A Bayesian machine scientist to aid in the solution of challenging scientific problems
Open accessThe DOI matches its Crossref record; 44 of the 44 references it cites were found in a registry or an index.
Abstract
A Bayesian machine scientist uncovers closed-form mathematical models from data.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Figures

![Fig. 2 The prior probability distribution generates plausible expressions. (A) We extracted 4080 mathematical expressions from Wikipedia pages listed under the category “List of scientific equations named after people”, for example, the page devoted to “Newton’s law of universal gravitation” (see also table S2). (B) After parsing each expression (Materials and Methods), we counted the number of each operation in the expression and the square of these numbers. In Newton’s law of gravitation, we counted two products, one division, and one square. The complete list of operations is listed in table S1. Then, we parametrized the prior distributions p(fi) and fit the parameters so as to generate expressions that have realistic numbers (and squares) of each operation (Materials and Methods). (C to H) We ran the MCMC to draw sets of 4080 synthetic expressions from the prior distribution. We compared these sets of 4080 synthetic expressions with the 4080 empirical expressions obtained from Wikipedia. In particular, for each set of 4080 expressions, we counted the mean number 〈no〉 of each operation o per expression [(C) sine, (D) logarithm, (E) sum], or the mean 〈no²〉 of the square of those numbers (F to H) and plot their probability density function (PDF; pink). The red vertical line indicates the mean of the distribution, which is close to the empirical observation, represented by a dashed black line. The empirical value is always well within the expected ranges of the expressions generated by the MCMC. The results in the figure correspond to expressions with up to five variables and eight parameters.](https://figures.cochat.ai/paper-figures/d0f994b87cbe4b6f8c9b956669cd6462/a-bayesian-machine-scientist-to-aid-in-the-solution-of-challenging-scientific-figure-2.jpg)
![Fig. 3 Recovery of an arbitrary expression from synthetic data. We drew an expression with two variables (x1 and x2) and two parameters (θ1 and θ2) from the prior distribution p(fi); the drawn function is F(x1,x2; θ1,θ2) = x1(θ1 + x2) cos (x1)/[θ2 log (θ2)], with θ1 = −1.19 and θ2 = 0.29. With this function, we generated 400 synthetic points y^k=F(x1^k,x2^k;θ1,θ2)+ε^k, with (x1^k,x2^k) uniformly chosen in [−2,2]2, and ε ∼ N(0,1) normally distributed [points in (A) and (D) to (G); the color of the points corresponds to the value of y as indicated by the color bar]. We fed these synthetic data to the Bayesian machine scientist and run 2500 steps of the MCMC algorithm. (A) After 2500 steps, the most plausible model identified (blue surface) is f(x1, x2; θ1, θ2) = x1(c1 + c2x2) cos x1, which coincides with the correct model up to indeterminacies in the dependency on the parameters that cannot be resolved without varying their values (see movie S1 for the complete evolution of the MCMC). (B) The blue line indicates the evolution of the description length L during the MCMC (up to an irrelevant additive constant that comes from the prior normalization and therefore affects all points equally). Lower description lengths correspond to more plausible models. The MCMC starts from an expression with high description length, but after 1000 to 1500 steps, it equilibrates and samples from the stationary distribution p(fi∣D). (C) The blue line indicates the evolution of the mean absolute error of each model sampled by the MCMC with respect to the true model. The error is calculated over a grid in [−2,2]2, that is, not at the observed points. (D to G) The MCMC is prone to getting trapped in local maxima of the plausibility (local minima of the description length). To explore the expression space more efficiently, we used parallel tempering (38), a technique used in statistical physics to study disordered systems with rugged energy landscapes (Materials and Methods). Besides the main MCMC in (A), parallel tempering keeps a number of parallel MCMCs that sample the distributions p(fi,Tk) ∝ exp [−B(fi)/2Tk + log p(fi)] parametrized by “temperatures” Tk (Materials and Methods). In our case, higher temperatures correspond to simpler expressions. At each MCMC step, we attempted a swap of expressions at contiguous Tk using Metropolis’ rule. The evolution of the description length and the mean absolute error of the parallel MCMCs are represented in (B) and (C) with the same colors as in (D) to (G).](https://figures.cochat.ai/paper-figures/a3232a7e475f43718cd71b76083be883/a-bayesian-machine-scientist-to-aid-in-the-solution-of-challenging-scientific-figure-3.jpg)
![Fig. 4 Recovery of the differential equations governing the Rössler system. (A) We generated synthetic data for the Rössler system (equations shown in the panels) with a = b = 0.2 and c = 5.7. We assume that (x, y, z) are measured without error, but the functions that we aim to recover (x.,y.,z.) are measured with two levels of Gaussian noise [e.g., x.^k=F(x^k,y^k,z^k)+ε^k with εk ∼ N(0, σε)]: orange, σε = 1; blue, σε = 5. (B) Trajectories in velocity space (x.,y.,z.) for the levels of noise that we considered. (C to J) For each level of noise, we sampled 10,000 models for x. using the machine scientist and the same for y. and z.. (C and G), For each level of noise, we plot the most plausible model obtained. (D to F and H to J) We plot the distribution of description lengths (up to an irrelevant additive constant) for the models sampled. (D to F) For low noise, the most plausible model among those sampled always coincides with the true model, as indicated by the dashed vertical line. (H to J) For high noise, x. was recovered exactly, whereas for y. and z., the most plausible models corresponded to regularized versions of the true models, in which terms that are comparatively small because of the factor 0.2 are dropped. In all cases, the true model (dashed vertical line) is, at least, among the most plausible ones.](https://figures.cochat.ai/paper-figures/149444b9563248999b73ff6060385d75/a-bayesian-machine-scientist-to-aid-in-the-solution-of-challenging-scientific-figure-4.jpg)
