![](https://img-blog.csdnimg.cn/20201014180756928.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
Introduction to AI
@yuqing_wang
这个作者很懒,什么都没留下…
展开
-
Proability and Bayes’ NET
Probability and Bayes’ NETProbabilistic ReasoningGeneral situation of Uncertaintyobseved variables(evidence)Agent knows certain things about the state of the worldunobserved variablesAgent needs to reason about other aspectsmodelAgent knows someth原创 2020-12-18 14:11:22 · 652 阅读 · 0 评论 -
Markov Models and Hidden Markov Models
Markov Models模型构成initial distrbutiontransition modelstationary assumption: 转移概率保持不变Markov propertyMini-Forward Algorithmmarginalize:o(dj)o(d^j)o(dj)存储Stationary distributionHidden Markov Modelsobserve some evidence at timestep influence the bel原创 2020-12-17 10:53:45 · 134 阅读 · 0 评论 -
reinforcement learning
Reinforcement learningoffline planning: agent have full knowledge of both transition function and rewards functiononline planning: agent have no prior knowledge transition function and rewards functionmust do exploritation and receives feedbacksample原创 2020-12-17 09:42:19 · 269 阅读 · 0 评论 -
Markov Decision processes
Non-deterministic searchMarkov Decision processpropertya set of states Sa set of actions Astart statepossibly one or more terminal statesa discount factorγ\gammaγtransition function T(s,a,s′)T(s,a,s')T(s,a,s′): a probablity functionreward functio原创 2020-12-08 15:56:45 · 307 阅读 · 0 评论 -
Adversarial search
Gamesadversarial search problems/games: our agents have one or more adversaries who attempt to keep them from reaching their goalstype of gamesactions:deterministic or stochastic outcomesnumber of playerszero-sum games: all the utility will be a co原创 2020-12-07 12:07:29 · 614 阅读 · 0 评论 -
Constraint Satisfaction Problems
Constraint Satisfation problemsplanning v.s. identificationplanning problems: search problemfollowing path will result a goalpath has various costheuristics give problems a guidestate is a black-box,and goal test is performed on stateidentification原创 2020-12-07 10:45:07 · 803 阅读 · 0 评论 -
search problem
Search problemsconsista state space: all possilbel states in the worldworld state: all info about the state in the worldsearch state: only info needed for planninge.g. Pacman pathing / pacman eating dotssuccessor function:input: action,stateoutput原创 2020-12-05 15:54:14 · 628 阅读 · 0 评论 -
Introduction
Rational decisionsrational: maximize achieving pre-defined goals(maximize expected utility)rationality only concerned decisions not the processgoals: the utility of the outcomeAgentagent: perceives and actsrational agent: an entity that has goals原创 2020-12-03 18:40:25 · 80 阅读 · 0 评论