Using Game Theory to Understand Machine Learning
Machine learning is a prominent field of computer science. It helps in creating human-like intelligence in several day-to-day products to automate actions. On the other hand, the Game theory is a field of mathematics to understand the behavior of different players in a particular setting with some constraints. It can be used to improve and better understand the concepts of Machine Learning.
In this blog, I will discuss two concepts of Machine learning using Game theory.
The first concept is the Training of GANs Model. In the GANs, there are two models generative and discriminative. In terms of game theory, there are two players generative and discriminative. The generative player tries to generate more fake images that look real. The discriminative player tries to improve the algorithm to distinguish between fake and real images. Both the players will improve their strategies (models) until they reach their until they reach Nash Equibrilllium.
The second concept is the model using MARL (multi-agent Reinforcement Learning). Reinforcement learning is a concept in which a model trains itself using the inputs from the surrounding environment. But when multiple agents are involved in the environment, we need to consider their interaction also. In terms of game theory, all the agents are the player. All the players will improve their strategies (models) until they Nash Equibrilllium or have perfect collaboration.
Through this blog, I discussed two concepts of machine learning that can be better explained using Game theory. Concepts like Nash Equibrillium in Game theory (that are discussed in class) helps in better explanation and then better modeling of mathematical concepts in the area of ML. Through game theory, we can help in the better development of ML algorithms as they help us in understanding them through practical examples.
Articles used (citation):
https://towardsdatascience.com/game-theory-in-artificial-intelligence-57a7937e1b88
http://cs229.stanford.edu/proj2012/AgrawalJaiswal-WhenMachineLearningMeetsAIandGameTheory.pdf
https://www.cs.cmu.edu/~mblum/search/AGTML35.pdf