene, cross-validation, or external benchmark is used to demonstrate that the synthesis generalizes beyond the fitted input, so the claimed capability reduces to simulation from a fitted model labeled as synthesis.

full rationale

The ARIMA-GARCH portion is a standard fit-and-simulate pipeline: parameters are estimated from a gene's Huffman-encoded trend, and new sequences are generated from that same fitted process. Standing alone, that is descriptive rather than circular, since the paper's stated goal there is synthesis rather than out-of-sample prediction. The load-bearing circularity is in Section 4.3 and the Conclusion. Equation (6) defines the 'counteracting' PDF by flipping the signs of off-diagonal covariance entries of the GMM fitted to the original sequence; therefore the asserted negative statistical correlation is a direct consequence of the definition. No sequence is generated from p(x′) and verified to be negatively correlated with the original, and no biological mechanism links a sign-flipped 2-D marginal covariance to functional counteraction. The arbirariness of the Huffman numeric labeling makes the sign-flip construction code-dependent, so the 'counteracting' PDF is not even a well-defined invariant of the genome sequence. The synthesis claim in Section 4.2 is also partly circular, because the synthesized sequence is built from the original gene's fitted coefficients and original cyclic component. Overall, the central biological conclusion reduces by construction to the covariance sign flip, giving a partial but significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central contribution is a pipeline of standard tools; every stage introduces fitted parameters (ARIMA and GARCH coefficients per gene, HP filter smoothing, GMM parameters) and domain assumptions about the meaningfulness of the numeric representation. No independent benchmarks are used.

free parameters (6)
  • ARIMA coefficients (per gene) = e.g., CAV1: -0.4369, 0.1312, -0.1334, -0.8718
    Fitted to each gene's trend component; central to synthesis.
  • GARCH coefficients (per gene) = e.g., CAV1: 0.8991, 0.0189, 7.40609e-06
    Fitted to ARIMA residuals; model volatility.
  • HP filter smoothing parameter lambda = not stated
    Required by HP filter to define trend/cyclic split; no value given.
  • ARIMA orders (p,d,q) and GARCH orders = selected by AICc
    Model order selection is data-driven and affects all fits.
  • Huffman letter-to-number mapping = A=0, T=1, G=2, C=3
    Arbitrary mapping; authors claim no effect but provide no proof.
  • GMM parameters (weights, means, covariances) = not reported for HIV
    GMM fit to two-dimensional marginal; used to define counteracting PDF.
assumptions (5)
  • domain assumption Genome sequence can be treated as a discrete time series
    Introduction states 'genome sequences could also be treated as a discrete signal'.
  • domain assumption Huffman-encoded decimal sequence preserves statistically relevant information
    Section 2.1.1 claims Huffman coding is 'gene-specific coding', but no evidence is provided.
  • domain assumption HP filter decomposes the sequence into heteroskedastic trend and homoskedastic cyclic components
    Section 2.1.2 and Fig. 2, but no statistical test supports this split.
  • standard math ARIMA-GARCH error terms are i.i.d. standard normal
    Section 2.2.1 states errors are assumed i.i.d. normal, a standard but untested assumption.
  • ad hoc to paper The two-dimensional marginal PDF of the numeric sequence is sufficient for counteraction
    Section 4.3 assumes a 2D marginal without justification for sufficiency or biological relevance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A New Framework For Spatial Modeling And Synthesis of Genome Sequence." pith.science (2026). https://pith.science/paper/WS53RRAL

@misc{pith2026190803342,
  author       = {Pith},
  title        = {Pith review of: A New Framework For Spatial Modeling And Synthesis of Genome Sequence},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WS53RRAL}},
  note         = {Machine review of arXiv:1908.03342}
}
read the original abstract

This paper provides a framework in order to statistically model sequences from human genome, which is allowing a formulation to synthesize gene sequences. We start by converting the alphabetic sequence of genome to decimal sequence by Huffman coding. Then, this decimal sequence is decomposed by HP filter into two components, trend and cyclic. Next, a statistical modeling, ARIMA-GARCH, is implemented on trend component exhibiting heteroskedasticity, autoregressive integrated moving average (ARIMA) to capture the linear characteristics of the sequence and later, generalized autoregressive conditional heteroskedasticity (GARCH) is then appropriated for the statistical nonlinearity of genome sequence. This modeling approach synthesizes a given genome sequence regarding to its statistical features. Finally, the PDF of a given sequence is estimated using Gaussian mixture model and based on estimated PDF, we determine a new PDF presenting sequences that counteract statistically the original sequence. Our strategy is performed on several genes as well as HIV nucleotide sequence and corresponding results is presented.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 27 canonical work pages

