370
个编辑
更改
跳到导航
跳到搜索
第995行：
第995行： − − − 第1,004行：
第1,001行： − AI已经开发出许多工具来解决计算机科学中最困难的问题。下面将讨论其中一些最常用的方法。+ − − − ===搜索和优化 Search and optimization === − − − − − − + − − − − − − 第1,029行：
第1,012行： − + − 第1,037行：
第1,019行： − 对于大多数真实世界的问题，简单的穷举搜索很难满足要求: 搜索空间(要搜索的位置数)很快就会增加到天文数字。结果就是搜索速度太慢或者永远不能完成。对于许多问题，解决方法是使用“'''<font color=#ff8000>启发式 Heuristics</font>''' ”或“'''<font color=#ff8000>经验法则 Rules of Thumb</font>''' ” ，优先考虑那些更有可能达到目标的选择，并且在较短的步骤内完成。在一些搜索方法中，启发式方法还可以完全移去一些不可能通向目标的选择(称为“修剪搜索树”)。启发式为程序提供了解决方案所在路径的“最佳猜测”。启发式做不到去更小的样本中搜索解。+ 第1,044行：
第1,026行： − + 第1,058行：
第1,040行： − 进化计算用了优化搜索的形式。例如，他们可能从一群有机体(猜测)开始，然后让它们变异和重组，选择适者继续生存 (改进猜测)。经典的进化算法包括遗传算法、基因表达编程和遗传编程。+ − − − − − − − − − − − − − − − 第1,084行：
第1,051行： − 逻辑被用来知识表示和解决问题，还可以应用到其他问题上。例如，satplan 算法使用逻辑进行规划，归纳逻辑编程是一种学习方法。+ − − − 第1,093行：
第1,057行： − AI研究中使用了多种不同形式的逻辑。命题逻辑包含诸如“或”和“否”这样的真值函数。一阶逻辑增加了量词和谓词，可以表达关于对象、对象属性和对象之间的关系。模糊集合论给诸如“爱丽丝老了”(或是富有的、高的、饥饿的)这样模糊的表述赋予了一个“真实程度”(介于0到1之间)，这些表述在语言上很模糊，不能完全判定为正确或错误。模糊逻辑在控制系统中得到了成功应用，使专家能够制定模糊规则，比如“如果你正以较快的速度接近终点站，那么就增加列车的制动压力”;这些模糊的规则可以在系统内用数值细化。模糊逻辑无助于扩展知识库;许多AI研究者质疑链接模糊逻辑推理的有效性。+ 第1,099行：
第1,063行： − − '''<font color=#ff8000>缺省逻辑 Default Logics</font>'''、'''<font color=#ff8000>非单调逻辑 Non-monotonic Logics</font>'''、'''<font color=#ff8000>限制逻辑 Circumscription</font>'''和'''<font color=#ff8000>模态逻辑 Modal Logics</font>'''。对多智能体系统中出现的矛盾或不一致的陈述进行建模的逻辑也已经被设计出来，例如'''<font color=#ff8000>次协调逻辑 Paraconsistent Logics.</font>'''。 − 默认逻辑、非单调逻辑和边界都用逻辑形式来解决缺省推理和限定问题。一些逻辑扩展被用于处理特定的知识领域，例如:描述逻辑;情景演算、事件演算和用于表示事件和时间的流畅演算；因果演算;信念演算(信念修正);和模态逻辑。人们也设计了对多智能体系统中出现的矛盾或不一致陈述进行建模的逻辑，如次协调逻辑。 − − − − − ===不确定推理的概率方法 Probabilistic methods for uncertain reasoning === − − + + − 第1,123行：
第1,078行： − − − − − 第1,134行：
第1,084行： − + − − − 第1,145行：
第1,092行： − + − 第1,153行：
第1,099行： − + − − − ===分类器与统计学习方法 Classifiers and statistical learning methods === − − − − − + − − − − − − 第1,177行：
第1,110行： − + + − + − − A classifier can be trained in various ways; there are many statistical and machine learning approaches. The decision tree is perhaps the most widely used machine learning algorithm. Other widely used classifiers are the neural network, − − 分类器可以通过多种方式进行训练;，比如许多统计学和机器学习方法。决策树可能是应用最广泛的机器学习算法。其他使用广泛的分类器还有神经网络。 − − − [[k-nearest neighbor algorithm]],{{efn|The most widely used analogical AI until the mid-1990s{{sfn|Domingos|2015|p=187}}}}<ref name="K-nearest neighbor algorithm"/> − − k-nearest neighbor algorithm,}} − − K最近邻算法 − − [[kernel methods]] such as the [[support vector machine]] (SVM),{{efn|SVM displaced k-nearest neighbor in the 1990s{{sfn|Domingos|2015|p=188}}}}<ref name="Kernel methods"/> − − kernel methods such as the support vector machine (SVM),}} − − 例如支持向量机的核心方法 − − − Gaussian mixture model, and the extremely popular naive Bayes classifier.}} Classifier performance depends greatly on the characteristics of the data to be classified, such as the dataset size, distribution of samples across classes, the dimensionality, and the level of noise. Model-based classifiers perform well if the assumed model is an extremely good fit for the actual data. Otherwise, if no matching model is available, and if accuracy (rather than speed or scalability) is the sole concern, conventional wisdom is that discriminative classifiers (especially SVM) tend to be more accurate than model-based classifiers such as "naive Bayes" on most practical data sets. − − '''<font color=#ff8000> 高斯混合模型 Gaussian Mixture Mode</font>'''，以及非常流行的'''<font color=#ff8000>朴素贝叶斯分类器 Naive Bayes Classifier</font>'''。分类器的分类效果在很大程度上取决于待分类数据的特征，如数据集的大小、样本跨类别的分布、维数和噪声水平。如果假设的模型很符合实际数据，则基于这种模型的分类器就能给出很好的结果。否则，传统观点认为如果没有匹配模型可用，而且只关心准确性(而不是速度或可扩展性) ，在大多数实际数据集上鉴别分类器(尤其是支持向量机)往往比基于模型的分类器(如“朴素贝叶斯”)更准确。 + − − − − − − − − − 第1,225行：
第1,129行： − − − − 第1,236行：
第1,136行： − 神经网络的诞生受到人脑神经元结构的启发。一个简单的“神经元”N 接受来自其他神经元的输入，每个神经元在被激活(或者说“放电”)时，都会对N是否应该被激活按一定的权重赋上值。学习的过程需要一个根据训练数据调整这些权重的算法; 一个被称为“相互放电，彼此联系”简单的算法在一个神经元激活触发另一个神经元的激活时增加两个连接神经元之间的权重。神经网络中形成一种分布在一个共享的神经元子网络中的”概念”，这些神经元往往一起放电; ”腿”的概念可能和”脚”概念的子网络相结合，后者包括”脚”的发音。神经元有一个连续的激活谱; 此外，神经元还可以用非线性的方式处理输入，而不是简单地加权求和。现代神经网络可以学习连续函数甚至的数字逻辑运算。神经网络早期的成功包括预测股票市场和自动驾驶汽车(1995年)。2010年代，神经网络使用深度学习取得巨大进步，也因此将AI推向了公众视野里，并促使企业对AI投资急速增加; 例如2017年与AI相关的并购交易规模是2015年的25倍多。+ 第1,428行：
第1,328行： − + − −
→工具 Tools
==工具 Tools ==
==工具 Tools ==
AI has developed many tools to solve the most difficult problems in computer science. A few of the most general of these methods are discussed below.
AI has developed many tools to solve the most difficult problems in computer science. A few of the most general of these methods are discussed below.
AI领域已经开发出许多工具来解决计算机科学中最困难的问题。下面将讨论其中一些最常用的方法。