![Fig. 5 The machine scientist gives accurate closed-form models and out-of-sample predictions for the Nikuradse dataset. The dataset contains measurements of the turbulent friction λ in rough pipes as a function of the Reynolds number xR and the inverse of the relative roughness xD. (A and B) We fed the machine scientist with the Nikuradse dataset (gray) and sample more than 18,000 models using MCMC as described in the text. The two graphs show the fit of two of the most plausible models (red), indicated at the top of each panel. For comparison, we show the best models identified by alternative machine scientists Eureqa (green) and EFS (orange). The inset in each graph shows the mean absolute error (MAE) of the machine scientist (MS), Eureqa (EU), and EFS (EF) (Materials and Methods), and error bars indicate the 95% confidence interval for the mean. (C and D) We fed the machine scientist with the data points in gray and sample 20,000 models using MCMC. We made predictions (red) for the unobserved data (black) using the median predictive model (see text). For comparison, we show the predictions of Eureqa, EFS, and Gaussian processes (GP) (Materials and Methods). The inset in each graph shows the mean absolute error for all approaches. (E) We considered, again, the models obtained from the whole data as in (A) and (B) but asked what the median predictive model predicts in the fully turbulent regime, well above the highest observed Reynold number (31). Despite the fact that individual sampled models have different behaviors in this region (fig. S5), the median predictive model [obtained as in (C) and (D)] predicts that turbulent friction stays constant at large Reynolds numbers, in agreement with theory (31–33). (F) The machine scientist also predicts the correct scaling (D/r)−1/3 for the turbulent friction in this regime (dashed line). (G) Scaling of the Nikuradse dataset proposed by Prandtl (with ks⁺(Re,r/D)=Re×√(λ/32)×D/r) (30, 33) and most plausible model identified by the machine scientist for the universal scaling function. (H) When unscaled, this simple universal function provides a good description of all the data.](https://figures.cochat.ai/paper-figures/77eb2836ac0a4d58b2cce7d719a210b1/a-bayesian-machine-scientist-to-aid-in-the-solution-of-challenging-scientific-figure-5.jpg)
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Both “machine scientists” (96, 97) and large language models hold the promise to help encapsulate the knowledge in the scientific literature for the use of scientists and the lay public.”
From The entities enabling scientific fraud at scale are large, resilient, and growing rapidly (2025) · cites this paper as [97]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Scikit-learn: Machine Learning in Python
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Müller, A., Nothman, J., Louppe, G., et al. (2012) · arXiv · doi:10.48550/arxiv.1201.0490 · Cited by 71,016
Estimating the Dimension of a Model
Schwarz, G. (1978) · The Annals of Statistics · doi:10.1214/aos/1176344136 · Cited by 47,033 · cited 3×
“where B( fi) is the Bayesian information criterion (BIC) of expression fi and can be readily calculated from the data (Supplementary text S1) (11, 17).”
BackgroundEquation of State Calculations by Fast Computing Machines
Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., Teller, A. H., & Teller, E. (1953) · The Journal of Chemical Physics · doi:10.1063/1.1699114 · Cited by 37,192
Information theory and statistical mechanics
Jaynes, E. T. (1957) · Physical Review · doi:10.1103/physrev.106.620 · Cited by 12,931 · cited 3×
“In particular, we choose a prior that generates expressions with the same average number of each type of operation [and their squares (23, 24)] as in the empirical corpus and, given this constraint, satisfies the maximum entropy principle (14, 23) (Materials and Methods; Supplementary text S2, fig.”
MethodNWChem: A comprehensive and scalable open-source solution for large scale molecular simulations
Valiev, M., Bylaska, E. J., Govind, N., Kowalski, K., Straatsma, T. P., Van Dam, H. J. J., Wang, D., Nieplocha, J., Apra, E., Windus, T. L., et al. (2010) · Computer Physics Communications · doi:10.1016/j.cpc.2010.04.018 · Cited by 5,779
Discovering governing equations from data by sparse identification of nonlinear dynamical systems
Brunton, S. L., Proctor, J. L., & Kutz, J. N. (2016) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1517384113 · Cited by 4,796 · cited 2×
“This approach is particularlysuited to learn differential equations ( 8 – 11 ), whose form often follows the assumption of linearity on relatively simple basis functions.”
BackgroundResultProbability theory the logic of science
Jaynes, E. T., & Bretthorst, G. L. (2003) · Cited by 4,654
Hoeting, J. A., Madigan, D., Raftery, A. E., & Volinsky, C. T. (1999) · Statistical Science · doi:10.1214/ss/1009212519 · Cited by 4,118
Probability Theory: The Logic of Science
Jaynes, E. T. (2003) · doi:10.1017/cbo9780511790423 · Cited by 4,045
An equation for continuous chaos
Rössler, O. E. (1976) · Physics Letters A · doi:10.1016/0375-9601(76)90101-8 · Cited by 4,012 · cited 1×
“Next, we investigate whether the machine is able to recover the differential equations that govern the Rössler system (25), andwhat is the effect of increasing observational noise in the equation discovery process.”
