Abstract
There are a wide variety of empirical settings that do not easily fit within an optimization framework and for which results that are “good,” but not necessarily the “best,” are the only practical option. When business firms compete, for instance, a single firm can dominate a market by “just” being better than its competitors. Over the past decade, reinforcement learning has built on this perspective with considerable success. Although several features of reinforcement learning are some distance from our full regression approach, its promise motivates a brief discussion. Reinforcement learning also is sometimes included as a component of deep learning.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
In our gridlock example, the set of possible decisions is fixed and does not depend on what choices you make. Among the most exciting applications of tree search algorithms involve playing against some opponent who reacts to your decisions and counters by changing the mix of choices you have available. The setting is adversarial. A game of checkers is a simple example. With each alternative move, the available decisions and their consequences can change. The most impressive performance to date is by Google’s AlphaGo AI that beat the best Go player in the world. A discussion of adversarial reinforcement learning is beyond the scope of this book in part because the regression formulation is a stretch. The setting is a game. See Silver et al. (2016). It is very cool stuff.
- 2.
There is some disagreement about whether genetic algorithms should be seen as reinforcement learning, and there are indeed some important differences (Sutton and Barto 2018: 8–9). This section can be productively read even if genetic algorithms are at least a distant cousin to reinforcement learning.
- 3.
The manner in which the initial population is generated often does not matter a great deal. For example, an initial population of 500 could be composed of 500 identical network specifications. Variation in the population is introduced later.
- 4.
The new form was designed by Susan B. Sorenson in collaboration with the local police department (Berk and Sorenson 2019).
- 5.
References
Affenseller, M., Winkler, S., Wagner, S., & Beham, A. (2009). Genetic algorithms and genetic programming: Modern concepts and practical applications. New York: Chapman & Hall.
Berk, R. A., & Bleich, J. (2013) Statistical procedures for forecasting criminal behavior: A comparative assessment. Journal of Criminology and Public Policy, 12(3), 515–544.
Berk, R. A., & Sorenson, S. (2019). An algorithmic approach to forecasting rare violent events: An illustration based on IPV perpetration. arXiv: 1903.00604v1.
Choudhary, A. (2019). Introduction to Monte Carlo tree search: The game-changing algorithm behind DeepMind’s AlphaGo. Analytics Vidhya, Jan 23. https://www.analyticsvidhya.com/blog/2019/01/monte-carlo-tree-search-introduction-algorithm-deepmind-alphago/
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. Cambridge: MIT Press.
Lapan, M. (2018). Deep reinforcement learning hands on. Birmingham: Packt Publishing.
Mitchell, M. (1998). An introduction to genetic algorithms. Cambridge: MIT Press.
Proellochs, N. (2019). Reinforcement learning in R. https://cran.r-project.org/web/packages/ReinforcementLearning/vignettes/ReinforcementLearning.html
Scrucca, L. (2014). GA: A package for genetic algorithms in R. Journal of Statistical Software, 53(4), 1–37.
Scrucca, L. (2017). On some extensions to GA package: Hybrid optimisation, parallelisation and islands evolution. The R Journal, 9(1), 187–206.
Silver, D., Huang, A., Maddison, C.J., Guez, A., Sifre, L., van den Driessche, G., et al. (2016). Mastering the game of go with deep neural networks and tree search. Nature, 529(28), 484–489.
Sutton, R. S., & Barto, A. G. (2018). Reinforcement learning (2nd ed.). Cambridge: A Bradford Book.
Umbarkar, A. J., & Sheth, P. D. (2015). Crossover operators in genetic algorithms: A review. ICTACT Journal of Soft Computing, 6(1), 1083–1092.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this chapter
Cite this chapter
Berk, R.A. (2020). Reinforcement Learning and Genetic Algorithms. In: Statistical Learning from a Regression Perspective. Springer Texts in Statistics. Springer, Cham. https://doi.org/10.1007/978-3-030-40189-4_9
Download citation
DOI: https://doi.org/10.1007/978-3-030-40189-4_9
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-40188-7
Online ISBN: 978-3-030-40189-4
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)