===搜索和优化===
{{Main|Search algorithm|Mathematical optimization|Evolutionary computation}}
{{Main|Search algorithm|Mathematical optimization|Evolutionary computation}}
Many problems in AI can be solved in theory by intelligently searching through many possible solutions:<ref name="Search"/> [[#Deduction, reasoning, problem solving|Reasoning]] can be reduced to performing a search. For example, logical proof can be viewed as searching for a path that leads from [[premise]]s to [[Logical consequence|conclusions]], where each step is the application of an [[inference rule]].<ref name="Logic as search"/> [[Automated planning and scheduling|Planning]] algorithms search through trees of goals and subgoals, attempting to find a path to a target goal, a process called [[means-ends analysis]].<ref name="Planning as search"/> [[Robotics]] algorithms for moving limbs and grasping objects use [[local search (optimization)|local searches]] in [[Configuration space (physics)|configuration space]].<ref name="Configuration space"/> Many [[machine learning|learning]] algorithms use search algorithms based on [[optimization (mathematics)|optimization]].
Many problems in AI can be solved in theory by intelligently searching through many possible solutions:<ref name="Search"/> [[#Deduction, reasoning, problem solving|Reasoning]] can be reduced to performing a search. For example, logical proof can be viewed as searching for a path that leads from [[premise]]s to [[Logical consequence|conclusions]], where each step is the application of an [[inference rule]].<ref name="Logic as search"/> [[Automated planning and scheduling|Planning]] algorithms search through trees of goals and subgoals, attempting to find a path to a target goal, a process called [[means-ends analysis]].<ref name="Planning as search"/> [[Robotics]] algorithms for moving limbs and grasping objects use [[local search (optimization)|local searches]] in [[Configuration space (physics)|configuration space]].<ref name="Configuration space"/> Many [[machine learning|learning]] algorithms use search algorithms based on [[optimization (mathematics)|optimization]].
Many problems in AI can be solved in theory by intelligently searching through many possible solutions: Reasoning can be reduced to performing a search. For example, logical proof can be viewed as searching for a path that leads from premises to conclusions, where each step is the application of an inference rule. Planning algorithms search through trees of goals and subgoals, attempting to find a path to a target goal, a process called means-ends analysis. Robotics algorithms for moving limbs and grasping objects use local searches in configuration space. Many learning algorithms use search algorithms based on optimization.
Many problems in AI can be solved in theory by intelligently searching through many possible solutions: Reasoning can be reduced to performing a search. For example, logical proof can be viewed as searching for a path that leads from premises to conclusions, where each step is the application of an inference rule. Planning algorithms search through trees of goals and subgoals, attempting to find a path to a target goal, a process called means-ends analysis. Robotics algorithms for moving limbs and grasping objects use local searches in configuration space. Many learning algorithms use search algorithms based on optimization.
AI中的许多问题可以通过智能地搜索许多可能的解决方案而在理论上得到解决: 推理可以简化为执行一次搜索。例如，逻辑证明可以看作是寻找一条从前提到结论的路径，其中每一步都用到了推理规则。规划算法通过搜索目标和子目标的树，试图找到一条通往目标的路径，这个过程称为目的手段分析。机器人学中移动肢体和抓取物体的算法使用的是位形空间的局部搜索。许多学习算法使用基于优化的搜索算法。
AI中的许多问题可以通过智能地搜索许多可能的解决方案而在理论上得到解决<ref name="Search"/>: 推理可以简化为执行一次搜索。例如，逻辑证明可以看作是寻找一条从前提到结论的路径，其中每一步都用到了推理规则。规划算法通过搜索目标和子目标的树，试图找到一条通往目标的路径，这个过程称为“目的-手段”分析。机器人学中移动肢体和抓取物体的算法使用的是位形空间的局部搜索。许多学习算法也使用到了基于优化的搜索算法。
Simple exhaustive searches are rarely sufficient for most real-world problems: the search space (the number of places to search) quickly grows to astronomical numbers. The result is a search that is too slow or never completes. The solution, for many problems, is to use "heuristics" or "rules of thumb" that prioritize choices in favor of those that are more likely to reach a goal and to do so in a shorter number of steps. In some search methodologies heuristics can also serve to entirely eliminate some choices that are unlikely to lead to a goal (called "pruning the search tree"). Heuristics supply the program with a "best guess" for the path on which the solution lies. Heuristics limit the search for solutions into a smaller sample size.