Background
Research works printed but not linked: 3
- W. La Cava, L. Spector, K. Danai, Proceedings of the Genetic and Evolutionary Computation Conference 2016, GECCO ‘16 (ACM, 2016), pp. 741–748.
- I. Arnaldo, K. Krawiec, U.-M. O’Reilly, Proceedings of the 2014 Annual Conference on Genetic and Evolutionary Computation, GECCO ‘14 (ACM, 2014), pp. 879–886.
- Hoeting J. A., Madigan D., Raftery A. E., Volinsky C. T., Bayesian model averaging: A tutorial. Stat. Sci. 14, 382417 (1999).
Cited by
Scientific discovery in the age of artificial intelligence.
Wang, H., Fu, T., Du, Y., Gao, W., Huang, K., Liu, Z., Chandak, P., Liu, S., Van Katwyk, P., Deac, A., et al. (2023) · Nature · doi:10.1038/s41586-023-06221-2 · Cited by 1,823
Discovering Symbolic Models from Deep Learning with Inductive Biases
Cranmer, M., Sánchez‐González, Á., Battaglia, P., Xu, R., Cranmer, K. S., Spergel, D. N., & Ho, S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.11287 · Cited by 250
“Other community packages such as [26, 27, 28, 29, 30, 31, 32, 33, 34], could likely also be used and achieve similar results (although [30] appears to not fit constants in equations, so could not be used here).”
Discovering causal relations and equations from data
Camps‐Valls, G., Gerhardus, A., Ninad, U., Varando, G., Martius, G., Balaguer‐Ballester, E., Vinuesa, R., Díaz, E., Zanna, L., & Runge, J. (2023) · Physics Reports · doi:10.1016/j.physrep.2023.10.005 · Cited by 201
Tenachi, W., Ibata, R., & Diakogiannis, F. I. (2023) · The Astrophysical Journal · doi:10.3847/1538-4357/ad014c · Cited by 199
“One can exploit such prior knowledge by formulating it as an in situ prior ( Guimerà et al. 2020; Kim et al. 2021 ) or by learning on it in a supervised manner using transformer learning techniques ( Biggio et al. 2021; Kamienny et al. 2022; Vastl et al. 2022; Bendinelli et al. 2023; Kamienny et…”
Data, measurement and empirical methods in the science of science
Liu, L., Jones, B. F., Uzzi, B., & Wang, D. (2023) · Nature Human Behaviour · doi:10.1038/s41562-023-01562-4 · Cited by 182
“Indeed, predicting the direction of new discoveries represents one of the most promising avenues for machine learning models, with neural networks being applied widely to biology 208 , physics 209,210 , mathematics 211 , chemistry 212 , medicine 213 and clinical applications 214 .”
BackgroundThe turning point and end of an expanding epidemic cannot be precisely forecast
Castro, M., Ares, S., Cuesta, J. A., & Manrubia, S. C. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2007868117 · Cited by 161
Rediscovering orbital mechanics with machine learning
Lemos, P., Jeffrey, N., Cranmer, M., Ho, S., & Battaglia, P. (2023) · Machine Learning Science and Technology · doi:10.1088/2632-2153/acfa63 · Cited by 197
Accelerating organic solar cell material's discovery: high-throughput screening and big data
Rodríguez‐Martínez, X., Pascual‐San‐José, E., & Campoy‐Quiles, M. (2021) · Energy & Environmental Science · doi:10.1039/d1ee00559f · Cited by 147
The entities enabling scientific fraud at scale are large, resilient, and growing rapidly
Richardson, R. A. K., Hong, S. S., Byrne, J. A., Stoeger, T., & Amaral, L. A. N. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2420092122 · Cited by 129
Perspectives for self-driving labs in synthetic biology
Martín, H. G., Radivojević, T., Zucker, J., Bouchard, K. E., Sustarich, J. M., Peisert, S., Arnold, D., Hillson, N. J., Babnigg, G., Martí, J. M., et al. (2023) · Current Opinion in Biotechnology · doi:10.1016/j.copbio.2022.102881 · Cited by 129
“[34] (Guimera et al): A stimulating demonstration of the power of “machine scientists”, able to extract closed mathematical models automatically out of data.”
Statistical inference links data and theory in network science
Peel, L., Peixoto, T. P., & Domenico, M. D. (2022) · Nature Communications · doi:10.1038/s41467-022-34267-9 · Cited by 142
“One possibility is to try to infer the dynamical rules together with the network itself, which may be possible to approximate in a nonparametric fashion, using, for example, Bayesian symbolic regression 70 .”
MethodComplex systems in the spotlight: next steps after the 2021 Nobel Prize in Physics
Bianconi, G., Arenas, À., Biamonte, J., Carr, L. D., Kahng, B., Kertész, J., Kurths, J., Lü, L., Masoller, C., Motter, A. E., et al. (2023) · Journal of Physics Complexity · doi:10.1088/2632-072x/ac7f75 · Cited by 126
Self-citationCombining data and theory for derivable scientific discovery with AI-Descartes
Cornelio, C., Dash, S., Austel, V., Josephson, T. R., Goncalves, J., Clarkson, K. L., Megiddo, N., Khadir, B. E., & Horesh, L. (2023) · Nature Communications · doi:10.1038/s41467-023-37236-y · Cited by 147
A review of mechanistic learning in mathematical oncology
Metzcar, J., Jutzeler, C. R., Macklin, P., Köhn‐Luque, A., & Brüningk, S. C. (2024) · Frontiers in Immunology · doi:10.3389/fimmu.2024.1363144 · Cited by 98
“Bayesian approaches applied to genetic programming have also proven successful in situations where existing non-Bayesian approaches failed (84).”
