Random Error Sampling-based Recurrent Neural Network Architecture Optimization

Andr\'es Camero; Enrique Alba; Jamal Toutouh

REVIEW 4 major objections 5 minor 50 references

Random Error Sampling-based Recurrent Neural Network Architecture Optimization

T0 review · 4 major / 5 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 A training-free score can replace error-based fitness in RNN architecture search, matching trained search at roughly half the optimization time.

desk verdict A new and credible training-free fitness for RNN architecture search, with the central proxy assumption left only indirectly tested and one unfair baseline muddying the main comparison. read the letter →

arxiv 1909.02425 v3 pith:A3ACLLUZ submitted 2019-09-04 cs.NE cs.LG

Andrés Camero , Jamal Toutouh , Enrique Alba This is my paper · ORCID

classification cs.NEcs.LG

keywords neuroevolutionrecurrentneuralnetworksarchitectureoptimizationmeanabsoluteerrorrandomsamplingtraining-freeevaluationevolutionaryalgorithmLSTMtimeseriesprediction

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 claims that a recurrent neural network architecture can be chosen without training any candidate: sample random weight sets, record their mean absolute errors, fit a truncated normal distribution, and use the probability that the error falls below a threshold as the fitness in an evolutionary search. This method, called RESN, is reported to match training-based architecture optimization on sine-wave and waste-generation prediction, to beat a published neuroevolutionary baseline on coal-flame prediction, and to cut optimization time roughly in half. The reason to care is that architecture search is normally dominated by training cost; if a cheap untrained proxy can rank architectures correctly, much larger search spaces become feasible within the same compute budget.

What carries the argument

The central object is the Mean Absolute Error Random Sampling (MRS) score $p_t$. For a candidate architecture, the algorithm generates several sets of random normally distributed weights, evaluates the network's MAE on the data for each set, fits a truncated normal distribution to those MAE values, and computes the probability that a sample lies below a threshold (0.01 in this paper). That probability serves as the fitness guiding an evolutionary search with binary tournament selection, cell-count and layer-count mutations, and self-adapting mutation parameters; because no gradient training happens during search, candidate evaluation is cheap. Once the search ends, the single best architecture is trained with Adam.

What would settle it

Rank a set of candidate architectures by the MRS probability $p_t$ and also by their test error after full training; if the rank correlation is close to zero on a fresh dataset, the search signal cannot be doing the claimed work. The paper reports no such direct check.

Watch

Extended reading notes

Core claim

RESN's central claim is that the MRS score $p_t$ — the estimated probability that a randomly weighted instance of an architecture has mean absolute error below a threshold — can stand in for actual training error during architecture search. The evolutionary algorithm maximizes $p_t$ over LSTM stack configurations, and only the single final winner is trained, using Adam. The reported consequence is that this training-free search matches or improves on training-based and neuroevolutionary baselines: mean MAE 0.105 versus 0.142 on sine prediction, median MAE 0.073 versus 0.073 on waste prediction, and average MSE 0.005208 versus 0.024432 on coal-flame prediction, while reducing optimization time by about half and producing results comparable to a human-expert-designed load-forecasting solution.

Load-bearing premise

The whole method rests on assuming that the probability of a random-weight version of a network scoring low error tells you how good that network will be after it is actually trained — a link the paper relies on but never directly measures.

Editorial extensions

If this is right

Architecture search can be run without training candidates, so the same time budget can explore far more architectures than training-based search.
The search phase and the training phase are cleanly separated: the search uses a training-free proxy, and only the final architecture incurs training cost.
On the waste-generation problem, RESN matches the short-training baseline in median MAE while roughly halving the total optimization time.
On coal-flame prediction, the paper reports an error about one order of magnitude lower than the EXALT neuroevolutionary baseline it compares against.
The evolutionary framework is not tied to the MRS fitness: the paper notes that the evaluation function can be swapped for any other fitness, so the search structure is reusable.