Simple exhaustive searches are rarely sufficient for most real-world problems: the search space (the number of places to search) quickly grows to astronomical numbers. The result is a search that is too slow or never completes. The solution, for many problems, is to use "heuristics" or "rules of thumb" that prioritize choices in favor of those that are more likely to reach a goal and to do so in a shorter number of steps. In some search methodologies heuristics can also serve to entirely eliminate some choices that are unlikely to lead to a goal (called "pruning the search tree"). Heuristics supply the program with a "best guess" for the path on which the solution lies. Heuristics limit the search for solutions into a smaller sample size.
对于大多数真实世界的问题，简单的穷举搜索<ref name="Uninformed search"/>很难满足要求: 搜索空间(要搜索的位置数)很快就会增加到天文数字。结果就是搜索速度太慢或者永远不能完成。对于许多问题，解决方法是使用“'''<font color=#ff8000>启发式 Heuristics</font>''' ”或“'''<font color=#ff8000>经验法则 Rules of Thumb</font>''' ” ，优先考虑那些更有可能达到目标的选择，并且在较短的步骤内完成。在一些搜索方法中，启发式方法还可以完全移去一些不可能通向目标的选择(称为“修剪搜索树”)。启发式为程序提供了解决方案所在路径的“最佳猜测”。启发式把搜索限制在了更小的样本规模里。。
A very different kind of search came to prominence in the 1990s, based on the mathematical theory of optimization. For many problems, it is possible to begin the search with some form of a guess and then refine the guess incrementally until no more refinements can be made. These algorithms can be visualized as blind hill climbing: we begin the search at a random point on the landscape, and then, by jumps or steps, we keep moving our guess uphill, until we reach the top. Other optimization algorithms are simulated annealing, beam search and random optimization.
A very different kind of search came to prominence in the 1990s, based on the mathematical theory of optimization. For many problems, it is possible to begin the search with some form of a guess and then refine the guess incrementally until no more refinements can be made. These algorithms can be visualized as blind hill climbing: we begin the search at a random point on the landscape, and then, by jumps or steps, we keep moving our guess uphill, until we reach the top. Other optimization algorithms are simulated annealing, beam search and random optimization.
在20世纪90年代，一种非常不同的基于数学最优化理论的搜索引起了人们的注意。对于许多问题，可以从某种形式的猜测开始搜索，然后逐步细化猜测，直到无法进行更多的细化。这些算法可以喻为盲目地爬山: 我们从地形上的一个随机点开始搜索，然后，通过跳跃或登爬，我们将猜测继续向山上移动，直到我们到达山顶。其他的优化算法有 '''<font color=#ff8000>模拟退火算法</font>''' 、'''<font color=#ff8000>定向搜索</font>''' 和'''<font color=#ff8000>随机优化</font>''' 。
在20世纪90年代，一种非常不同的基于数学最优化理论的搜索引起了人们的注意。对于许多问题，可以从某种形式的猜测开始搜索，然后逐步细化猜测，直到无法进行更多的细化。这些算法可以喻为盲目地爬山: 我们从地形上的一个随机点开始搜索，然后，通过跳跃或登爬，我们把猜测点继续向山上移动，直到我们到达山顶。其他的优化算法有 '''<font color=#ff8000>模拟退火算法</font>''' 、'''<font color=#ff8000>定向搜索</font>''' 和'''<font color=#ff8000>随机优化</font>''' 。<ref name="Optimization search"/>
Evolutionary computation uses a form of optimization search. For example, they may begin with a population of organisms (the guesses) and then allow them to mutate and recombine, selecting only the fittest to survive each generation (refining the guesses). Classic evolutionary algorithms include genetic algorithms, gene expression programming, and genetic programming.
Evolutionary computation uses a form of optimization search. For example, they may begin with a population of organisms (the guesses) and then allow them to mutate and recombine, selecting only the fittest to survive each generation (refining the guesses). Classic evolutionary algorithms include genetic algorithms, gene expression programming, and genetic programming.
[[演化计算]]用到了优化搜索的形式。例如，他们可能从一群有机体(猜测)开始，然后让它们变异和重组，选择适者继续生存 (改进猜测)。经典的演化算法包括遗传算法、基因表达编程和遗传编程。<ref name="Genetic programming"/> Alternatively, distributed search processes can coordinate via [[swarm intelligence]] algorithms. Two popular swarm algorithms used in search are [[particle swarm optimization]] (inspired by bird [[flocking (behavior)|flocking]]) and [[ant colony optimization]] (inspired by [[ant trail]]s).<ref name="Society based learning"/><ref>{{cite book|author1=Daniel Merkle|author2=Martin Middendorf|editor1-last=Burke|editor1-first=Edmund K.|editor2-last=Kendall|editor2-first=Graham|title=Search Methodologies: Introductory Tutorials in Optimization and Decision Support Techniques|date=2013|publisher=Springer Science & Business Media|isbn=978-1-4614-6940-7|language=en|chapter=Swarm Intelligence}}</ref>
===逻辑 Logic ===
===逻辑 Logic ===
{{Main|Logic programming|Automated reasoning}}
{{Main|Logic programming|Automated reasoning}}
[[Logic]]<ref name="Logic"/> is used for knowledge representation and problem solving, but it can be applied to other problems as well. For example, the [[satplan]] algorithm uses logic for [[automated planning and scheduling|planning]]<ref name="Satplan"/> and [[inductive logic programming]] is a method for [[machine learning|learning]].<ref name="Symbolic learning techniques"/>
[[Logic]]<ref name="Logic"/> is used for knowledge representation and problem solving, but it can be applied to other problems as well. For example, the [[satplan]] algorithm uses logic for [[automated planning and scheduling|planning]]<ref name="Satplan"/> and [[inductive logic programming]] is a method for [[machine learning|learning]].<ref name="Symbolic learning techniques"/>
Logic is used for knowledge representation and problem solving, but it can be applied to other problems as well. For example, the satplan algorithm uses logic for planning and inductive logic programming is a method for learning.