BackgroundUsing generative AI to investigate medical imagery models and datasets
Lang, O., Stupp, D., Traynis, I., Cole-Lewis, H. J., Bennett, C. R., Lyles, C. R., Lau, C., Irani, M., Semturs, C., Webster, D. R., et al. (2024) · EBioMedicine · doi:10.1016/j.ebiom.2024.105075 · Cited by 79
“Given this study was retrospective and used de-identi fi ed datasets, the need for further review was waived by the Advarra Institutional Review Board (IRB).”
MethodAutomating the practice of science: Opportunities, challenges, and implications
Musslick, S., Bartlett, L., Chandramouli, S., Dubova, M., Gobet, F. R., Griffiths, T. L., Hullman, J. R., King, R. D., Kutz, J. N., Lucas, C. G., et al. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2401238121 · Cited by 102
Learning dynamical models of single and collective cell migration: a review
Brückner, D. B., & Broedersz, C. P. (2024) · Reports on Progress in Physics · doi:10.1088/1361-6633/ad36d2 · Cited by 79
“…can be constrained by taking into account the symmetries of the system [85, 86], by applying Bayesian approaches to the fit-complexity trade-off [173], or by applying sparsity constraints to detect the relevant terms, such as SINDy [78, 79], which has recently been generalised to stochastic…”
MethodProbabilistic grammars for equation discovery
Brence, J., Todorovski, L., & Džeroski, S. (2021) · Knowledge-Based Systems · doi:10.1016/j.knosys.2021.107077 · Cited by 76
Process design within planetary boundaries: Application to CO2 based methanol production
Vázquez, D., & Guillén‐Gosálbez, G. (2021) · Chemical Engineering Science · doi:10.1016/j.ces.2021.116891 · Cited by 50
Predicting the photocurrent–composition dependence in organic solar cells
Rodríguez‐Martínez, X., Pascual‐San‐José, E., Fei, Z., Heeney, M., Guimerà, R., & Campoy‐Quiles, M. (2021) · Energy & Environmental Science · doi:10.1039/d0ee02958k · Cited by 47
Self-citationHuman mobility is well described by closed-form gravity-like models learned automatically from data
Cabanas-Tirapu, O., Danús, L., Moro, E., Sales‐Pardo, M., & Guimerà, R. (2025) · Nature Communications · doi:10.1038/s41467-025-56495-5 · Cited by 63
Self-citationWadekar, D. S., Thiele, L., Villaescusa-Navarro, F., Hill, J. C., Cranmer, M., Spergel, D. N., Battaglia, N., Anglés‐Alcázar, D., Hernquist, L., & Ho, S. (2023) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2202074120 · Cited by 50
Interpretable Machine Learning in Physics: A ReviewPreprint
Wetzel, S. J., Ha, S., Iten, R., Klopotek, M., & Liu, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.23616 · Cited by 56
“…search for mathematical expressions in symbolic regression can be made more efficient by using a corpus of closed-form mathematical models from Wikipedia to derive the posterior probability of each expression based on basic probabilistic principles and explicit approximations as discussed in [327].”
Loh, J. Y. Y., Wang, A., Mohan, A., Tountas, A. A., Gouda, A., Tavasoli, A., & Ozin, G. A. (2024) · Advanced Science · doi:10.1002/advs.202306604 · Cited by 41
SINDy for delay-differential equations: application to model bacterial zinc response
Sandoz, A., Ducret, V., Gottwald, G. A., Vilmart, G., & Perron, K. (2023) · Proceedings of the Royal Society A Mathematical Physical and Engineering Sciences · doi:10.1098/rspa.2022.0556 · Cited by 45
Catch-22s of reservoir computing
Zhang, Y., & Cornelius, S. P. (2023) · Physical Review Research · doi:10.1103/physrevresearch.5.033213 · Cited by 41
Simplifying polylogarithms with machine learning
Dersy, A., Schwartz, M. D., & Zhang, X. (2023) · International Journal of Data Science in the Mathematical Sciences · doi:10.1142/s2810939223500028 · Cited by 46
Percolation on feature-enriched interconnected systems
Artime, O., & Domenico, M. D. (2021) · Nature Communications · doi:10.1038/s41467-021-22721-z · Cited by 46
“In other dynamical processes different from the SIS model the functional expressions used here might not work, but eventually one can always follow similar steps or even apply the BMS to the two-dimensional empirical data to directly compute an approximation to P ( k , F ) .”
