REVIEW 4 major objections 5 minor 1 cited by
Optimizing Carbon Footprint in ICT through Swarm Intelligence with Algorithmic Complexity
T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read A formula deriving swarm algorithms' CO2 emissions from their computational complexity, ranking them from 5.25% to 7.87%.
desk verdict A novel-looking complexity formula for swarm-algorithm CO2 that is actually an uncalibrated index; Table I is a normalized output of the formula, not a measurement. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the product $H(N_p) \cdot sf(N_i)$, with $H(N_p)=\prod_{i=1}^{N_p} i^i$ and $sf(N_i)=\prod_{j=1}^{N_i} j!$; these hyperfactorial and superfactorial functions convert particle and iteration counts into a sharply growing measure of computational work. The formula multiplies this measure by factor groups for hyperparameters, swarm topologies, and boundary-handling methods, and then by $t_{\mathrm{unit}} P_h \eta e_r$, so that algorithmic parameters alone yield a carbon figure. The mechanism's role is to replace empirical energy measurement with a parameter-derived complexity score that can be compared across algorithms on a percentage scale.
What would settle it
Run PSO, the Firefly Algorithm, and Fast Bacterial Swarming on identical hardware for identical optimization tasks, meter the energy in kWh, convert it to kg CO2 using regional grid intensity, and compare against the paper's predictions; a mismatch in ranking or ratio would show the formula is not measuring emissions.
Extended reading notes
Core claim
The paper's central claim is that the CO2 emissions of a swarm algorithm equal the product of a hyperfactorial of particle count, a superfactorial of iteration count, factor groups for hyperparameters, topologies, and boundary handling, unit computation time, hardware power, utilization, and regional emission factor, as stated in Eq. (1)–(4). Normalizing these results as percentages produces a unified scale on which the Bat Algorithm scores 5.25% and Fast Bacterial Swarming scores 7.87%. The authors report that this ranking shows algorithmic simplicity correlates with lower environmental impact: hybrid algorithms carry higher computational overhead, while stochastic and random-search methods are less resource-intensive. They further claim that this integration of swarm characteristics and CO2 emissions surpasses the threshold of existing models.
Load-bearing premise
The load-bearing premise is that Eq. (1)–(4) is proportional to actual computational energy consumption, so if that proportionality fails, the percentages in Table I are an arbitrary index rather than carbon emissions.
Editorial extensions
If this is right
- Algorithmic simplicity becomes a usable proxy for carbon impact: the paper's ranking places stochastic and random-search methods below hybrid ones.
- ICT teams could compare swarm algorithms on a single complexity-percentage axis without first running them on metered hardware.
- Hybrid algorithms, despite their performance advantages, carry an environmental overhead that should be weighed when selecting an optimizer.
- The formula gives a first deterministic route from hyperparameter settings to an emissions estimate, a step toward carbon-aware algorithm selection.
- Refinement of the formula for real-world applications is the paper's stated next step, implying the current percentages are a starting scale rather than settled constants.
Reading between the lines
- A natural extension not pursued in the paper is to calibrate the hyperfactorial and superfactorial exponents against metered energy data; until then, the percentages should be read as complexity indices, not validated emissions constants.
- The same parameter-to-emissions structure could be adapted to non-swarm machine-learning models by replacing the hyperfactorial-superfactorial product with an operation-count or FLOP-based complexity measure.
- Because the paper configures all algorithms with standard parameters, a sensitivity analysis varying hyperparameters would reveal whether the ranking is stable or an artifact of the chosen settings.