Reading between the lines

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

If the MRS proxy holds across problem types, the same trick could be applied to other training-heavy searches, such as convolutional network or transformer architecture search, by replacing MAE with a task-appropriate error metric.
The threshold and number of random samples are taken from the original MRS paper without tuning; varying them could change ranking quality, and finding a principled way to set them might make the proxy more reliable.
Since only the final solution is trained, the time savings of RESN should grow with dataset size — training-based search becomes prohibitive on large datasets while the MRS forward-pass cost stays comparatively small.
A direct testable extension is to combine the training-free proxy with learning-curve extrapolation: use MRS for early pruning of bad architectures and reserve expensive training for the survivors.

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.

Desk Editor's Note

read the letter

Short version: RESN is a genuine new combination—MRS as a training-free fitness inside an evolutionary search, with only the winner trained. The paper is clearly written, the code is public, and the experiments use 30 runs and Wilcoxon tests. The soft spots are real but not fatal: one baseline selects on test error, the MRS proxy is inherited rather than measured, and “state-of-the-art” appears once too often. The random-search baseline actually gives the central claim indirect support.

The main thing you should know: this is not just another neuroevolution paper. The authors promote their earlier MRS idea from an architecture-comparison tool to the fitness function in a (mu+lambda) EA. That is a legitimate step. Four problems, repeated runs, statistical tests, and a reimplementable algorithm description. Credit where due.

Where it gets softer. E1.i defines GDET by training each candidate one epoch and evaluating it on the test set. That is not a valid training-based baseline; it has seen test labels. It doesn’t sink the paper—if anything, it gives GDET an unfair advantage and RESN still wins—but it muddies the claim that RESN beats a purely training-based optimizer on equal terms. E2 compares to published EXALT and EXAMM numbers rather than re-running them; acceptable but weaker than a head-to-head. The authors note EXAMM’s MSE is 0.0016 while RESN is 0.0052, then still say “state-of-the-art” in the conclusions. The body is more honest; the conclusion overreaches.

The bigger question is whether MRS actually predicts trained performance. The paper does not report a correlation between pt and post-training MAE/MSE, and the MRS parameters come from the earlier paper without tuning. Strictly speaking, that is the load-bearing assumption. But the paper includes E1.iii, a random search baseline that trains candidates. On the waste problem, RESN (mean MAE 0.079) beats that random search (0.091), p=0.017. If the MRS signal were pure noise, RESN would be picking architectures at random and should not beat a selection-from-trained-candidates baseline. So the worry that gains stem entirely from final Adam training does not hold up for that dataset. Still, an ablation replacing MRS with random fitness, plus a simple correlation plot, would settle the proxy question directly.

Who is this for? Anyone working on cheap architecture search or neuroevolution for RNNs. It deserves serious peer review and a request for revision, not a desk reject. Send it out; the authors need to fix the baseline, tone down the SOTA claim, and measure the proxy, but the core idea is sound and worth engaging.

Referee Report

4 major / 5 minor

Summary. The paper proposes RESN, an evolutionary algorithm for RNN architecture optimization that uses the Mean Absolute Error Random Sampling (MRS) as a training-free fitness function. MRS estimates, for a given architecture, the probability that a randomly weighted instantiation achieves MAE below a threshold, and this probability is used to guide an (µ+λ) evolutionary search. Only the final selected architecture is trained with Adam (or SGD for the coal-flame experiment). The authors evaluate RESN on four prediction tasks: sine wave, waste generation, coal-flame intensity, and load forecasting, comparing against a gradient-based architecture optimizer (GDET), a random search, two published neuroevolutionary methods (EXALT and EXAMM), and expert-designed solutions. They report that RESN matches or improves on these baselines while halving optimization time, and conclude that MRS is a very useful heuristic.

Significance. RESN addresses a real bottleneck in neural architecture search: the cost of training every candidate. If the MRS proxy is valid, the approach offers a significant speedup and could make evolutionary architecture optimization practical for RNNs. The paper includes a public code repository and uses statistical tests on repeated runs. However, the empirical evidence for the central mechanism is incomplete: the MRS proxy's correlation with post-training error is never measured, the main same-protocol baseline leaks test-set information, and the 'state-of-the-art' claim is contradicted by the EXAMM numbers quoted in the same section. These gaps currently prevent the paper from fully supporting its conclusions.

major comments (4)

