A Method for Improving the Performance of Ensemble Neural Networks by Introducing Randomization into Their Training Data
Abstract
:1. Introduction
2. Methodology Introduction
2.1. Creating a Neural Network Solution to Predict Sleep
2.1.1. Ensemble MSE Calculation
- For each neural network NNg = NN1–NN100 Do
- Train NNg through 20 epochs.
- EndFor
- For each species Sn = S1–Sm Do
- For each neural network NNg = NN1–NN100 Do
- Run NNg with input data from Sn.
- Calculate the squared error in SWS and in PS (using output from NNg).
- EndFor
- Calculate (for the species Sn) the mean (across all NN1–NN100) of the squared error in SWS and the mean (across all NN1–NN100) of the squared error in PS.
- EndFor
- Calculate the mean (across all species S1–Sm) of the MSE in SWS predictions.
- Calculate the mean (across all species S1–Sm) of the MSE in PS predictions.
2.1.2. Node Sensitivity Study
- For number of hidden layer nodes = 2–20 Do
- Call the ensemble MSE calculation described in Section 2.1.1.
- EndFor
- Compare the MSE in SWS predictions and PS predictions across all node numbers.
2.2. Assessing the Impact of Each Independent Variable on the SWS and PS Predictions of an Ensemble of 100 Neural Networks Trained on the Same Data Set
Impacts Study Based on 5% Perturbations of Inputs Acting on An Ensemble of 100 Neural Networks Trained All on the Same Data Set
- Call the ensemble MSE calculation described in Section 2.1.1.
- For each input variable i = 1–7 Do
- For each species Sn = S1–Sm Do
- Perturb the variable xi,n by 5%
- Calculate the MSE in the SWS predictions and the MSE in the PS predictions when feeding the perturbed variable as an input to the ensemble of NN1–100 already-trained neural networks.
- EndFor
- EndFor
- Calculate the proportional MSE impact of each perturbation in the input variable averaged across all species S1–Sm.
- Normalize the MSE impacts separately for each output variable (SWS and PS), with 100 being the greatest impact.
2.3. Assessing the Impact of Each Independent Variable on the SWS and PS Predictions of an Ensemble of 100 Neural Networks Trained on a Diversified Training Set
2.3.1. Simulating Sample Error in the Training Data with Full Range Sampling
- Set standard deviations for p = 9 observations of n = 39 species. This study uses for i = 1–p and n = 1–n. This could be improved with attention to each attribute and species, but 5% of the mean is considered reasonable. This is at the lower end of standard deviations among brain sizes and lifespans in mammals (commensurate with cows) but at the upper end of standard deviations of gestation time (commensurate with gorillas and humans). In total, 5% of the mean represents approximately 15–45 min of short-wave sleep or up to 30 min of paradoxical sleep for most of the animals in this study.
- For j = 1–10 Do
- Perturb every variable xi,m (independent and dependent variables ranging from I = 1–p and species m = 1–n) in the training set by a randomly generated amount whereby the mean value produced by the random number generator is 1 and the standard deviation is . Save the perturbed value in an expanded input set of 390 training data points (n = 39 species by jmax = 10 perturbations).
- EndFor
- For each neural network NNg = NN1–NN100 Do
- For each species Sn = S1–Sm Do
- Set j = a random integer ,
- Select data set j for species Sn in the training set of neural network NNg.
- EndFor
- Train NNg through 20 epochs.
- EndFor
2.3.2. Impacts Study Based on 5% Perturbations of Inputs Acting on an Ensemble of 100 Neural Networks Each Trained on a Data Set Randomized with Full Range Sampling (FRS)
- Call the function of Section 2.3.1 to simulate sample error in the training data with FRS.
- For each independent variable i = 1–7 Do
- For each species Sn = S1–Sm Do
- Perturb the variable xi,n by 5%
- Calculate the MSE in the SWS predictions and the MSE in the PS predictions when feeding the perturbed variable as an input to the ensemble of NN1–100 already-trained neural networks.
- EndFor
- EndFor
- Calculate the proportional MSE impact of each perturbation in the input variable averaged across all species.
- Normalize the MSE impacts separately for each output variable (SWS and PS), with 100 being the greatest impact.