- If calibrated, the formula could feed cloud scheduling policies that route jobs to lower-footprint algorithms or greener regions, a practical consequence the paper leaves implicit.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a deterministic formula, Eq. (1)-(4) in Section II.B, for quantifying the CO2 emissions of swarm-intelligence algorithms. The formula multiplies a hyperfactorial of the number of particles, a superfactorial of the number of iterations, products of hyperparameter, topology, and boundary-handling factors, and a hardware/energy term (tunit, Ph, η, er). From this, Table I reports "Comp (%)" values between 5.25% and 7.87% for 35 algorithms and claims that simpler stochastic algorithms have lower emissions while hybrid algorithms have higher emissions. The stated contribution is an unprecedented framework for measuring the environmental impact of swarm algorithms in ICT. The paper gives no measured emissions, no parameter values, no normalization rule, no benchmark comparison, and no error analysis.
Significance. If the proposed formula were validated against measured energy consumption, a complexity-based proxy for comparing the carbon footprints of swarm algorithms would be a useful practical tool for green AI. The paper also usefully draws attention to the multiplicity of algorithmic factors, hyperparameters, topologies, and boundary-handling choices that can affect computational cost. However, the central claim is unsupported: Eq. (1)-(4) is introduced without derivation or calibration, the inputs to the formula are never specified, and Table I is presented without any reproducible protocol. As it stands, the manuscript provides an illustrative index defined by an arbitrary formula, not a measurement of CO2 emissions. The lack of empirical grounding, reproducibility, and internal consistency between the "prototype" language and the strong claims in the abstract prevents the result from being assessed at the level required for publication.
major comments (4)
- [II.B, Eq. (1)-(4)] The core claim that Eq. (1)-(4) quantifies CO2 emissions is not supported because the formula is ad hoc and uncalibrated. H(Np) and sf(Ni) are dimensionless combinatorial functions with super-exponential growth; for instance, increasing the particle count from Np to Np+1 multiplies H(Np) by (Np+1)^(Np+1), which for Np=30 is an astronomically large factor. No argument or empirical comparison establishes that energy consumption follows such growth. The factors hk, tl, and bm are also left completely unspecified in both value and units, yet they multiply a quantity asserted to be measured in kg CO2. Without a derivation, a calibration step, or a comparison to measured energy, the formula is an arbitrary index rather than a model of emissions.
- [Table I] The "Comp (%)" values in Table I are not reproducible from the paper. No values are given for Np, Ni, hk, tl, bm, tunit, Ph, η, or er for any of the 35 algorithms, and no normalization rule is specified that converts the kg-CO2 output of Eq. (1)-(4) into percentages. The reader therefore cannot reconstruct any entry of the table, test the ranking, or verify that the differences (e.g., 5.25% for Bat Algorithm vs. 7.87% for Fast Bacterial Swarming) reflect the formula rather than unspecified choices. This is a load-bearing gap because the entire empirical conclusion of the paper rests on these percentages.
- [III] Section III states that "computational experiments were conducted under controlled conditions, measuring energy consumption with precision," but the manuscript reports no measurement protocol, no hardware configuration, no raw energy measurements, no comparison between formula outputs and measured values, and no error bars. Since the central claim is that the authors "quantified the environmental impact of various swarm algorithms," the absence of any experimental data makes that claim untestable. Either the experiments and data must be reported in full, or the claims must be reduced to a theoretical proposal.
- [II.B and IV] The paper itself describes the formula as "a prototype" (Section II.B) and concludes that "future work should focus on refining the formula for real-world applications" (Section IV). This language is incompatible with the abstract's assertion that complexity percentages "ranged from 5.25% to 7.87%" and that the authors "were able to quantify the environmental impact of various swarm algorithms." The strong claims need to be either supported by validation or explicitly downgraded to an illustrative complexity index with no claim of quantitative accuracy.
minor comments (5)
- [Abstract and Introduction] The text contains numerous unusual and nonstandard phrasings such as "apodictic indication," "asyndetic integration," "veracious influence on carbonic effluvia," and "consanguineous relationship." These should be replaced with plain scientific language for clarity.
- [Eq. (1)-(4)] The equation is split across four numbered lines, making it appear as four separate statements. It should be presented as a single equation with the definition of all symbols immediately following.