  1. [1]

    INTRODUCTION Considering the increasingly enormous amount of human and also animals genomic data and assuming that genome sequences could also be treated as a discrete signal, raise the idea that signal processing can play an important role in interpreting the genome information [1, 2]. Here we an- alyze and model genome sequence which is represented by s...

    work page

  2. [2]

    PROBLEM STA TEMENT AND FORMULA TION In order for statistical modeling of a genome sequence on a computer, decimal numbers are preferred, on the other hand, Huffman coding is a variable-length coding commonly ac- cepted for lossless data compression providing a short bi- nary sequence; in entropy sense, easily converted to a decimal number. By so doing, a ...

    work page Pith review arXiv 1908

  3. [3]

    FITTING THE BEST ARIMA-GARCH MODEL To perform our method, we selected 18 genes that have most interaction in human breast cancer; the sequence of each gene is downloaded from United State National Center for Biotech- Fig. 2 . Normalized Time Varying Variance of Trend and Cyclic components of the two genes, CA V1 (blue) and CD34 (orange); horizontal axis r...

    work page

  4. [4]

    Modeling the sequences of 18 genes of human genome Order of ARIMA model for each gene is selected based on best AIC

    RESULTS 4.1. Modeling the sequences of 18 genes of human genome Order of ARIMA model for each gene is selected based on best AIC. Best ARCH/GARCH model, is then fitted to residuals of ARIMA. Results of sample gene, gene CA V1, are presented in tables and figures. For all of genes we achieved mathematical formulation extracted from ARIMA- GARCH modeling; her...

    work page

  5. [5]

    counteracting the original sequence x is also modeled by a GMM as follows p(x′) = M∑ i=1 wiN (x′,µ i, Σ′ i), 0<w i < 1 (5) Fig. 4. Gaussian mixture approximation to PDF of HIV se- quence. Fig. 5. Gaussian mixture approximation to PDF of highly un- correlated HIV sequence; this PDF could be considered as the basis for searching sequences that are statistic...

    work page

  6. [6]

    Then perform sec- ond step, statistical modeling, to capture linear and nonlinear characteristics of genomic sequence

    CONCLUSION In this paper, a new framework is presented in which, firstly we propose a preprocessing step to prepare 18 genomic se- quences for further statistical modeling. Then perform sec- ond step, statistical modeling, to capture linear and nonlinear characteristics of genomic sequence. Furthermore, we pro- pose an approach to synthesize a new sequence...

    work page

  7. [7]

    Different genomic signal processing methods for eukaryotic gene prediction: A systematic review,

    Mai S Mabrouk, Safaa M Naeem, and Mohamed A El- dosoky, “Different genomic signal processing methods for eukaryotic gene prediction: A systematic review,” Biomedical Engineering: Applications, Basis and Com- munications, vol. 29, no. 01, pp. 1730001, 2017

    work page 2017

  8. [8]

    The role of signal-processing concepts in genomics and pro- teomics,

    PP Vaidyanathan and Byung-Jun Yoon, “The role of signal-processing concepts in genomics and pro- teomics,” Journal of the Franklin Institute , vol. 341, no. 1, pp. 111–135, 2004

    work page 2004

Show all 27 references
  1. [9]

    A novel numerical mapping method based on entropy for digitizing dna se- quences,

    Bihter Das and Ibrahim Turkoglu, “A novel numerical mapping method based on entropy for digitizing dna se- quences,” Neural Computing and Applications, pp. 1–9, 2017

    2017

  2. [10]

    Preprocessing and signal processing techniques on genomic data se- quences,

    Muhammaed Talha Naseem, KR Aravind Britto, Mustafa Musa Jaber, VS Balaji, G Rajkumar, K Narasimhan, V Elamaran, et al., “Preprocessing and signal processing techniques on genomic data se- quences,” Biomedical Research, pp. 1–1, 2017

    2017

  3. [11]

    Information-and communication the- ory in molecular biology,

    Martin Bossert, “Information-and communication the- ory in molecular biology,” 2017

    2017

  4. [12]

    A com- parative study on dna-based cryptosystem,

    M Thangavel, P Varalakshmi, and R Sindhuja, “A com- parative study on dna-based cryptosystem,” in Hand- book of Research on Recent Developments in Intelligent Communication Application, pp. 496–528. IGI Global, 2017

    2017

  5. [13]

    A periodic pattern of mrna secondary structure created by the genetic code,

    Svetlana A Shabalina, Aleksey Y Ogurtsov, and Niko- lay A Spiridonov, “A periodic pattern of mrna secondary structure created by the genetic code,”Nucleic Acids Re- search, vol. 34, no. 8, pp. 2428–2437, 2006

