Title Bayesovske igre
Author Vanessa Keranović
Mentor Lavoslav Čaklović (mentor)
Committee member Lavoslav Čaklović (predsjednik povjerenstva)
Committee member Zlatko Drmač (član povjerenstva)
Committee member Marcela Hanzer (član povjerenstva)
Committee member Hrvoje Šikić (član povjerenstva)
Granter University of Zagreb Faculty of Science (Department of Mathematics) Zagreb
Defense date and country 2017-09-25, Croatia
Scientific / art field, discipline and subdiscipline NATURAL SCIENCES Mathematics
Abstract U ovome radu obrađene su igre s nepotpunom informacijom, koje još zovemo i Bayesovske igre. Bayesovske igre podijelili smo na 2 podvrste, statičke i dinamičke, gdje smo svaku podvrstu obradili u zasebnom poglavlju te potom posvetili još jedno poglavlje primjeni Bayesovskih igara u stvarnom životu, aukcijama. Statičke Bayesovske igre sastoje se od simultanih poteza igrača nakon čega igra završava. Ovo poglavlje započinjemo motivacijskim primjerom, čime uočavamo potrebne definicije kako bi se početna igra transformirala u onu koju možemo riješiti. U istom poglavlju uvodimo osnovne definicije i pravila koja se kasnije poopćavaju u dinamičkim Bayesovskim igrama. Te osnovne definicije i pravila odnose se između ostalog na uvođenje tipa igrača, čime se igra s nepotpunom informacijom (o funkciji ispate igrača) svodi na igru s nesavršenom informacijom (o tipu igrača). Definiramo Bayesovu ravnotežu te uočavamo da je ona ekvivalentna ravnoteži koju bismo dobili nakon "signala", odnosno nakon što igrači upoznaju svoje tipove. Također, uočavamo da je Bayesova ravnoteža zapravo Nashova ravnoteža ekvivalentne transformirane igre. Poglavlje završavamo modificiranim poznatim primjerom
Rat spolova, gdje je broj tipova neprekidan. U dinamičkim igrama potezi nisu nužno simultani. Dinamičke Bayesovske igre prvo predstavljamo pomoću posebnog slučaja signalizirajućih igara, gdje također započinjemo motivacijskim primjerom te uvodimo grafički prikaz igre koji još zovemo i ekstenzivna forma. Potom prelazimo na općenite dinamičke Bayesovske igre, gdje uz poopćenje postojećih definicija iz poglavlja o statičkim Bayesovskim igrama, uvodimo modifikaciju podigara originalnih igara iz dinamičkih igara s potpunom informacijom. Nadalje, pomoću te modifikacije definiramo PBE, posebnu vrstu Bayesove ravnoteže za dinamičke Bayesovske igre te poglavlje završavamo zanimljivim primjerom gdje koristimo novo definirane pojmove. U posljednjem poglavlju bavimo se aukcijama, prilično uspješnom primjenom Bayesovskih igara. Dotičemo se različitih vrsta aukcija (koje se još uvijek danas koriste u svijetu) te se bavimo njihovim ravnotežama. Na kraju, dajemo primjer aukcije gdje, osim ravnoteža igrača, promatramo i položaj prodavača, odnosno koja od dviju aukcija bi za njega bila profitabilnija. Ovime završavamo rad.
Abstract (english) In this paper we are studying games with incomplete information, also known as Bayesian games. We have divided Bayesian games into two types, static and dynamic, where each of the types are studied in separate chapters. There is one more chapter dedicated to an application of Bayesian games in real life, auctions. Static Bayesian games consist of simultaneous moves made by players, after which the game ends. This chapter we open with a motivational example that shows us necessary definitions to transform a game we start with into one we can solve. In the same chapter, we give definitions and rules about Bayesian games that we later on, for dynamic Bayesian games, generalize. These definitions and rules refer to introducing types of players, which transforms a game with incomplete information (about the utility function) into a game with imperfect information (about the types of players). By defining Bayesian equilibrium we notice the equivalence between equilibria calculated before and after "the signal", the moment when the players find out their type. We also show that Bayesian equilibrium is actually Nash equilibrium of the equivalent, transformed game. We finish the chapter about static Bayesian games with a modified example of Battle of the Sexes, where the number of types of players is continuous. In dynamic games player's moves aren't necessarily simultaneous. Chapter Dynamic Bayesian games we begin by introducing a special case of dynamic Bayesian games, "signaling" games. Here as well we start with a motivational example and a graphical representation of the game, also called an extensive form. We then study dynamic Bayesian games in general, where in addition to generalized definitions introduced in the chapter about static Bayesian games, we define a modification of a subgame in a game with complete information. Further more, using that modification we define PBE (perfect Bayesian equilibrium), a special kind of Bayesian equilibrium and we finish the chapter with an interesting example of a perjury trap where we apply newly defined terms. The last chapter deals with auctions, a very successful application of Bayesian games. We consider different types of auctions (which are still used nowadays) and their equilibria. We finish the chapter with an example of an auction where we in addition to finding equilibria regarding players, we also consider the position the seller is in and which of the two auctions considered in the example are more profitable for him.
Keywords
igre s nepotpunom informacijom
Bayesovske igre
statičke igre
dinamičke igre
Bayesova ravnoteža
Nashova ravnoteža ekvivalentne transformirane igre
aukcije
Keywords (english)
games with incomplete information
Bayesian games
static Bayesian games
dynamic Bayesian games
Bayesian equilibrium
Nash equilibrium of the equivalent transformed game
auctions
perfect Bayesian equilibrium
PBE
Language croatian
URN:NBN urn:nbn:hr:217:358031
Study programme Title: Finance and Business Mathematics Study programme type: university Study level: graduate Academic / professional title: magistar/magistra matematike (magistar/magistra matematike)
Type of resource Text
File origin Born digital
Access conditions Open access
Terms of use
Created on 2018-01-31 12:55:00