InfluentialTowards data-driven discovery of governing equations in geosciences
Song, W., Jiang, S., Camps‐Valls, G., Williams, M., Zhang, L., Reichstein, M., Vereecken, H., He, L., Hu, X., & Shi, L. (2024) · Communications Earth & Environment · doi:10.1038/s43247-024-01760-6 · Cited by 41
Fundamental limits to learning closed-form mathematical models from data
Fajardo-Fontiveros, O., Reichardt, I., Ríos, H. R. D. L., Duch, J., Sales‐Pardo, M., & Guimerà, R. (2023) · Nature Communications · doi:10.1038/s41467-023-36657-z · Cited by 42
Self-citationSaadun, A. J., Pablo-García, S., Paunović, V., Li, Q., Sabadell-Rendón, A., Kleemann, K., Krumeich, F., López, N., & Pérez-Ramírez, J. (2020) · ACS Catalysis · doi:10.1021/acscatal.0c00679 · Cited by 30
Evolving scientific discovery by unifying data and background knowledge with AI Hilbert
Cory-Wright, R., Cornelio, C., Dash, S., Khadir, B. E., & Horesh, L. (2024) · Nature Communications · doi:10.1038/s41467-024-50074-w · Cited by 53
Finding physical insights in catalysis with machine learning
Liu, C.-Y., & Senftle, T. P. (2022) · Current Opinion in Chemical Engineering · doi:10.1016/j.coche.2022.100832 · Cited by 29
“Other promising SR approaches are available, such as AI Feynman [32] and SR with Bayesian optimization [33], artificial bee colony [34], and deep learning [35], which have not yet been applied widely in the catalysis field.”
Efficient generator of mathematical expressions for symbolic regression
Mežnar, S., Džeroski, S., & Todorovski, L. (2023) · Machine Learning · doi:10.1007/s10994-023-06400-2 · Cited by 38
“Classical symbolic regression approaches (Cranmer, 2023; Guimerà et al., 2020; Schmidt & Lipson, 2009), based on evolutionary algorithms, use stochastic generators of expression trees: At the beginning, the expression trees are randomly sampled, and later on, they are transformed using the…”
BackgroundHidden layers of density dependence in consumer feeding rates
Stouffer, D. B., & Novak, M. (2021) · Ecology Letters · doi:10.1111/ele.13670 · Cited by 31
Vázquez, D., Guimerà, R., Sales‐Pardo, M., & Guillén‐Gosálbez, G. (2021) · Sustainable Production and Consumption · doi:10.1016/j.spc.2021.12.025 · Cited by 28
Self-citationPriors for symbolic regression
Bartlett, D. J., Desmond, H., & Ferreira, P. G. (2023) · Genetic and Evolutionary Computation Conference Companion (GECCO Companion) · doi:10.1145/3583133.3596327 · Cited by 27
Simulation-based inference on virtual brain models of disorders
Hashemi, M., Ziaeemehr, A., Woodman, M., Fousek, J., Petkoski, S., & Jirsa, V. (2024) · Machine Learning Science and Technology · doi:10.1088/2632-2153/ad6230 · Cited by 31
“…evaluates and incorporates uncertainties in the parameters and observations to drive meaningful conclusions from the data, with a broad range of applications (Bonomi et al 2016, Hashemi et al 2018, Khalvati et al 2019, Guimer ` a et al 2020, Broderick et al 2023, Sip et al 2023, Wang et al 2023).”
Optimal inflationary potentials
Sousa, T., Bartlett, D. J., Desmond, H., & Ferreira, P. G. (2024) · Physical review. D/Physical review. D. · doi:10.1103/physrevd.109.083524 · Cited by 26
Chatterjee, T., Shaw, A. D., Friswell, M. I., & Khodaparast, H. H. (2023) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2023.110858 · Cited by 26
Hybrid analytical surrogate-based process optimization via Bayesian symbolic regression
Jog, S., Vázquez, D., dos Santos, L. F., Caballero, J. A., & Guillén‐Gosálbez, G. (2023) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2023.108563 · Cited by 28
Machine-assisted agent-based modeling: Opening the black box
Taghikhah, F. R., Voinov, A. A., Filatova, T., & Polhill, G. J. (2022) · Journal of Computational Science · doi:10.1016/j.jocs.2022.101854 · Cited by 28
Dong, Z., Kong, K., Matchev, K. T., & Matcheva, K. (2023) · Physical review. D/Physical review. D. · doi:10.1103/physrevd.107.055018 · Cited by 26
“[11] Roger Guimerà, Ignasi Reichardt, Antoni AguilarMogas, Francesco A.”
BackgroundMachine learning uncovers analytical kinetic models of bioprocesses
Forster, T., Vázquez, D., Müller, C., & Guillén‐Gosálbez, G. (2024) · Chemical Engineering Science · doi:10.1016/j.ces.2024.120606 · Cited by 26
The Inefficiency of Genetic Programming for Symbolic Regression
Kronberger, G., de França, F. O., Desmond, H., Bartlett, D. J., & Kammerer, L. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70055-2_17 · Cited by 23
“Symbolic regression results for this dataset, predicting the relationship between turbulent friction λ and scaled roughness, have been reported in [23,49].”
Discovering physical laws with parallel symbolic enumeration
Ruan, K., Xu, Y., Gao, Z.-F., Liu, Y., Guo, Y., Wen, J.-R., & Sun, H. (2025) · Nature Computational Science · doi:10.1038/s43588-025-00904-8 · Cited by 38
“Genetic programming (GP) and related algorithms 1,8 explore expression spaces by applying genetic operations to tree-based representations, while Bayesian methods 9 employ probabilistic sampling to identify promising expressions.”