Logic is used for knowledge representation and problem solving, but it can be applied to other problems as well. For example, the satplan algorithm uses logic for planning and inductive logic programming is a method for learning.
逻辑<ref name="Logic"/>被用来表示知识和解决问题，还可以应用到其他问题上。例如，satplan 算法就使用逻辑进行规划<ref name="Satplan"/>。另外，归纳逻辑编程是一种学习方法。
Several different forms of logic are used in AI research. [[Propositional logic]]<ref name="Propositional logic"/> involves [[truth function]]s such as "or" and "not". [[First-order logic]]<ref name="First-order logic"/> adds [[quantifier (logic)|quantifiers]] and [[predicate (mathematical logic)|predicates]], and can express facts about objects, their properties, and their relations with each other. [[Fuzzy set theory]] assigns a "degree of truth" (between 0 and 1) to vague statements such as "Alice is old" (or rich, or tall, or hungry) that are too linguistically imprecise to be completely true or false. [[Fuzzy logic]] is successfully used in [[control system]]s to allow experts to contribute vague rules such as "if you are close to the destination station and moving fast, increase the train's brake pressure"; these vague rules can then be numerically refined within the system. Fuzzy logic fails to scale well in knowledge bases; many AI researchers question the validity of chaining fuzzy-logic inferences.{{efn|"There exist many different types of uncertainty, vagueness, and ignorance... [We] independently confirm the inadequacy of systems for reasoning about uncertainty that propagates numerical factors according to only to which connectives appear in assertions."<ref>{{cite journal|last1=Elkan|first1=Charles|title=The paradoxical success of fuzzy logic|journal=IEEE Expert|date=1994|volume=9|issue=4|pages=3–49|doi=10.1109/64.336150|citeseerx=10.1.1.100.8402}}</ref>}}<ref name="Fuzzy logic"/><ref>{{cite news|title=What is 'fuzzy logic'? Are there computers that are inherently fuzzy and do not apply the usual binary logic?|url=https://www.scientificamerican.com/article/what-is-fuzzy-logic-are-t/|accessdate=5 May 2018|work=Scientific American|language=en}}</ref>
Several different forms of logic are used in AI research. [[Propositional logic]]<ref name="Propositional logic"/> involves [[truth function]]s such as "or" and "not". [[First-order logic]]<ref name="First-order logic"/> adds [[quantifier (logic)|quantifiers]] and [[predicate (mathematical logic)|predicates]], and can express facts about objects, their properties, and their relations with each other. [[Fuzzy set theory]] assigns a "degree of truth" (between 0 and 1) to vague statements such as "Alice is old" (or rich, or tall, or hungry) that are too linguistically imprecise to be completely true or false. [[Fuzzy logic]] is successfully used in [[control system]]s to allow experts to contribute vague rules such as "if you are close to the destination station and moving fast, increase the train's brake pressure"; these vague rules can then be numerically refined within the system. Fuzzy logic fails to scale well in knowledge bases; many AI researchers question the validity of chaining fuzzy-logic inferences.{{efn|"There exist many different types of uncertainty, vagueness, and ignorance... [We] independently confirm the inadequacy of systems for reasoning about uncertainty that propagates numerical factors according to only to which connectives appear in assertions."<ref>{{cite journal|last1=Elkan|first1=Charles|title=The paradoxical success of fuzzy logic|journal=IEEE Expert|date=1994|volume=9|issue=4|pages=3–49|doi=10.1109/64.336150|citeseerx=10.1.1.100.8402}}</ref>}}<ref name="Fuzzy logic"/><ref>{{cite news|title=What is 'fuzzy logic'? Are there computers that are inherently fuzzy and do not apply the usual binary logic?|url=https://www.scientificamerican.com/article/what-is-fuzzy-logic-are-t/|accessdate=5 May 2018|work=Scientific American|language=en}}</ref>
Several different forms of logic are used in AI research. Propositional logic}}
Several different forms of logic are used in AI research. Propositional logic}}
AI研究中使用了多种不同形式的逻辑。命题逻辑<ref name="Propositional logic"/>包含诸如“或”和“否”这样的真值函数。一阶逻辑<ref name="First-order logic"/>增加了量词和谓词，可以表达关于对象、对象属性和对象之间的关系。模糊集合论给诸如“爱丽丝老了”(或是富有的、高的、饥饿的)这样模糊的表述赋予了一个“真实程度”(介于0到1之间)，这些表述在语言上很模糊，不能完全判定为正确或错误。模糊逻辑在控制系统中得到了成功应用，使专家能够制定模糊规则，比如“如果你正以较快的速度接近终点站，那么就增加列车的制动压力”;这些模糊的规则可以在系统内用数值细化。但是，模糊逻辑无助于扩展知识库，许多AI研究者质疑把模糊逻辑和推理结合起来的有效性。{{efn|"There exist many different types of uncertainty, vagueness, and ignorance... [We] independently confirm the inadequacy of systems for reasoning about uncertainty that propagates numerical factors according to only to which connectives appear in assertions."<ref>{{cite journal|last1=Elkan|first1=Charles|title=The paradoxical success of fuzzy logic|journal=IEEE Expert|date=1994|volume=9|issue=4|pages=3–49|doi=10.1109/64.336150|citeseerx=10.1.1.100.8402}}</ref>}}<ref name="Fuzzy logic"/><ref>{{cite news|title=What is 'fuzzy logic'? Are there computers that are inherently fuzzy and do not apply the usual binary logic?|url=https://www.scientificamerican.com/article/what-is-fuzzy-logic-are-t/|accessdate=5 May 2018|work=Scientific American|language=en}}</ref>
Default logics, non-monotonic logics and circumscription and modal logics. Logics to model contradictory or inconsistent statements arising in multi-agent systems have also been designed, such as paraconsistent logics.
