Machine Learning and Auctions
Scientists have used machine learning for a myriad of different purposes, making interesting contributions in fields as diverse as medicine and art. In basic terms, machine learning refers to teaching a computer to perform a task based on providing it with training examples (Hardesty). The linked article describes how scientists from the Technical University of Munich (TUM) are now using it as a novel way to study complex game theory and achieve results previously not attainable with traditional methods (Malewar). Specifically, the researchers are studying auction theory, a facet of game theory that is primarily used to study markets, and applying neural networks to solve complex problems in the field.
Until now, only simple auctions could be solved precisely to find a Bayes Nash Equilibrium, or a place where parties can not “improve their expected utility by changing their strategy.” Simple auctions involve multiple, simultaneous bids on a single item. However, many real-word auction situations are more complicated than this, and the article highlights one particular type of complex auction, wireless spectrum auctions by governments worldwide, which involves simultaneous bids on multiple items. By applying neural networks, the researchers at TUM developed an algorithm to find an equilibrium point for precisely this type of problem. Neural networks, also known as deep learning, are a high performing type of artificial intelligence.
This article ties into the class because the algorithm is specifically made to try and find the Nash Equilibrium in a complex auction game. Thanks to this algorithm, bidders can get as close to an equilibrium as possible, something previously impossible. In class, we learned how to find a Nash Equilibrium in a game where all players know all the moves and payoffs in a simultaneous game for one item. Specifically, the algorithm works to find the equilibrium in multi-item markets with each item affecting other items.
Sources:https://www.techexplorist.com/understanding-complex-markets-using-machine-learning/40980/
https://www.nature.com/articles/s42256-021-00365-4
https://news.mit.edu/2017/explained-neural-networks-deep-learning-0414
-Camila Orr