3. Results
4. Discussion
4.1. Node Number Sensitivity Study
4.2. Neural Network Ensembles
4.3. Randomized Training Sets with Full Range Sampling
4.4. Interpretations of Randomised Data and Variations in Data
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Tadeusiewicz, R. Neural networks as a tool for modeling of biological systems. Bio-Algorithms Med. Syst. 2015, 11, 135–144. [Google Scholar] [CrossRef]
- Dalzeil, B.D.; Morales, J.M.; Fryxell, J.M. Fitting probability distributions to animal movement trajectories: Using artificial neural networks to link distance, resources, and memory. Am. Nat. 2008, 172, 248–258. [Google Scholar] [CrossRef] [PubMed]
- Tracey, J.A.; Zhu, J.; Crooks, K.R. Modeling and inference of animal movement using artificial neural networks. Env. Ecol. Stats 2011, 18, 393–410. [Google Scholar] [CrossRef]
- Jeantet, L.; Vigon, V.; Geiger, S.; Chevallier, D. Fully convolutional neural network: A solution to infer animal behaviours from multi-sensor data. Ecol. Model. 2021, 450, 109555. [Google Scholar] [CrossRef]
- Fang, C.; Zhang, T.M.; Zeng, H.K.; Huang, J.D.; Cuan, K.X. Pose estimation and behaviour classification of broiler chickens based on deep neural networks. Comp. Electron. Agric. 2021, 180, 105863. [Google Scholar] [CrossRef]
- Allison, T.; Cicchetti, D.V. Sleep in Mammals: Ecological and Constitutional Correlates. Science 1976, 194, 732–734. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Pan, Z.X.; Pan, D.J.; Sun, P.Y.; Zhang, M.S.; Zuberbuhler, A.D.; Jung, B. Spectroscopic quantitation of amino acids by using artificial neural networks combined with factor analysis. Spectrochim. Acta Part A 1997, 53, 1629–1632. [Google Scholar]
- Zhang, Y.X. Artificial neural networks based on principal component analysis input selection for clinical pattern recognition analysis. Talanta 2007, 73, 68–75. [Google Scholar] [CrossRef] [PubMed]
- Ding, S.F.; Jia, W.K.; Su, C.Y.; Zhang, L.W.; Liu, L.L. Research of neural network algorithm based on factor analysis and cluster analysis. Neural Comput. Appl. 2011, 20, 297–302. [Google Scholar] [CrossRef]
- Crick, Francis, and Graeme Mitchison. REM sleep and neural nets. Behav. Brain Res. 1995, 69, 147–155. [Google Scholar] [CrossRef] [PubMed]
- Tagluk, M.E.; Akin, M.; Sezgin, N. Classıfıcation of sleep apnea by using wavelet transform and artificial neural networks. Expert. Syst. Appl. 2010, 37, 1600–1607. [Google Scholar] [CrossRef]
- Opitz, D.W.; Shavlik, J.W. Actively Searching for an Effective Neural-Network Ensemble. Connect. Sci. 1996, 8, 3–4. [Google Scholar] [CrossRef]
- Sagi, O.; Rokach, L. Ensemble learning: A survey. WIREs Data Min. Knowl. Discov. 2018, 8, 255–272. [Google Scholar] [CrossRef]
- StatLib, Carnegie Mellon University. Available online: https://lib.stat.cmu.edu/datasets/ (accessed on 29 October 2022).
- MATLAB and Statistics Toolbox; ver. R2022a; The Mathworks, Inc.: Natick, MA, USA, 2022.
- Sobhanifard, Y. Hybrid modelling of the consumption of organic foods in Iran using exploratory factor analysis and an artificial neural network. BJF 2017, 120, 44–58. [Google Scholar] [CrossRef]
- Efron, B. Bootstrap methods: Another look at the jackknife. Ann. Stat. 1979, 7, 1–26. [Google Scholar] [CrossRef]
- Microsoft Excel; ver. 16.62; Microsoft Corp.: Redmond, WA, USA, 2022.