Default logics, non-monotonic logics and circumscription and modal logics. Logics to model contradictory or inconsistent statements arising in multi-agent systems have also been designed, such as paraconsistent logics.
'''<font color=#ff8000>缺省逻辑 Default Logics</font>'''、'''<font color=#ff8000>非单调逻辑 Non-monotonic Logics</font>'''、'''<font color=#ff8000>限制逻辑 Circumscription</font>'''和'''<font color=#ff8000>模态逻辑 Modal Logics</font>'''<ref name="Default reasoning and non-monotonic logic"/>，都用逻辑形式来解决缺省推理和限定问题。一些逻辑扩展被用于处理特定的知识领域，例如:'''<font color=#ff8000>描述逻辑 Description Logics</font>'''<ref name="Representing categories and relations"/> 、情景演算、事件演算、'''<font color=#ff8000>流态演算 Fluent Calculus</font>'''（用于表示事件和时间）<ref name="Representing time"/>、因果演算<ref name="Representing causation"/>、信念演算(信念修正)<ref>"The Belief Calculus and Uncertain Reasoning", Yen-Teh Hsia</ref>、和模态逻辑<ref name="Representing knowledge about knowledge"/>。人们也设计了对多主体系统中出现的矛盾或不一致陈述进行建模的逻辑，如次协调逻辑。
===不确定推理的概率方法===
{{Main|Bayesian network|Hidden Markov model|Kalman filter|Particle filter|Decision theory|Utility theory}}
{{Main|Bayesian network|Hidden Markov model|Kalman filter|Particle filter|Decision theory|Utility theory}}
[[Expectation-maximization clustering of Old Faithful eruption data starts from a random guess but then successfully converges on an accurate clustering of the two physically distinct modes of eruption.]]
[[Expectation-maximization clustering of Old Faithful eruption data starts from a random guess but then successfully converges on an accurate clustering of the two physically distinct modes of eruption.]]
Many problems in AI (in reasoning, planning, learning, perception, and robotics) require the agent to operate with incomplete or uncertain information. AI researchers have devised a number of powerful tools to solve these problems using methods from probability theory and economics.
Many problems in AI (in reasoning, planning, learning, perception, and robotics) require the agent to operate with incomplete or uncertain information. AI researchers have devised a number of powerful tools to solve these problems using methods from probability theory and economics.
AI中的许多问题(在推理、规划、学习、感知和机器人技术方面)要求智能体在信息不完整或不确定的情况下进行操作。AI研究人员从概率论和经济学的角度设计了许多强大的工具来解决这些问题。
AI中的许多问题(在推理、规划、学习、感知和机器人技术方面)要求主体在信息不完整或不确定的情况下进行操作。AI研究人员从概率论和经济学的角度设计了许多强大的工具来解决这些问题。
Bayesian networks are a very general tool that can be used for various problems: reasoning (using the Bayesian inference algorithm), learning (using the expectation-maximization algorithm),}} planning (using decision networks) and perception (using dynamic Bayesian networks). Probabilistic algorithms can also be used for filtering, prediction, smoothing and finding explanations for streams of data, helping perception systems to analyze processes that occur over time (e.g., hidden Markov models or Kalman filters). Compared with symbolic logic, formal Bayesian inference is computationally expensive. For inference to be tractable, most observations must be conditionally independent of one another. Complicated graphs with diamonds or other "loops" (undirected cycles) can require a sophisticated method such as Markov chain Monte Carlo, which spreads an ensemble of random walkers throughout the Bayesian network and attempts to converge to an assessment of the conditional probabilities. Bayesian networks are used on Xbox Live to rate and match players; wins and losses are "evidence" of how good a player is. AdSense uses a Bayesian network with over 300 million edges to learn which ads to serve.
Bayesian networks are a very general tool that can be used for various problems: reasoning (using the Bayesian inference algorithm), learning (using the expectation-maximization algorithm),}} planning (using decision networks) and perception (using dynamic Bayesian networks). Probabilistic algorithms can also be used for filtering, prediction, smoothing and finding explanations for streams of data, helping perception systems to analyze processes that occur over time (e.g., hidden Markov models or Kalman filters). Compared with symbolic logic, formal Bayesian inference is computationally expensive. For inference to be tractable, most observations must be conditionally independent of one another. Complicated graphs with diamonds or other "loops" (undirected cycles) can require a sophisticated method such as Markov chain Monte Carlo, which spreads an ensemble of random walkers throughout the Bayesian network and attempts to converge to an assessment of the conditional probabilities. Bayesian networks are used on Xbox Live to rate and match players; wins and losses are "evidence" of how good a player is. AdSense uses a Bayesian network with over 300 million edges to learn which ads to serve.