    2006

  6. [14]

    Genome-wide random regression analysis for parent-of-origin effects of body composition allometries in mouse,

    Jingli Zhao, Shuling Li, Lijuan Wang, Li Jiang, Run- qing Yang, and Yuehua Cui, “Genome-wide random regression analysis for parent-of-origin effects of body composition allometries in mouse,” Scientific Reports, vol. 7, pp. 45191, 2017

    2017

  7. [15]

    Dnabit compress–genome compression algorithm,

    Pothuraju Rajarajeswari and Allam Apparao, “Dnabit compress–genome compression algorithm,” Bioinfor- mation, vol. 5, no. 8, pp. 350, 2011

    2011

  8. [16]

    Optimized relative lempel-ziv compression of genomes,

    Shanika Kuruppu, Simon J Puglisi, and Justin Zo- bel, “Optimized relative lempel-ziv compression of genomes,” in Proceedings of the Thirty-Fourth Aus- tralasian Computer Science Conference-Volume 113 . Australian Computer Society, Inc., 2011, pp. 91–98

    2011

  9. [17]

    Dna coding using finite- context models and arithmetic coding,

    Armando J Pinho, Ant ´onio JR Neves, Carlos AC Bas- tos, and Paulo JSG Ferreira, “Dna coding using finite- context models and arithmetic coding,” in 2009 IEEE International Conference on Acoustics, Speech and Sig- nal Processing. IEEE, 2009, pp. 1693–1696

    2009

  10. [18]

    Comparative performance evaluation of svd-based im- age compression,

    Farhang Yeganegi, Vahid Hassanzade, and SM Ahadi, “Comparative performance evaluation of svd-based im- age compression,” in Electrical Engineering (ICEE), Iranian Conference on. IEEE, 2018, pp. 464–469

    2018

  11. [19]

    The practice of application-oriented personnel training in the course of information theory and coding,

    Jin He, Xuewen Ding, and Li Li, “The practice of application-oriented personnel training in the course of information theory and coding,” DEStech Transactions on Social Science, Education and Human Science, , no. eemt, 2017

    2017

  12. [20]

    Toward a better compression for dna sequences using huffman encoding,

    Anas Al-Okaily, Badar Almarri, Sultan Al Yami, and Chun-Hsi Huang, “Toward a better compression for dna sequences using huffman encoding,” Journal of Com- putational Biology, vol. 24, no. 4, pp. 280–288, 2017

    2017

  13. [21]

    The hodrick–prescott filter, the slutzky effect, and the distortionary effect of filters,

    Torben Mark Pedersen, “The hodrick–prescott filter, the slutzky effect, and the distortionary effect of filters,” Journal of Economic Dynamics and Control, vol. 25, no. 8, pp. 1081–1101, 2001

    2001

  14. [22]

    A modification of the hp filter aiming at reducing the end-point bias,

    Pierre-Alain Bruchez, “A modification of the hp filter aiming at reducing the end-point bias,” Swiss Financial Administration, Working Paper, 2003

    2003

  15. [23]

    Generalized autoregressive conditional heteroskedasticity,

    Tim Bollerslev, “Generalized autoregressive conditional heteroskedasticity,” Journal of econometrics , vol. 31, no. 3, pp. 307–327, 1986

    1986

  16. [24]

    Arima-garch model- ing for epileptic seizure prediction,

    Salman Mohamadi, Hamidreza Amindavar, and SM Ali Tayaranian Hosseini, “Arima-garch model- ing for epileptic seizure prediction,” in Acoustics, Speech and Signal Processing (ICASSP), 2017 IEEE International Conference on. IEEE, 2017, pp. 994–998

    2017

  17. [25]

    Detection and statistical modeling of birth- death anomaly,

    Salman Mohamadi, Farhang Yeganegi, and Nasser M Nasrabadi, “Detection and statistical modeling of birth- death anomaly,” arXiv preprint arXiv:1906.11788 , 2019

    1906 arXiv

  18. [26]

    Statistical methods for hiv dynamic studies in aids clinical trials,

    Hulin Wu, “Statistical methods for hiv dynamic studies in aids clinical trials,” Statistical methods in medical research, vol. 14, no. 2, pp. 171–192, 2005

    2005

  19. [27]

    Improved adaptive gaussian mixture model for background subtraction,

    Zoran Zivkovic, “Improved adaptive gaussian mixture model for background subtraction,” in Pattern Recogni- tion, 2004. ICPR 2004. Proceedings of the 17th Interna- tional Conference on. IEEE, 2004, vol. 2, pp. 28–31

    2004

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.