[Section 4.2.1, E1.i] The GDET baseline selects candidate architectures by computing MAE on the test set during optimization, which leaks test information and makes the comparison invalid for RQ1. A legitimate training-based baseline should use a validation split. The Wilcoxon p-value of 0.001 in Section 5 therefore cannot be interpreted as evidence that RESN matches or exceeds training-based optimization.
[Section 5, E2 (Table 8)] The text claims 'state-of-the-art performance' and 'ten times more accurate than the state-of-the-art' after comparing RESN to EXALT, but the same paragraph reports EXAMM averages of 0.001690 and 0.001601, which are both better than RESN's 0.005208. Since EXAMM is a neuroevolutionary method on the same dataset, the unqualified state-of-the-art claim is not supported by the paper's own numbers.
[Section 3, Algorithms 1-2 and Section 6] The central premise that the MRS probability pt is a reliable predictor of post-training performance is never tested. The paper reports no correlation between pt and final MAE/MSE and no ablation that removes or randomizes the MRS signal in the evolutionary loop. Consequently, the conclusion in Section 6 that 'the MRS is a very useful heuristic' is unsupported; the reported error performance might be attributable to the evolutionary operators and the final training of the selected architecture.
[Section 5, E4 (Table 9)] The comparison to expert-designed solutions is based on matching RESN's minimum over 30 runs (1.370) against the minima of the expert methods, whose means are not reported (NA). This is not a valid algorithm-level comparison; moreover, RESN's minimum is not better than the WKNNRW minimum (1.323). The claim that RESN is 'comparable to a human expert' needs a fairer comparison (e.g., means or distributions).

minor comments (5)

[Abstract] The abstract and full text should be consistent on the number of prediction problems; the first abstract says three, the full text says four.
[Section 4.2.1] The GDET baseline's use of the test set for selection should be explicitly flagged as a methodological limitation, as the current text does not acknowledge the leakage.
[Table 9] The table would be clearer if the source studies' full statistics were reported; the NA entries make it hard to judge the comparison.
[Section 5] The code repository link should include a version/commit identifier and a description of the computational environment to support reproducibility.
[Section 2.2] The MRS threshold (0.01) and weight distribution are fixed across datasets of different scales; the paper should discuss whether this is appropriate or whether the threshold should be normalized per problem.

Circularity Check

0 steps flagged · score 0.0 of 10

No constructional circularity: RESN's benchmarked claims are supported by external empirical comparisons, and its MRS fitness is not defined in terms of the predicted post-training error.

full rationale

The paper's derivation chain is empirical rather than self-referential. The MRS score pt (Algorithm 1) is computed by sampling random-weight MAE values and fitting a truncated normal; it is used as the RESN fitness in Equation 1 and Algorithm 2, while the reported errors are measured only after the final Adam/SGD training of the selected architecture. Nothing in Equation 1 or Algorithm 1 defines pt as the post-training error, so the final MAE/MSE results are not forced by construction. The E1.i GDET baseline, the E1.ii Short training comparison, E2 EXALT/EXAMM comparison, and E4 expert-design comparison all rely on independently run experiments or published numbers, and the paper does not fit any parameter to those targets. The paper does inherit MRS and its parameter values from same-author references [10,12] and explicitly defers MRS parameter tuning to future work, and it does not report a correlation between pt and final training error; that is a verification and independence concern, not an equation-level circularity. Accordingly, no circular step meeting the quoted-reduction standard is present.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper's main conceptual debt is a validity assumption for MRS inherited from the authors' prior work; the remaining entries are standard modeling and experimental assumptions. No invented entities are introduced.

free parameters (4)

MRS THRESHOLD = 0.01 (fixed across all problems)
Hand-chosen in prior self-cited MRS work; pt is P(MAE < 0.01) under random weights, so the search direction depends on this value. The paper acknowledges tuning is out of scope.
MRS MAX_SAMPLES = 100
Number of random weight samples per architecture; set from [10], not tuned for the four datasets.
MRS random weight distribution = mu=0, sigma=1
Random weight distribution for MRS inherited from [10]; no sensitivity analysis is provided.
Evolutionary hyperparameters = cell_mut_p=0.2, layer_mut_p=0.2, max_step=5 or 15, population_size=10, offspring_size=10, max_eval=100
Set from prior work [39] and self-adaptation is intended to reduce sensitivity; still, these are hand-chosen and not fitted to the target datasets.