'''<font color=#ff8000>贝叶斯网络 Bayesian Networks </font>''' 是一个非常通用的工具，可用于各种问题: 推理(使用贝叶斯推断算法) ，学习(使用期望最大化算法) ，规划(使用决策网络)和感知(使用动态贝叶斯网络)。概率算法也可以用于滤波、预测、平滑和解释数据流，帮助传感系统分析随时间发生的过程(例如，隐马尔可夫模型或'''<font color=#ff8000>卡尔曼滤波器 Kalman Filters</font>''')。与符号逻辑相比，形式化的贝叶斯推断逻辑运算量很大。为了使推论易于处理，大多数观察值必须彼此有条件地独立。含有方块或其他“圈”(无向循环)的复杂图形可能需要比如马尔科夫蒙特卡洛图的复杂方法，这种方法将一组随机行走遍布整个贝叶斯网络，并试图收敛到对条件概率的评估。贝叶斯网络在 Xbox Live 上被用来评估和匹配玩家; 胜率是证明一个玩家有多有优秀的“证据”。AdSense使用一个有超过3亿条边的贝叶斯网络来学习广告推广的最佳时机。
'''<font color=#ff8000>[[贝叶斯网络]] Bayesian Networks </font>''' 是一个非常通用的工具，可用于各种问题: 推理(使用贝叶斯推断算法) ，学习(使用期望最大化算法) ，规划(使用决策网络)和感知(使用动态贝叶斯网络)。概率算法也可以用于滤波、预测、平滑和解释数据流，帮助传感系统分析随时间发生的过程(例如，隐马尔可夫模型或'''<font color=#ff8000>卡尔曼滤波器 Kalman Filters</font>''')。与符号逻辑相比，形式化的贝叶斯推断逻辑运算量很大。为了使推理易于处理，大多数观察值必须彼此条件独立。含有菱形或其他“圈”(无向循环)的复杂图形可能需要比如马尔科夫-蒙特卡罗图的复杂方法，这种方法将一组随机行走遍布整个贝叶斯网络，并试图收敛到对条件概率的评估。贝叶斯网络在 Xbox Live 上被用来评估和匹配玩家：胜率是证明一个玩家有多有优秀的“证据”。AdSense使用一个有超过3亿条边的贝叶斯网络来学习如何推送广告的。
A key concept from the science of economics is "utility": a measure of how valuable something is to an intelligent agent. Precise mathematical tools have been developed that analyze how an agent can make choices and plan, using decision theory, decision analysis, and information value theory. These tools include models such as Markov decision processes, dynamic decision networks, game theory and mechanism design.
A key concept from the science of economics is "utility": a measure of how valuable something is to an intelligent agent. Precise mathematical tools have been developed that analyze how an agent can make choices and plan, using decision theory, decision analysis, and information value theory. These tools include models such as Markov decision processes, dynamic decision networks, game theory and mechanism design.
经济学中的一个关键概念是“效用” :这是一种衡量某物对于一个智能智能体的价值的方法。人们运用决策理论、决策分析和信息价值理论开发出了精确的数学工具来分析智能体应该如何选择和计划。这些工具包括马尔可夫决策过程、动态决策网络、博弈论和机制设计等模型。
经济学中的一个关键概念是“效用” :这是一种衡量某物对于一个智能主体的价值的方法。人们运用决策理论、决策分析和信息价值理论开发出了精确的数学工具来分析智能主体应该如何选择和计划。这些工具包括马尔可夫决策过程、动态决策网络、博弈论和机制设计等模型。
===分类器与统计学习方法===
{{Main|Classifier (mathematics)|Statistical classification|Machine learning}}
{{Main|Classifier (mathematics)|Statistical classification|Machine learning}}
The simplest AI applications can be divided into two types: classifiers ("if shiny then diamond") and controllers ("if shiny then pick up"). Controllers do, however, also classify conditions before inferring actions, and therefore classification forms a central part of many AI systems. [[Classifier (mathematics)|Classifiers]] are functions that use [[pattern matching]] to determine a closest match. They can be tuned according to examples, making them very attractive for use in AI. These examples are known as observations or patterns. In supervised learning, each pattern belongs to a certain predefined class. A class can be seen as a decision that has to be made. All the observations combined with their class labels are known as a data set. When a new observation is received, that observation is classified based on previous experience.<ref name="Classifiers"/>
The simplest AI applications can be divided into two types: classifiers ("if shiny then diamond") and controllers ("if shiny then pick up"). Controllers do, however, also classify conditions before inferring actions, and therefore classification forms a central part of many AI systems. [[Classifier (mathematics)|Classifiers]] are functions that use [[pattern matching]] to determine a closest match. They can be tuned according to examples, making them very attractive for use in AI. These examples are known as observations or patterns. In supervised learning, each pattern belongs to a certain predefined class. A class can be seen as a decision that has to be made. All the observations combined with their class labels are known as a data set. When a new observation is received, that observation is classified based on previous experience.<ref name="Classifiers"/>
The simplest AI applications can be divided into two types: classifiers ("if shiny then diamond") and controllers ("if shiny then pick up"). Controllers do, however, also classify conditions before inferring actions, and therefore classification forms a central part of many AI systems. Classifiers are functions that use pattern matching to determine a closest match. They can be tuned according to examples, making them very attractive for use in AI. These examples are known as observations or patterns. In supervised learning, each pattern belongs to a certain predefined class. A class can be seen as a decision that has to be made. All the observations combined with their class labels are known as a data set. When a new observation is received, that observation is classified based on previous experience.
The simplest AI applications can be divided into two types: classifiers ("if shiny then diamond") and controllers ("if shiny then pick up"). Controllers do, however, also classify conditions before inferring actions, and therefore classification forms a central part of many AI systems. Classifiers are functions that use pattern matching to determine a closest match. They can be tuned according to examples, making them very attractive for use in AI. These examples are known as observations or patterns. In supervised learning, each pattern belongs to a certain predefined class. A class can be seen as a decision that has to be made. All the observations combined with their class labels are known as a data set. When a new observation is received, that observation is classified based on previous experience.