- Trichakis, I.; Nikolos, I.; Karatzas, G.P. Comparison of bootstrap confidence intervals for an ANN model of a karstic aquifer response. Hydrol. Process 2011, 25, 2827–2836. [Google Scholar] [CrossRef]
- Richards, B.; Emekwuru, N. Using machine learning to predict synthetic fuel spray penetration from limited experimental data without computational fluid dynamics. In Proceedings of the ICESF International Conference on Energy and Sustainable Futures, Coventry, UK, 7–8 September 2022. [Google Scholar]
- NIST/SEMATECH e-Handbook of Statistical Methods; NIST: Gaithersburg, MD, USA, 2012. [CrossRef]
- Zepelin, H.; Siegel, J.M.; Tobler, I. Mammalian sleep. Chapter 8. In Principles and Practice of Sleep Medicine, 4th ed.; Roth, K.M.H.T., Dement, W.C., Eds.; Elsevier/Saunders: Philadelphia, PA, USA, 2005; pp. 91–100. [Google Scholar]
Neural Network | Mean Squared Error * | |
---|---|---|
Node Number | SWS | PS |
2 | 48.59% | 50.74% |
3 | 46.87% | 42.06% |
4 | 39.13% | 36.34% |
5 | 39.19% | 29.87% |
6 | 38.92% | 29.75% |
7 | 33.98% | 32.54% |
8 | 33.13% | 26.41% |
9 | 35.11% | 25.37% |
10 | 31.33% | 27.04% |
11 | 31.16% | 23.27% |
12 | 32.73% | 22.35% |
13 | 32.56% | 21.38% |
14 | 23.60% | 17.56% |
15 | 31.50% | 19.45% |
16 | 28.51% | 19.15% |
17 | 28.55% | 16.78% |
18 | 30.02% | 18.53% |
19 | 29.23% | 15.32% |
20 | 25.20% | 17.78% |
Present Study 1 Normalized MSE Impacts | Allison and Cicchetti 2 Correlation Coefficients | |||
---|---|---|---|---|
Input Variable | SWS | PS | SWS | PS |
Body weight | 10 | 2 | −0.712 | −0.370 |
Brain weight | 15 | 4 | −0.679 | −0.435 |
Lifespan | 33 | 14 | −0.377 | −0.342 |
Gestation time | 33 | 24 | −0.589 | −0.651 |
Predation index | 88 | 57 | −0.369 | −0.536 |
Sleep exposure index | 58 | 31 | −0.580 | −0.591 |
Overall danger index | 100 | 100 | −0.542 | −0.686 |
Normalized MSE Impacts on Ensembles with FRS Diversified Training Data 1 | Change in Absolute MSE Impact When Using FRS Diversified Training Data 2 | |||
---|---|---|---|---|
Input Variable | SWS | PS | SWS | PS |
Body weight | 6 | 10 | −0.61 × 10−2 | 0.39 × 10−2 |
Brain weight | 16 | 24 | −0.46 × 10−2 | 1.07 × 10−2 |
Lifespan | 42 | 58 | −0.61 × 10−2 | 1.97 × 10−2 |
Gestation time | 35 | 41 | −1.07 × 10−2 | −0.59 × 10−2 |
Predation index | 100 | 95 | −2.47 × 10−2 | −1.63 × 10−2 |
Sleep exposure index | 50 | 98 | −2.62 × 10−2 | 2.32 × 10−2 |
Overall danger index | 83 | 100 | −4.73 × 10−2 | −7.44 × 10−2 |
Totals | −12.5 × 10−2 | −3.91 × 10−2 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Richards, B.; Emekwuru, N. A Method for Improving the Performance of Ensemble Neural Networks by Introducing Randomization into Their Training Data. Knowledge 2023, 3, 307-319. https://doi.org/10.3390/knowledge3030021
Richards B, Emekwuru N. A Method for Improving the Performance of Ensemble Neural Networks by Introducing Randomization into Their Training Data. Knowledge. 2023; 3(3):307-319. https://doi.org/10.3390/knowledge3030021
Chicago/Turabian StyleRichards, Bryn, and Nwabueze Emekwuru. 2023. "A Method for Improving the Performance of Ensemble Neural Networks by Introducing Randomization into Their Training Data" Knowledge 3, no. 3: 307-319. https://doi.org/10.3390/knowledge3030021