assumptions (4)

domain assumption The MRS score pt is a valid proxy for the post-training error of an RNN architecture.
The entire search uses pt as fitness (Algorithm 2, Evaluate calls Algorithm 1); the paper cites [10] but does not directly measure correlation between pt and trained MAE. This is the load-bearing premise.
standard math A truncated normal distribution adequately describes the distribution of random-weight MAE values.
Algorithm 1 fits a truncated normal to samples before computing pt; this distributional assumption is inherited from [10] and not tested per problem.
domain assumption Search-space bounds (min/max layers, cells, look-back) contain good architectures for each problem.
Different bounds per experiment (Tables 2-5) are set by the authors; no sensitivity analysis is given.
domain assumption Baseline results taken from published tables are directly comparable.
EXALT [38], EXAMM [37], Short training [45], and Lang et al. [46] results are quoted; differences in data splits, training epochs, and hardware are not controlled.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Random Error Sampling-based Recurrent Neural Network Architecture Optimization." pith.science (2026). https://pith.science/paper/A3ACLLUZ

@misc{pith2026190902425,
  author       = {Pith},
  title        = {Pith review of: Random Error Sampling-based Recurrent Neural Network Architecture Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A3ACLLUZ}},
  note         = {Machine review of arXiv:1909.02425}
}

read the original abstract

Recurrent neural networks are good at solving prediction problems. However, finding a network that suits a problem is quite hard because their performance is strongly affected by their architecture configuration. Automatic architecture optimization methods help to find the most suitable design, but they are not extensively adopted because of their high computational cost. In this work, we introduce the Random Error Sampling-based Neuroevolution (RESN), an evolutionary algorithm that uses the mean absolute error random sampling, a training-free approach to predict the expected performance of an artificial neural network, to optimize the architecture of a network. We empirically validate our proposal on three prediction problems, and compare our technique to training-based architecture optimization techniques and to neuroevolutionary approaches. Our findings show that we can achieve state-of-the-art error performance and that we reduce by half the time needed to perform the optimization.

Figures

Figures reproduced from arXiv: 1909.02425 by the authors.

**Figure 1.** The global scheme of RESN y(t) = A · sin(2π · f · t + φ) (2) The waste generation prediction problem, introduced by Ferrer and Alba [44], consists of predicting the filling level of 217 recycling bins located in the metropolitan area of a city in Spain. Each filling level is recorded daily for one year. Therefore, given the historical data (i.e., 217 input values per day), the problem is to predict the next day (i.… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 45 canonical work pages

[1]

LeCun, Y

Y . LeCun, Y . Bengio, G. Hinton, Deep learning, nature 521 (2015) 436

work page 2015
[2]

Litjens, T

G. Litjens, T. Kooi, B. E. Bejnordi, A. A. A. Setio, F. Ciompi, M. Ghafoo- rian, J. A. van der Laak, B. van Ginneken, C. I. Sánchez, A survey on deep learning in medical image analysis, Medical image analysis 42 (2017) 60–88

work page 2017
[3]

S. Min, B. Lee, S. Yoon, Deep learning in bioinformatics, Brieﬁngs in bioinformatics 18 (2017) 851–869. 8

work page 2017
[4]

Bengio, P

Y . Bengio, P. Simard, P. Frasconi, Learning long-term dependencies with gradient descent is diﬃcult, IEEE tran on neural networks 5 (1994) 157–166

work page 1994
[5]

Pascanu, T

R. Pascanu, T. Mikolov, Y . Bengio, On the diﬃculty of training recurrent neural networks, in: Proceedings of the 30th International Conference on International Conference on Machine Learning - V olume 28, ICML 13, JMLR.org, 2013, pp. III–1310–III–1318

work page 2013
[6]

J. S. Bergstra, R. Bardenet, Y . Bengio, B. Kégl, Algorithms for hyper- parameter optimization, in: J. Shawe-Taylor, R. S. Zemel, P. L. Bartlett, F. Pereira, K. Q. Weinberger (Eds.), Advances in Neural Information Processing Systems 24, Curran Associates, Inc., 2011, pp. 2546–2554

work page 2011
[7]