- [Table I] The table caption and the table body are not fully self-contained: some algorithm names such as "Bees Algorithms" and "Wolf Search" are ambiguous, the abbreviations are only partially expanded, and the exact meaning of "Comp (%)" is not defined in the caption or the text.
- [Abstract and Methodology] The abstract states that the work constructs "a convex optimization problem," but no convex optimization problem is formulated or solved anywhere in the paper. Either provide the formulation or remove this claim.
- [References] Reference [2] is cited as "Itu-T" and would be better cited as the ITU-T L.1470 recommendation with the standard institutional authorship. Several other references are cited vaguely in the introduction without clear point-by-point support; consider tightening the related-work discussion.
Circularity Check
The CO2 'quantification' is the defining formula itself: Table I's percentages are a normalization of Eq. (1)-(4), so the emission ranking is forced by construction rather than measured.
-
self definitional
[Section II.B, Eqs. (1)-(4); Section III, Table I; Section IV]
"CO2 = H(Np) × sf(Ni) × (∏ hk) × (∏ tl) × (∏ bm) × tunit × Ph × η × er ... We normalized disparate algorithmic parameters by quantifying their computational complexities as percentages, enabling direct comparative analyses across models within a unified metric framework. ... The emissions reflect computational complexity, where higher percentages indicate more resource-intensive models."
Eq. (1)-(4) defines CO2 entirely as a product of H(Np), sf(Ni), and multiplicative factor groups, with no calibration, no measured energy data, and no independent validation. Table I reports 'complexity percentages' that are described as a normalization of these same computational complexities, so the reported percentages are just the defining formula rescaled. The conclusion that hybrid algorithms have higher emissions and stochastic algorithms lower emissions therefore follows from the multiplicative structure of the definition and the chosen category labels, not from any empirical measurement.
full rationale
The paper's central claim is that Eq. (1)-(4) quantifies CO2 emissions and that Table I ranks algorithms from 5.25% (Bat Algorithm) to 7.87% (Fast Bacterial Swarming). But the only definition of the reported quantity is Eq. (1)-(4) itself: CO2 is defined as H(Np)·sf(Ni)·(products of hyperparameter, topology, and boundary factors)·tunit·Ph·η·er. Table I is then presented as 'normalized... computational complexities as percentages' of this same quantity. No measured energy data, no calibration, and no independent benchmark are provided; the Discussion asserts that energy was 'measuring... with precision' but reports no kWh values, no parameter settings, and no normalization rule. Therefore the percentages and the resulting algorithm ranking are a rescaling of the defining formula. The conclusion that hybrid models emit more and stochastic models emit less restates the multiplicative structure of the formula and the category labels rather than being discovered from measurements. This is definitional circularity, not a minor self-citation issue. The few self-citations in the paper (e.g., [3]) are not load-bearing. The 'prototype' caveat and the statement that future work should refine the formula are honest, but they do not cure the fact that the reported 'prediction' is, by construction, the output of the same equation that defines the quantity.
Assumptions & free parameters
free parameters (4)
- Np and Ni per algorithm
- hk hyperparameter factors
- tl topology factors
- bm boundary-handling factors
assumptions (4)
- domain assumption Hyperfactorials and superfactorials of particle and iteration counts are proportional to computational work.
- ad hoc to paper Dimensionless factor groups for hyperparameters, topologies, and boundary handling multiply the complexity term and contribute to emissions multiplicatively.
- ad hoc to paper Normalized percentages of the formula support cross-algorithm comparison.
- domain assumption ICT contributes about 6% of global emissions.
