DNA-Pattern-Inspired Machine Learning Weight Optimisation
DOI:
https://doi.org/10.63282/3117-5481/AIJCST-V4I3P105Keywords:
DNA Computing, Weight Optimization, Neural Architecture Search, Genetic Algorithms, Enzymatic Learning, Triplet Codon Encoding, Molecular Neural Networks, Evolutionary Computation, Deep Learning, Biologically Inspired Optimization, Strand Displacement, Particle Swarm OptimizationAbstract
Combining molecular biology with machine learning has given rise to a new paradigm in which structural and functional characteristics of deoxyribonucleic acid (DNA) are used as design principles for optimising the weights of neural networks. This paper combines improvements from the last decade reported until 2020 in the fields of DNA-inspired weight encoding, enzymatic update rules, strand-displacement computation and evolutionary neural architecture search to provide a holistic framework, namely DNA-Pattern Inspired Machine Learning Weight Optimisation (DPIWMO). The results show that enzymatic weight update schemes [2] outperform conventional genetic algorithm (GA) based schemes [13] by 25.7% in terms of the number of parameters and 98.1% in terms of accuracy on benchmark logic problems. Winner-take-all (WTA) molecular networks with a 100-class classification [3] illustrate that the DNA substrate provides a search space of more than 10¹´ candidate weight configurations. It could be demonstrated that DNA computing-inspired deep architecture search [17] reached a top-1 accuracy of 67.4% on ImageNet, 3.2% better than the NEAT-based methods [10] which had only 16.4 million parameters. The convergence of network weights was improved by 18% and the training time was shortened by 2.6 hours for the Triplet codon encoding GAs than for binary encoded GAs [16]. The theoretical limit is explored, the No-Free-Lunch (NFL) theorem [14], which inspires use of hybrid approaches that integrate DNA-sequence motifs with gradient-based refinement. Cross domain utility is also demonstrated in applications for genomic sequence mining [15] which reaches 97.1% accuracy on 200,000 sequences. The proposed framework can be readily extended to the design of automated neural networks that are biologically plausible.
References
[1] L. M. Adleman, "Molecular computation of solutions to combinatorial problems," Science, vol. 266, no. 5187, pp. 1021–1024, 1994.
[2] C. Baek, S.-W. Lee, B.-J. Lee, D.-H. Kwak, and B.-T. Zhang, "Enzymatic weight update algorithm for DNA-based molecular learning," Molecules, vol. 24, no. 7, Art. 1409, 2019.
[3] K. M. Cherry and L. Qian, "Scaling up molecular pattern recognition with DNA-based winner-take-all neural networks," Nature, vol. 559, no. 7714, pp. 370–376, 2018.
[4] J. H. Holland, Adaptation in Natural and Artificial Systems. Cambridge, MA: MIT Press, 1992.
[5] Y.-F. Huang, C.-M. Wang, and S.-W. Liou, "Discovering weighted patterns in intron sequences using self-adaptive harmony search and back-propagation algorithms," The Scientific World Journal, vol. 2013, Art. 249034, 2013.
[6] J. Kennedy and R. C. Eberhart, "Particle swarm optimization," in Proc. IEEE Int. Conf. Neural Networks, Perth, WA, Australia, 1995, vol. 4, pp. 1942–1948.
[7] F. H. F. Leung, H. K. Lam, S. H. Ling, and P. K. S. Tam, "Tuning of the structure and parameters of a neural network using an improved genetic algorithm," IEEE Trans. Neural Netw., vol. 14, no. 1, pp. 79–88, Jan. 2003.
[8] Z. Lu et al., "Multiobjective evolutionary design of deep convolutional neural networks for image classification," IEEE Trans. Evol. Comput., vol. 25, no. 2, pp. 277–291, Apr. 2021.
[9] L. Qian, E. Winfree, and J. Bruck, "Neural network computation with DNA strand displacement cascades," Nature, vol. 475, no. 7356, pp. 368–372, 2011.
[10] K. O. Stanley and R. Miikkulainen, "Evolving neural networks through augmenting topologies," Evol. Comput., vol. 10, no. 2, pp. 99–127, 2002.
[11] Y. Sun, B. Xue, M. Zhang, and G. G. Yen, "Evolving deep convolutional neural networks for image classification," IEEE Trans. Evol. Comput., vol. 24, no. 2, pp. 394–407, Apr. 2020.
[12] Y. Sun, B. Xue, M. Zhang, G. G. Yen, and J. Lv, "Automatically designing CNN architectures using the genetic algorithm for image classification," IEEE Trans. Cybern., vol. 50, no. 9, pp. 3840–3854, Sep. 2020.
[13] D. Whitley, "A genetic algorithm tutorial," Stat. Comput., vol. 4, no. 2, pp. 65–85, 1994.
[14] D. H. Wolpert and W. G. Macready, "No free lunch theorems for optimization," IEEE Trans. Evol. Comput., vol. 1, no. 1, pp. 67–82, Apr. 1997.
[15] A. Yang et al., "Review on the application of machine learning algorithms in the sequence data mining of DNA," Front. Bioeng. Biotechnol., vol. 8, Art. 1032, 2020.
[16] X. Yang, S. Deng, M. Ji, J. Zhao, and W. Zheng, "Neural network evolving algorithm based on the triplet codon encoding method," Genes, vol. 9, no. 12, Art. 626, 2018.
[17] G. Zhong et al., "DNA computing inspired deep networks design," Neurocomputing, vol. 378, pp. 305–314, 2020.
[18] G. Zhong, T. Li, W. Liu, and Y. Chen, "Structure learning of deep networks via DNA computing algorithm," presented at AAAI Workshop Eng. Dependable Secure Mach. Learn. Syst., 2018.
[19] Taluri, R. (2021). Cloud-Native Architectures for Enterprise Financial Data Management, Analytics, and Regulatory Reporting Compliance. International Journal of Emerging Trends in Computer Science and Information Technology, 2(2), 101-111. https://doi.org/10.63282/3050-9246.IJETCSIT-V2I2P112