Camero, J

A. Camero, J. Toutouh, D. H. Stolﬁ, E. Alba, Evolutionary deep learning for car park occupancy prediction in smart cities, in: Intl. Conf. on Learning and Intelligent Optimization, Springer, 2018, pp. 386–401

work page 2018
[8]

Jozefowicz, W

R. Jozefowicz, W. Zaremba, I. Sutskever, An empirical exploration of recurrent network architectures, in: Proceedings of the 32Nd International Conference on International Conference on Machine Learning - V olume 37, ICML’15, JMLR.org, 2015, pp. 2342–2350

work page 2015

Show all 50 references

[9]

V . K. Ojha, A. Abraham, V . Snášel, Metaheuristic design of feedforward neural networks: A review of two decades of research, Engineering Applications of Artiﬁcial Intelligence 60 (2017) 97–116

2017
[10]

Camero, J

A. Camero, J. Toutouh, E. Alba, Low-cost recurrent neural network expected performance evaluation, Preprint arXiv:1805.07159 (2018)

2018 arXiv
[11]

Domhan, J

T. Domhan, J. T. Springenberg, F. Hutter, Speeding up automatic hyperpa- rameter optimization of deep neural networks by extrapolation of learning curves, in: Proceedings of the 24th International Conference on Artiﬁcial Intelligence, IJCAI’15, AAAI Press, 2015, pp. 3460–3468

2015
[12]

Camero, J

A. Camero, J. Toutouh, E. Alba, Comparing deep recurrent networks based on the mae random sampling, a ﬁrst approach, in: Conference of the Spanish Association for Artiﬁcial Intelligence, Springer, 2018, pp. 24–33

2018
[13]

Z. C. Lipton, J. Berkowitz, C. Elkan, A critical review of recurrent neural networks for sequence learning, arXiv preprint arXiv:1506.00019 (2015)

2015 arXiv
[14]

J. F. Kolen, S. C. Kremer, Gradient Flow in Recurrent Nets: The Diﬃculty of Learning LongTerm Dependencies, Wiley-IEEE Press, 2001, pp. 464– 479

2001
[15]

Hochreiter, J

S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural computa- tion 9 (1997) 1735–1780

1997
[16]

E. Z. Ramos, M. Nakakuni, E. Yfantis, Quantitative measures to evaluate neural network weight initialization strategies, in: 2017 IEEE 7th An- nual Computing and Communication Workshop and Conference (CCWC), 2017, pp. 1–7

2017
[17]

Haykin, Neural networks and learning machines, volume 3, Pearson Upper Saddle River, NJ, USA:, 2009

S. Haykin, Neural networks and learning machines, volume 3, Pearson Upper Saddle River, NJ, USA:, 2009

2009
[18]

Albelwi, A

S. Albelwi, A. Mahmood, A framework for designing the architectures of deep convolutional neural networks, Entropy 19 (2017) 242

2017
[19]

S. C. Smithson, G. Yang, W. J. Gross, B. H. Meyer, Neural networks designing neural networks: multi-objective hyper-parameter optimization, in: Computer-Aided Design (ICCAD), 2016 IEEE /ACM International Conference on, IEEE, 2016, pp. 1–8

2016
[20]

Back, Evolutionary Algorithms in Theory and Practice: Evolution Strate- gies, Evolutionary Programming, Genetic Algorithms, Oxford university press, 1996

T. Back, Evolutionary Algorithms in Theory and Practice: Evolution Strate- gies, Evolutionary Programming, Genetic Algorithms, Oxford university press, 1996

1996
[21]

E. Alba, J. Aldana, J. M. Troya, Full automatic ann design: A genetic approach, in: International Workshop on Artiﬁcial Neural Networks, Springer, 1993, pp. 399–404

1993
[22]

E. Alba, R. Martí, Metaheuristic procedures for training neural networks, volume 35, Springer Science & Business Media, 2006

2006
[23]

Zhining, P

Y . Zhining, P. Yunming, The genetic convolutional neural network model based on random sample, International Journal of u-and e-Service, Science and Technology 8 (2015) 317–326

2015
[24]

G. Rosa, J. Papa, A. Marana, W. Scheirer, D. Cox, Fine-tuning convo- lutional neural networks using harmony search, in: A. Pardo, J. Kittler (Eds.), Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, Springer Intl Pub, Cham, 2015, pp. 683–690