Cite this review
Pith. "Pith review of Optimizing Carbon Footprint in ICT through Swarm Intelligence with Algorithmic Complexity." pith.science (2026). https://pith.science/paper/OBIJPE3R
@misc{pith2026250117166,
author = {Pith},
title = {Pith review of: Optimizing Carbon Footprint in ICT through Swarm Intelligence with Algorithmic Complexity},
year = {2026},
howpublished = {\url{https://pith.science/paper/OBIJPE3R}},
note = {Machine review of arXiv:2501.17166}
}
read the original abstract
Global emissions from fossil fuel combustion and cement production were recorded in 2022, signaling a resurgence to pre-pandemic levels and providing an apodictic indication that emission peaks have not yet been achieved. Significant contributions to this upward trend are made by the Information and Communication Technology (ICT) industry due to its substantial energy consumption. This shows the need for further exploration of swarm intelligence applications to measure and optimize the carbon footprint within ICT. All causative factors are evaluated based on the quality of data collection; variations from each source are quantified; and an objective function related to carbon footprint in ICT energy management is optimized. Emphasis is placed on the asyndetic integration of data sources to construct a convex optimization problem. An apodictic necessity to prevent the erosion of accuracy in carbon footprint assessments is addressed. Complexity percentages ranged from 5.25% for the Bat Algorithm to 7.87% for Fast Bacterial Swarming, indicating significant fluctuations in resource intensity among algorithms. These findings suggest that we were able to quantify the environmental impact of various swarm algorithms.
Forward citations
Cited by 1 Pith paper
-
A Biomimetic Way for Coral-Reef-Inspired Swarm Intelligence for Carbon-Neutral Wastewater Treatment
A coral-reef-inspired swarm neural network is claimed to outperform seven baselines on simulated wastewater treatment metrics, but the supporting code, data, and physical validation are not provided.
Reference graph
Works this paper leans on
-
[1]
Monitoring globa l carbon emissions in 2022,
Z. Liu, Z. Deng, S. Davis, and P . Ciais, “Monitoring globa l carbon emissions in 2022,” Nature Reviews Earth & Environment , vol. 4, no. 4, pp. 205–206, Mar. 2023. [Online]. Available: https://www.nature.com/articles/s43017-023-00406-z
work page 2022
-
[2]
L. Itu-T, “Greenhouse gas emissions trajectories for th e information and communication technology sector compatible with the un fccc paris agreement,” document itu-t 1. 1470 (01/20) , 2020
work page 2020
-
[3]
Critical review of biomimicry in the field of design,
N. Gerolimos, “Critical review of biomimicry in the field of design,” Hellanicus, 2022. [Online]. Available: http://hdl.handle.net/1161 0/25530
work page 2022
-
[4]
Swarm intelligence in cellular robo tic systems,
G. Beni and J. Wang, “Swarm intelligence in cellular robo tic systems,” in Robots and biological systems: towards a new bionics? Springer, 1993, pp. 703–712
work page 1993
-
[5]
D. K. K. Reddy, J. Nayak, H. Behera, V . Shanmuganathan, W. Viriy- asitavat, and G. Dhiman, “A systematic literature review on swarm intelligence based intrusion detection system: past, pres ent and future,” Archives of Computational Methods in Engineering , pp. 1–68, 2024
work page 2024
-
[6]
Qua ntifying the carbon emissions of machine learning,
A. Lacoste, A. Luccioni, V . Schmidt, and T. Dandres, “Qua ntifying the carbon emissions of machine learning,” 2019. [Online]. Available: https://arxiv.org/abs/1910.09700
arXiv 2019
-
[7]
D. K. K. Reddy, J. Nayak, H. S. Behera, V . Shanmuganathan, W. Viriyasitavat, and G. Dhiman, “A Systematic Literature R eview on Swarm Intelligence Based Intrusion Detection System: Past , Present and Future,” Archives of Computational Methods in Engineering , vol. 31, no. 5, pp. 2717–2784, Jul. 2024. [Online]. Availabl e: https://link.springer.com/10.100...