InfluentialThe need for verification in artificial intelligence-driven scientific discovery
Cornelio, C., Ito, T., Cory-Wright, R., Dash, S., & Horesh, L. (2026) · Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences · doi:10.1098/rsta.2024.0591 · Cited by 21
AI for identifying social norm violation
Neuman, Y., & Cohen, Y. (2023) · Scientific Reports · doi:10.1038/s41598-023-35350-x · Cited by 19
Distilling identifiable and interpretable dynamic models from biological data
Massonis, G., Villaverde, A. F., & Banga, J. R. (2023) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1011014 · Cited by 20
Pablo‐García, S., Sabadell‐Rendón, A., Saadun, A. J., Morandi, S., Pérez‐Ramírez, J., & López, N. (2022) · ACS Catalysis · doi:10.1021/acscatal.1c04345 · Cited by 19
Data-driven models for ground and excited states for Single Atoms on Ceria
Geiger, J., Sabadell‐Rendón, A., Daelman, N., & López, N. (2022) · npj Computational Materials · doi:10.1038/s41524-022-00852-1 · Cited by 18
AI-assisted discovery of quantitative and formal models in social science
Balla, J., Huang, S., Dugan, O., Dangovski, R., & Marin Soljacic, A. (2025) · Humanities and Social Sciences Communications · doi:10.1057/s41599-025-04405-x · Cited by 18
Vernet, A., & Fabregat, A. (2023) · Energies · doi:10.3390/en16062560 · Cited by 16
Negri, V., Vázquez, D., Sales‐Pardo, M., Guimerà, R., & Guillén‐Gosálbez, G. (2022) · ACS Omega · doi:10.1021/acsomega.2c04736 · Cited by 14
“Notably, for a fixed input flow, these variables are strongly connected to the cooling needs in the CO 2 compression stage and heating requirements in the stripper reboiler, which represent a large percentage of the overall cooling and heating, respectively.”
Self-citationBayesian Machine Scientist to Compare Data Collapses for the Nikuradse Dataset
Reichardt, I., Pallarès, J., Sales‐Pardo, M., & Guimerà, R. (2020) · Physical Review Letters · doi:10.1103/physrevlett.124.084503 · Cited by 16
Self-citationAlgebraic surrogate‐based process optimization using Bayesian symbolic learning
Forster, T., Vázquez, D., & Guillén‐Gosálbez, G. (2023) · AIChE Journal · doi:10.1002/aic.18110 · Cited by 13
Machine learning in physics: A short guide
Rodrigues, F. A. (2023) · Europhysics Letters (EPL) · doi:10.1209/0295-5075/ad0575 · Cited by 15
Geometric Complexity and the Information-Theoretic Comparison of Functional-Response Models
Novak, M., & Stouffer, D. B. (2021) · Frontiers in Ecology and Evolution · doi:10.3389/fevo.2021.740362 · Cited by 15
Racing Control Variable Genetic Programming for Symbolic Regression
Jiang, N., & Xue, Y. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i11.29187 · Cited by 12
Discovering network dynamics with neural symbolic regression
Yu, Z., Ding, J., & Li, Y. (2025) · Nature Computational Science · doi:10.1038/s43588-025-00893-8 · Cited by 28
“The process of finding the formula that best fits the given data is known as symbolic regression 10–14 .”
AIT Foundations of Structured Experience
Ruffini, G., & Lopez-Sola, E. (2022) · Journal of Artificial Intelligence and Consciousness · doi:10.1142/s2705078522500047 · Cited by 18
Cárdenas, S. D., Reznik, C. J., Ranaweera, R. S., Song, F., Chung, C. H., Fertig, E. J., & Gevertz, J. L. (2022) · npj Systems Biology and Applications · doi:10.1038/s41540-022-00244-7 · Cited by 12
Discovering interpretable models of scientific image data with deep learning
Soelistyo, C. J., & Lowe, A. R. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00682 · Cited by 13
“This includes the B ACON and D ALTON programs [21–23], and the series of genetic evolution algorithms developed afterward [9, 16, 20, 33, 40].”
MethodSymbolic Regression via Control Variable Genetic Programming
Jiang, N., & Xue, Y. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43421-1_11 · Cited by 11
Automated Scientific Discovery: From Equation Discovery to Autonomous Discovery Systems
Kramer, S., Cerrato, M., Brugger, J., Džeroski, S., & King, R. (2023) · arXiv · doi:10.48550/arxiv.2305.02251 · Cited by 10
“The Bayesian machine scientist [29] establishes the plausibility of models using explicit approximations to the exact marginal posterior over models and establishes its prior expectations about models by learning from a large empirical set of mathematical expressions.”
Forecasting Selected Commodities’ Prices with the Bayesian Symbolic Regression
Drachal, K., & Pawłowski, M. (2024) · International Journal of Financial Studies · doi:10.3390/ijfs12020034 · Cited by 11
Martingale Theory-Based Delay Bound Analysis for Multi-Hop Heterogeneous Satellite Networks
Zhang, Q., Zhu, Y., Zhou, D., Dong, Y., Guo, S., Lu, W., & Han, Z. (2024) · IEEE Transactions on Communications · doi:10.1109/tcomm.2024.3511710 · Cited by 10
“The MCMC method, which is widely employed to generate samples and construct Markov chains from arbitrary distributions, offers a potential solution to this issue [30], [31].”
Gravity-Bench-v1: A Benchmark on Gravitational Physics Discovery for AgentsPreprint
Koblischke, N., Jang, H., Kristen Menou, A., & Ali-Dib, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2501.18411 · Cited by 15
“…virtual environments for data-driven discovery or symbolic regression (e.g., Ma-jumder et al., 2024; Jansen et al., 2024; Udrescu & Tegmark, 2020; Guimer`a et al., 2020; Lemos et al., 2023; Shojaee et al., 2025), though they often focus on rediscovery of known phenomena, solving textbook-style…”
From reductionism to realism: holistic mathematical modelling for complex biological systems
Nartallo-Kaluarachchi, R., Lambiotte, R., & Goriely, A. (2025) · Journal of The Royal Society Interface · doi:10.1098/rsif.2025.0468 · Cited by 12
“Moreover, there exist some preliminary related approaches such as that of the Bayesian machine scientist [81], which explores and selects between candidate models using Monte-Carlo methods and Bayesian inference, penalising model-complexity, in order to perform system identification.”