最简单的AI应用程序可以分为两类: '''<font color=#ff8000>分类器 Classifiers</font>''' (“ if shiny then diamond”)和'''<font color=#ff8000>控制器 Controllers</font>''' (“ if shiny then pick up”)。然而，控制器在推断前也对条件进行分类，因此分类构成了许多AI系统的核心部分。分类器是使用匹配模式来判别最接近的类别的函数。它们可以根据例子进行调整，使它们在AI的应用中更有效。这些例子被称为观察或模式。在监督式学习中，每个模式都属于某个预定义的类别。可以把一个类看作是一个必须做出的决定。所有的观测和它们的类标签被称为数据集。当接收一个新观察时，这个观察结果将根据以前的经验被分类。
最简单的AI应用程序可以分为两类: '''<font color=#ff8000>分类器 Classifiers</font>''' (“若闪光，则为钻石”)和'''<font color=#ff8000>控制器 Controllers</font>''' (“若闪光，则捡起来”)。然而，控制器在推断前也对条件进行分类，因此分类构成了许多AI系统的核心部分。分类器一组是使用匹配模式来判断最接近的匹配的函数。它们可以根据样例进行性能调优，使它们在AI应用中更有效。这些样例被称为“观察”或“模式”。在监督学习中，每个模式都属于某个预定义的类别。可以把一个类看作是一个必须做出的决定。所有的样例和它们的对应的类别标签被称为数据集。当接收一个新样例时，它会被分类器根据以前的经验进行分类。
--[[用户:Thingamabob|Thingamabob]]（[[用户讨论:Thingamabob|讨论]]） 分类器(“ if shiny then diamond”)和控制器(“ if shiny then pick up”) 一句不能准确翻译
--[[用户:Thingamabob|Thingamabob]]（[[用户讨论:Thingamabob|讨论]]） 分类器(“ if shiny then diamond”)和控制器(“ if shiny then pick up”) 一句不能准确翻译
--[[用户:Qige96|Ricky]]（[[用户讨论:Qige96|讨论]]）已解决。这里主要是突出分类器的工作是下判断，而控制器的工作是做动作。
A classifier can be trained in various ways; there are many statistical and [[machine learning]] approaches. The [[decision tree learning|decision tree]]<ref name="Decision tree"/> is perhaps the most widely used machine learning algorithm.{{sfn|Domingos|2015|p=88}} Other widely used classifiers are the [[Artificial neural network|neural network]],<ref name="Neural networks"/>
A classifier can be trained in various ways; there are many statistical and [[machine learning]] approaches. The [[decision tree learning|decision tree]]<ref name="Decision tree"/> is perhaps the most widely used machine learning algorithm.{{sfn|Domingos|2015|p=88}} Other widely used classifiers are the [[Artificial neural network|neural network]],<ref name="Neural networks"/> [[Gaussian mixture model]],<ref name="Gaussian mixture model"/> and the extremely popular [[naive Bayes classifier]].{{efn|Naive Bayes is reportedly the "most widely used learner" at Google, due in part to its scalability.{{sfn|Domingos|2015|p=152}}}}<ref name="Naive Bayes classifier"/> Classifier performance depends greatly on the characteristics of the data to be classified, such as the dataset size, distribution of samples across classes, the dimensionality, and the level of noise. Model-based classifiers perform well if the assumed model is an extremely good fit for the actual data. Otherwise, if no matching model is available, and if accuracy (rather than speed or scalability) is the sole concern, conventional wisdom is that discriminative classifiers (especially SVM) tend to be more accurate than model-based classifiers such as "naive Bayes" on most practical data sets.<ref name="Classifier performance"/>{{sfn|Russell|Norvig|2009|loc=18.12: Learning from Examples: Summary}}
[[Gaussian mixture model]],<ref name="Gaussian mixture model"/> and the extremely popular [[naive Bayes classifier]].{{efn|Naive Bayes is reportedly the "most widely used learner" at Google, due in part to its scalability.{{sfn|Domingos|2015|p=152}}}}<ref name="Naive Bayes classifier"/> Classifier performance depends greatly on the characteristics of the data to be classified, such as the dataset size, distribution of samples across classes, the dimensionality, and the level of noise. Model-based classifiers perform well if the assumed model is an extremely good fit for the actual data. Otherwise, if no matching model is available, and if accuracy (rather than speed or scalability) is the sole concern, conventional wisdom is that discriminative classifiers (especially SVM) tend to be more accurate than model-based classifiers such as "naive Bayes" on most practical data sets.<ref name="Classifier performance"/>{{sfn|Russell|Norvig|2009|loc=18.12: Learning from Examples: Summary}}
分类器可以通过多种方式进行训练;，比如许多统计学和机器学习方法。决策树可能是应用最广泛的机器学习算法。其他使用广泛的分类器还有神经网络、<ref name="Neural networks"/>K最近邻算法{{efn|The most widely used analogical AI until the mid-1990s{{sfn|Domingos|2015|p=187}}}}<ref name="K-nearest neighbor algorithm"/>、核方法（比如支持向量机）{{efn|SVM displaced k-nearest neighbor in the 1990s{{sfn|Domingos|2015|p=188}}}}<ref name="Kernel methods"/>、'''<font color=#ff8000> 高斯混合模型 Gaussian Mixture Mode</font>'''<ref name="Gaussian mixture model"/>，以及非常流行的'''<font color=#ff8000>朴素贝叶斯分类器 Naive Bayes Classifier</font>'''<ref name="Naive Bayes classifier"/>。分类器的分类效果在很大程度上取决于待分类数据的特征，如数据集的大小、样本跨类别的分布、维数和噪声水平。如果假设的模型很符合实际数据，则基于这种模型的分类器就能给出很好的结果。否则，传统观点认为如果没有匹配模型可用，而且只关心准确性(而不是速度或可扩展性) ，在大多数实际数据集上鉴别分类器(尤其是支持向量机)往往比基于模型的分类器(如“朴素贝叶斯”)更准确。<ref name="Classifier performance"/>{{sfn|Russell|Norvig|2009|loc=18.12: Learning from Examples: Summary}}
===人工神经网络 Artificial neural networks ===
===人工神经网络 Artificial neural networks ===
{{Main|Artificial neural network|Connectionism}}
{{Main|Artificial neural network|Connectionism}}
A neural network is an interconnected group of nodes, akin to the vast network of [[neurons in the human brain.]]