-
[8]
Modeling and estimation of co2 emissions in china based on artificial intelligence,
P . Wang, Y . Zhong, and Z. Y ao, “Modeling and estimation of co2 emissions in china based on artificial intelligence,” Computational Intelligence and Neuroscience , vol. 2022, no. 1, p. 6822467, 2022
work page 2022
Show all 18 references
-
[9]
Prediction of co2 emissions in china by g eneralized regression neural network optimized with fruit fly optimiza tion algo- rithm,
H. Y ue and L. Bu, “Prediction of co2 emissions in china by g eneralized regression neural network optimized with fruit fly optimiza tion algo- rithm,” Environmental Science and Pollution Research , vol. 30, no. 33, pp. 80 676–80 692, 2023
2023
-
[10]
Modeling and foreca sting CO2 emissions in China and its regions using a novel ARIMA-LS TM model,
T. Wen, Y . Liu, Y . h. Bai, and H. Liu, “Modeling and foreca sting CO2 emissions in China and its regions using a novel ARIMA-LS TM model,” Heliyon, vol. 9, no. 11, Nov. 2023, publisher: Elsevier. [Online]. Available: https://doi.org/10.1016/j.heliyon.2023.e21241
2023 doi
-
[11]
Driving factors of co2 emis sions: further study based on machine learning,
S. Li, Y . W. Siu, and G. Zhao, “Driving factors of co2 emis sions: further study based on machine learning,” Frontiers in Environmental Science , vol. 9, p. 721517, 2021
2021
-
[12]
Comparative analysis of co2 emissions and economic perfor mance in the united states and china: Navigating sustainable development in the climate change era,
K. Mehmood, S. Tauseef Hassan, X. Qiu, and S. Ali, “Comparative analysis of co2 emissions and economic perfor mance in the united states and china: Navigating sustainable development in the climate change era,” Geoscience Frontiers , vol. 15, no. 5, p. 101843, 2024. [Online]. ...
2024
-
[13]
Forecasting chinese provincial c o2 emissions: A universal and robust new-information-based g rey model,
S. Ding and H. Zhang, “Forecasting chinese provincial c o2 emissions: A universal and robust new-information-based g rey model,” Energy Economics , vol. 121, p. 106685, 2023. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0140988323001834
2023
-
[14]
Sim- plified swarm optimization for hyperparameters of convolut ional neural networks,
W.-C. Y eh, Y .-P . Lin, Y .-C. Liang, C.-M. Lai, and C.-L. H uang, “Sim- plified swarm optimization for hyperparameters of convolut ional neural networks,” Computers & Industrial Engineering , vol. 177, p. 109076, 2023
2023
-
[15]
Counting carb on: A survey of factors influencing the emissions of machine learning,
A. S. Luccioni and A. Hernandez-Garcia, “Counting carb on: A survey of factors influencing the emissions of machine learning,” 2 023. [Online]. Available: https://arxiv.org/abs/2302.08476
-
[16]
Bio-inspired computing: Algorithms revi ew, deep analysis, and the scope of applications,
A. Darwish, “Bio-inspired computing: Algorithms revi ew, deep analysis, and the scope of applications,” Future Computing and Informatics Journal, vol. 3, no. 2, pp. 231–246, 2018
2018
-
[17]
Hybrid genetic and penguin search optimiz ation algorithm (ga-pseoa) for efficient flow shop scheduling solu tions,
T. Mzili, I. Mzili, M. E. Riffi, D. Pamucar, V . Simic, L. Ab ualigah, and B. Almohsen, “Hybrid genetic and penguin search optimiz ation algorithm (ga-pseoa) for efficient flow shop scheduling solu tions,” Facta Universitatis, Series: Mechanical Engineering , vol. 22, no. 1, pp. ...
2024
-
[18]
A brief review of nature-inspired algorithms for optimization,
I. Fister Jr, X.-S. Y ang, I. Fister, J. Brest, and D. Fist er, “A brief review of nature-inspired algorithms for optimization,” arXiv preprint arXiv:1307.4186, 2013
2013 arXiv
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.