Data-driven discovery of digital twins in biomedical research
Métayer, C., Ballesta, A., & Martinelli, J. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf722 · Cited by 15
Samal, S. K., Kumar, P., & Saha, S. K. (2023) · Physics of Fluids · doi:10.1063/5.0165497 · Cited by 9
Automating the discovery of partial differential equations in dynamical systems
Li, W., & Carvalho, R. (2024) · Machine Learning Science and Technology · doi:10.1088/2632-2153/ad682f · Cited by 10
Predicting Change in Emotion through Ordinal Patterns and Simple Symbolic Expressions
Neuman, Y., & Cohen, Y. (2022) · Mathematics · doi:10.3390/math10132253 · Cited by 9
Multi-Objective Optimization of a Hybrid Fossil/Renewable Carbon Methanol Cluster
Jog, S., Medrano‐García, J. D., Vázquez, D., & Guillén‐Gosálbez, G. (2025) · ACS Sustainable Chemistry & Engineering · doi:10.1021/acssuschemeng.4c06566 · Cited by 9
Symbolic Regression Approaches for the Direct Calculation of Pipe Diameter
Brkić, D., Praks, P., Praksová, R., & Kozubek, T. (2023) · Axioms · doi:10.3390/axioms12090850 · Cited by 8
Exploring the truth and beauty of theory landscapes with machine learning
Matchev, K. T., Matcheva, K., Pierre Ramond, B., & Verner, S. (2024) · Physics Letters B · doi:10.1016/j.physletb.2024.138941 · Cited by 11
SR-KAN: A Kolmogorov–Arnold Network guided symbolic regression framework
Bühler, M. A., & Guillén‐Gosálbez, G. (2026) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2026.109721 · Cited by 6
“Classical SR methods, such as genetic programming (GP) [4, 5, 6, 7], explore the space of symbolic expressions using evolutionary algorithms applied to expression trees [8].”
Limits of Inference in Complex Systems: When Stochastic Models Become Indistinguishable
Aguilar, J., Muñoz, M. A., & Azaele, S. (2026) · Physical Review X · doi:10.1103/tmkr-9kl2 · Cited by 6
“One of the preferred inference methodologies aims to identify the model that best describes the data, rather than estimating parameters within a constrained family of models (e.g., [23, 24]).”
AutoRA: Automated Research Assistant for Closed-Loop Empirical Research
Musslick, S., Andrew, B., Williams, C. C., Hewson, J. T. S., Li, S., Marinescu, I., Dubova, M., Dang, G. T., Strittmatter, Y., & Holland, J. G. (2024) · The Journal of Open Source Software · doi:10.21105/joss.06839 · Cited by 6
ON THE DEFINITION AND IMPORTANCE OF INTERPRETABILITY IN SCIENTIFIC MACHINE LEARNING
Rowan, C., & Doostan, A. (2026) · Journal of Machine Learning for Modeling and Computing · doi:10.1615/jmachlearnmodelcomput.2026062841 · Cited by 11
Niven, R. K. (2023) · Entropy · doi:10.3390/e25040617 · Cited by 6
Using machine learning to find exact analytic solutions to analytically posed physics problems
Ashhab, S. (2024) · Heliyon · doi:10.1016/j.heliyon.2024.e28124 · Cited by 8
“[16] and the abrupt transition between a low-noise regime and a high-noise regime observed in that context [21].”
BackgroundForster, T., Vázquez, D., Moreno-Palancas, I. F., & Guillén‐Gosálbez, G. (2024) · Computers & Chemical Engineering · doi:10.1016/j.compchemeng.2024.108630 · Cited by 5
Machine learning mathematical models for incidence estimation during pandemics
Fajardo-Fontiveros, O., Mattei, M., Burgio, G., Granell, C., Gómez, S., Arenas, À., Sales‐Pardo, M., & Guimerà, R. (2024) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1012687 · Cited by 6
Self-citationSeq-SVF: An unsupervised data-driven method for automatically identifying hidden governing equations
Wu, Z., Ye, H., Zhang, H., & Zheng, Y. (2023) · Computer Physics Communications · doi:10.1016/j.cpc.2023.108887 · Cited by 5
Sustainable Development Goals-Based Prospective Process Design Using Hybrid Modeling
Jog, S., Vázquez, D., dos Santos, L. F., Medrano‐García, J. D., & Guillén‐Gosálbez, G. (2025) · Industrial & Engineering Chemistry Research · doi:10.1021/acs.iecr.4c03563 · Cited by 8
Ai, L., Muggleton, S., Liang, S.-S., & Baldwin, G. S. (2025) · Machine Learning · doi:10.1007/s10994-025-06868-0 · Cited by 5
“Computational systems such as the BACON systems [42], the Lagramge systems [75], recent equation discovery systems [13,34,58] are primarily designed to formulate symbolic hypotheses from experimental results.”