A neural network is an interconnected group of nodes, akin to the vast network of [[neurons in the human brain.]]
Neural networks were inspired by the architecture of neurons in the human brain. A simple "neuron" N accepts input from other neurons, each of which, when activated (or "fired"), cast a weighted "vote" for or against whether neuron N should itself activate. Learning requires an algorithm to adjust these weights based on the training data; one simple algorithm (dubbed "fire together, wire together") is to increase the weight between two connected neurons when the activation of one triggers the successful activation of another. The neural network forms "concepts" that are distributed among a subnetwork of shared neurons that tend to fire together; a concept meaning "leg" might be coupled with a subnetwork meaning "foot" that includes the sound for "foot". Neurons have a continuous spectrum of activation; in addition, neurons can process inputs in a nonlinear way rather than weighing straightforward votes. Modern neural networks can learn both continuous functions and, surprisingly, digital logical operations. Neural networks' early successes included predicting the stock market and (in 1995) a mostly self-driving car. In the 2010s, advances in neural networks using deep learning thrust AI into widespread public consciousness and contributed to an enormous upshift in corporate AI spending; for example, AI-related M&A in 2017 was over 25 times as large as in 2015.
Neural networks were inspired by the architecture of neurons in the human brain. A simple "neuron" N accepts input from other neurons, each of which, when activated (or "fired"), cast a weighted "vote" for or against whether neuron N should itself activate. Learning requires an algorithm to adjust these weights based on the training data; one simple algorithm (dubbed "fire together, wire together") is to increase the weight between two connected neurons when the activation of one triggers the successful activation of another. The neural network forms "concepts" that are distributed among a subnetwork of shared neurons that tend to fire together; a concept meaning "leg" might be coupled with a subnetwork meaning "foot" that includes the sound for "foot". Neurons have a continuous spectrum of activation; in addition, neurons can process inputs in a nonlinear way rather than weighing straightforward votes. Modern neural networks can learn both continuous functions and, surprisingly, digital logical operations. Neural networks' early successes included predicting the stock market and (in 1995) a mostly self-driving car. In the 2010s, advances in neural networks using deep learning thrust AI into widespread public consciousness and contributed to an enormous upshift in corporate AI spending; for example, AI-related M&A in 2017 was over 25 times as large as in 2015.
神经网络的诞生受到人脑神经元结构的启发。一个简单的“神经元”''N'' 接受来自其他神经元的输入，每个神经元在被激活(或者说“放电”)时，都会对N是否应该被激活按一定的权重赋上值。学习的过程需要一个根据训练数据调整这些权重的算法; 一个被称为“相互放电，彼此联系”简单的算法在一个神经元激活触发另一个神经元的激活时增加两个连接神经元之间的权重。神经网络中形成一种分布在一个共享的神经元子网络中的”概念”，这些神经元往往一起放电; ”腿”的概念可能和”脚”概念的子网络相结合，后者包括”脚”的发音。神经元有一个连续的激活谱; 此外，神经元还可以用非线性的方式处理输入，而不是简单地加权求和。现代神经网络可以学习连续函数甚至的数字逻辑运算。神经网络早期的成功包括预测股票市场和自动驾驶汽车(1995年)。2010年代，神经网络使用深度学习取得巨大进步，也因此将AI推向了公众视野里，并促使企业对AI投资急速增加; 例如2017年与AI相关的并购交易规模是2015年的25倍多。
Proposed "universal intelligence" tests aim to compare how well machines, humans, and even non-human animals perform on problem sets that are generic as possible. At an extreme, the test suite can contain every possible problem, weighted by Kolmogorov complexity; unfortunately, these problem sets tend to be dominated by impoverished pattern-matching exercises where a tuned AI can easily exceed human performance levels.
Proposed "universal intelligence" tests aim to compare how well machines, humans, and even non-human animals perform on problem sets that are generic as possible. At an extreme, the test suite can contain every possible problem, weighted by Kolmogorov complexity; unfortunately, these problem sets tend to be dominated by impoverished pattern-matching exercises where a tuned AI can easily exceed human performance levels.
“通用智能”测试旨在比较机器、人类甚至非人类动物在尽可能通用的问题集上的表现。在极端情况下，测试集可以包含所有可能出现的问题，由柯氏复杂性赋权重; 可是这些问题集往往是用有限的模式匹配练习完成的，在这些练习中，优化过的AI可以轻易地超过人类。
“通用智能”测试旨在比较机器、人类甚至非人类动物在尽可能通用的问题集上的表现。在极端情况下，测试集可以包含所有可能出现的问题，由柯氏复杂性赋权重; 可是这些问题集往往是用有限的模式匹配练习完成的，在这些练习中，优化过的AI可以轻易地超过人类。
== 应用 Applications{{anchor|Goals}} ==
== 应用 Applications{{anchor|Goals}} ==