2015
[25]

L. R. Rere, M. I. Fanany, A. M. Arymurthy, Simulated annealing algorithm for deep learning, Procedia Computer Science 72 (2015) 137 – 144. The Third Information Systems International Conference 2015

2015
[26]

L. Rere, M. I. Fanany, A. M. Arymurthy, Metaheuristic algorithms for convolution neural network, Computational intelligence and neuroscience 2016 (2016)

2016
[27]

C. Wang, C. Xu, X. Yao, D. Tao, Evolutionary generative adversarial networks, IEEE Transactions on Evolutionary Computation 23 (2019) 921–934

2019
[28]

Schmiedlechner, I

T. Schmiedlechner, I. N. Z. Yong, A. Al-Dujaili, E. Hemberg, U.-M. O’Reilly, Lipizzaner: A System That Scales Robust Generative Adver- sarial Network Training, in: the 32nd Conference on Neural Information Processing Systems (NeurIPS 2018) Workshop on Systems for ML and Open S...

2018
[29]

Toutouh, E

J. Toutouh, E. Hemberg, U.-M. O’Reilly, Spatial evolutionary generative adversarial networks, in: Proceedings of the Genetic and Evolutionary Computation Conference, GECCO ’19, Association for Computing Ma- chinery, New York, NY , USA, 2019, p. 472–480. URL:https://doi. org/10...

2019
[30]

K. O. Stanley, R. Miikkulainen, Evolving neural networks through aug- menting topologies, Evolutionary computation 10 (2002) 99–127

2002
[31]

Larochelle, Y

H. Larochelle, Y . Bengio, J. Louradour, P. Lamblin, Exploring strategies for training deep neural networks, Journal of machine learning research 10 (2009) 1–40

2009
[32]

Miikkulainen, J

R. Miikkulainen, J. Liang, E. Meyerson, A. Rawal, D. Fink, O. Francon, B. Raju, H. Shahrzad, A. Navruzyan, N. Duﬀy, et al., Evolving deep neural networks, in: Artiﬁcial Intelligence in the Age of Neural Networks and Brain Computing, Elsevier, 2019, pp. 293–312

2019
[33]

Rawal, R

A. Rawal, R. Miikkulainen, Evolving deep lstm-based memory networks using an information maximization objective, in: Proceedings of the Genetic and Evolutionary Computation Conference 2016, ACM, 2016, pp. 501–508

2016
[34]

Liang, E

J. Liang, E. Meyerson, R. Miikkulainen, Evolutionary architecture search for deep multitask networks, in: Proceedings of the Genetic and Evolution- ary Computation Conference, GECCO ’18, ACM, New York, NY , USA, 2018, pp. 466–473. URL: http://doi.acm.org/10.1145/3205455. 32054...

2018
[35]

A. M. Ibrahim, N. H. El-Amary, Particle swarm optimization trained recurrent neural network for voltage instability prediction, Journal of Electrical Systems and Information Technology 5 (2018) 216 – 228

2018
[36]

ElSaid, F

A. ElSaid, F. E. Jamiy, J. Higgins, B. Wild, T. Desell, Using ant colony optimization to optimize long short-term memory recurrent neu- ral networks, in: Proceedings of the Genetic and Evolutionary Com- putation Conference, GECCO ’18, ACM, New York, NY , USA, 2018, pp. 13–20. ...

2018
[37]

Ororbia, A

A. Ororbia, A. ElSaid, T. Desell, Investigating recurrent neural network memory structures using neuro-evolution, in: Proceedings of the Genetic and Evolutionary Computation Conference, 2019, pp. 446–455

2019
[38]

ElSaid, S

A. ElSaid, S. Benson, S. Patwardhan, D. Stadem, T. Desell, Evolving recurrent neural networks for time series data prediction of coal plant pa- rameters, in: International Conference on the Applications of Evolutionary Computation (Part of EvoStar), Springer, 2019, pp. 488–503

2019
[39]

C. Doerr, Non-static parameter choices in evolutionary computation, in: Genetic and Evolutionary Computation Conference, GECCO 2017, Berlin, Germany, July 15-19, 2017, Companion Material Proceedings, ACM,

2017
[40]