A generalized model for estimating adsorption energies of single atoms on doped carbon materials
Minotaki, M. G., Geiger, J., Ruiz‐Ferrando, A., Sabadell‐Rendón, A., & López, N. (2024) · Journal of Materials Chemistry A · doi:10.1039/d3ta05898k · Cited by 7
Discovering Symmetries of ODEs by Symbolic Regression
Kahlmeyer, P., Merk, N., & Giesen, J. (2025) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v39i17.33948 · Cited by 5
The Agentic Leash: Extracting Causal Feedback Fuzzy Cognitive Maps with LLMs
Panda, A. K., Adigun, O., & Kosko, B. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-28469-3_34 · Cited by 4
Symbolically regressing dark matter halo profiles using weak lensing
Martín, A., Yasin, T., Bartlett, D. J., Desmond, H., & Ferreira, P. G. (2026) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/stag1394 · Cited by 5
Incorporating background knowledge in symbolic regression using a computer algebra system
Fox, C., Tran, N., Nacion, F. N., Sharlin, S., & Josephson, T. R. (2023) · Machine Learning: Science and Technology · doi:10.1088/2632-2153/ad4a1e · Cited by 10
“This improved consistency can be attributed to BSR’s tendency to generate higher-quality distributions of equations, as observed by Guimera [26].”
BackgroundMethodInfluentialStatistical patterns in the equations of physics and the emergence of a meta-law of nature
Constantin, A., Ferreira, P. G., Desmond, H., & Bartlett, D. J. (2026) · Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences · doi:10.1098/rsta.2025.0091 · Cited by 7
“…Lectures on Physics [4, 12]; one derived from the subset of physics equations in Wikipedia’s List of Scientific Equations Named after People [5, 13, 14]; one based on the comprehensive Encyclopaedia Inflationaris review of in-flationary cosmology [6]; and one compiled from The Cambridge…”
Passive scalar transfer rate at bubble interface in Carreau liquid in a transition regime
Kazemi, K., Vernet, A., Grau, F. X., Cito, S., & Fabregat, A. (2022) · International Journal of Multiphase Flow · doi:10.1016/j.ijmultiphaseflow.2022.104000 · Cited by 4
Creativity in AI: Progresses and Challenges
Ismayilzada, M., Paul, D., Bosselut, A., & van der Plas, L. (2024) · arXiv · doi:10.48550/arxiv.2410.17218 · Cited by 3
A simple preference aggregation rule explains how multidimensional identities shape social networks
Martín-Gutíerrez, S., van Dissel, M. N. C., & Karimi, F. (2026) · Communications Physics · doi:10.1038/s42005-026-02537-3 · Cited by 4
Knowledge-aware differential equation discovery with automated background knowledge extraction
Ivanchik, E., & Hvatov, A. (2025) · Information Sciences · doi:10.1016/j.ins.2025.122131 · Cited by 3
Dimension Reduction for Symbolic Regression
Kahlmeyer, P., Fischer, M. J., & Giesen, J. (2025) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v39i17.33947 · Cited by 3
Multiexpression Symbolic Regression and Its Circuit Design Case
Zhang, Y., Li, X., Hu, W., & Yen, G. G. (2024) · IEEE Transactions on Systems Man and Cybernetics Systems · doi:10.1109/tsmc.2024.3519675 · Cited by 3
Random Fuzzy-Rule Foams for Explainable AI
Panda, A. K., & Kosko, B. (2021) · Advances in intelligent systems and computing · doi:10.1007/978-3-030-81561-5_21 · Cited by 3
“So they do not yield explainable AI or XAI systems [1, 5, 6, 18].”
Background
Metadata
- Authors
- Guimerà, R., Reichardt, I., Aguilar-Mogas, A., Massucci, F. A., Miranda, M., Pallarès, J., & Sales-Pardo, M.
- Affiliations
- Institució Catalana de Recerca i Estudis Avançats (ES)
- Universitat Rovira i Virgili (ES)
- Department of Chemical Engineering, Universitat Rovira i Virgili, Tarragona 43007, Catalonia, Spain.
- SIRIS Academic (ES)
- Corresponding author
- Guimerà, Roger
- ORCID
- Venue
- Science Advances
- Published
- 2020-01-31
- Volume
- 6
- Issue
- 5
- Pages
- eaav6971-eaav6971
- ISSN
- 2375-2548 (ISSN-L)
- PMC
- PMC6994216
- OpenAlex
- W3003690504
- Cited by
- 136
- References
- 44
- Citations by year
- 2026: 17 · 2025: 24 · 2024: 32 · 2023: 31 · 2022: 15 · 2021: 13 · 2020: 4
- FWCI
- 11.11
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by-nc
- Locations
- Science Advances (publishedVersion, cc-by-nc)
- Europe PMC (PubMed Central) (publishedVersion, cc-by-nc)
- PubMed Central (publishedVersion, cc-by-nc)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- Rovira & Virgili repository (submittedVersion, cc-by)
- The Scholarship East Carolina University's Institutional Repository (East Carolina University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Ministerio de Economía y Competitividad, Gobierno de España (FIS2015-71563-ERC, FIS2016-78904-C3-P-1, DPI2016-75791-C2-1-P)
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Markov Chains and Monte Carlo Methods
- Physical Sciences › Computer Science › Artificial Intelligence › Gaussian Processes and Bayesian Inference
- Physical Sciences › Computer Science › Artificial Intelligence › Evolutionary Algorithms and Applications
- Keywords
- out-of-sample forecasting, interpretable models
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.