D. P. Kingma, J. Ba, Adam: A method for stochastic optimization, arXiv preprint arXiv:1412.6980 (2014)

2014 arXiv
[41]

Ferrer, E

J. Ferrer, E. Alba, Bin-ct: Urban waste collection based in predicting the container ﬁll level, arXiv preprint arXiv:1807.01603 (2018)

2018 arXiv
[42]

Chen, M.-W

B.-J. Chen, M.-W. Chang, et al., Load forecasting using support vector machines: A study on eunite competition 2001, IEEE tran on power systems 19 (2004) 1821–1830

2004
[43]

R. N. Bracewell, R. N. Bracewell, The Fourier transform and its applica- tions, volume 31999, McGraw-Hill New York, 1986

1986
[44]

Ferrer, E

J. Ferrer, E. Alba, BIN-CT: sistema inteligente para la gestión de la recogida de residuos urbanos, in: International Greencities Congress, 2018, pp. 117–128

2018
[45]

Camero, J

A. Camero, J. Toutouh, J. Ferrer, E. Alba, Waste generation prediction in smart cities through deep neuroevolution, in: Ibero-American Congress on Information Management and Big Data, Springer, 2018, pp. 192–204

2018
[46]

K. Lang, M. Zhang, Y . Yuan, X. Yue, Short-term load forecasting based on multivariate time series prediction and weighted neural network with random weights and kernels, Cluster Computing (2018) 1–9

2018
[47]

Camero, J

A. Camero, J. Toutouh, E. Alba, Dlopt: deep learning optimization library, arXiv preprint arXiv:1807.03523 (2018)

2018 arXiv
[48]

Chollet, et al., Keras, https://keras.io, 2015

F. Chollet, et al., Keras, https://keras.io, 2015. 9

2015
[49]

Abadi, P

M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, M. Isard, et al., Tensorﬂow: A system for large-scale machine learning., in: OSDI, volume 16, 2016, pp. 265–283. 10

2016
[2017]

doi:10.1145/3067695.3067707

Pith tools

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

[1] [1]

LeCun, Y

Y . LeCun, Y . Bengio, G. Hinton, Deep learning, nature 521 (2015) 436

work page 2015

[2] [2]

Litjens, T

G. Litjens, T. Kooi, B. E. Bejnordi, A. A. A. Setio, F. Ciompi, M. Ghafoo- rian, J. A. van der Laak, B. van Ginneken, C. I. Sánchez, A survey on deep learning in medical image analysis, Medical image analysis 42 (2017) 60–88

work page 2017

[3] [3]

S. Min, B. Lee, S. Yoon, Deep learning in bioinformatics, Brieﬁngs in bioinformatics 18 (2017) 851–869. 8

work page 2017

[4] [4]

Bengio, P

Y . Bengio, P. Simard, P. Frasconi, Learning long-term dependencies with gradient descent is diﬃcult, IEEE tran on neural networks 5 (1994) 157–166

work page 1994

[5] [5]

Pascanu, T

R. Pascanu, T. Mikolov, Y . Bengio, On the diﬃculty of training recurrent neural networks, in: Proceedings of the 30th International Conference on International Conference on Machine Learning - V olume 28, ICML 13, JMLR.org, 2013, pp. III–1310–III–1318

work page 2013

[6] [6]

J. S. Bergstra, R. Bardenet, Y . Bengio, B. Kégl, Algorithms for hyper- parameter optimization, in: J. Shawe-Taylor, R. S. Zemel, P. L. Bartlett, F. Pereira, K. Q. Weinberger (Eds.), Advances in Neural Information Processing Systems 24, Curran Associates, Inc., 2011, pp. 2546–2554

work page 2011

[7] [7]

Camero, J

A. Camero, J. Toutouh, D. H. Stolﬁ, E. Alba, Evolutionary deep learning for car park occupancy prediction in smart cities, in: Intl. Conf. on Learning and Intelligent Optimization, Springer, 2018, pp. 386–401

work page 2018

[8] [8]

Jozefowicz, W

R. Jozefowicz, W. Zaremba, I. Sutskever, An empirical exploration of recurrent network architectures, in: Proceedings of the 32Nd International Conference on International Conference on Machine Learning - V olume 37, ICML’15, JMLR.org, 2015, pp. 2342–2350

work page 2015