Artificial intelligence (AI) is a discipline that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence.

symbolism

connectionism

Behaviorism

Artificial intelligence research content

Application areas of artificial intelligence

Knowledge is human beings’ understanding and mastery of the natural world, human society, ways of thinking and laws of movement; it is the experience accumulated by human beings in long-term life and social practice, scientific research and experiments; it is the reorganization of the human brain through thinking , an information structure formed by associating relevant information obtained in practice.

Predicate logic is a logic based on the analysis of predicates in propositions. First-order predicate logic is the most intuitive type of predicate logic. Predicate is a language component used to describe the nature, status and relationship between individuals. For example, for the previous section If used

Most of the problem-solving methods used in artificial intelligence research use heuristic search methods. In other words, most problem solving methods are by searching within a certain possible solution space. Find an optimal solution to solve the problem. This problem representation and solution based on the solution space The solution is the state space representation

Productions are often used to describe facts, rules, and their The degree of uncertainty is suitable for representing factual knowledge and rule knowledge.

Semantic network is a network diagram that represents knowledge through concepts and semantic relationships (or semantic connections). From the perspective of graph theory, a semantic network is a directed graph with labels. Semantic network consists of nodes and arcs between nodes. Nodes represent various things, concepts, situations, attributes, states, things Parts and actions, etc.: Arc represents various semantic relationships between the nodes it connects. Both nodes and arcs must With labels, various objects and various semantic relationships between objects can be distinguished.

A frame is a data structure that describes the properties of an object. is a network composed of several nodes and relationships, in which objects Representing an object, event, or concept is a descriptive object The data structure is composed of several nodes and relationships. network. A frame consists of several structures called "slots". Each slot can be divided into several "sides" according to actual conditions.

Slot name 1: side name/1 side value 1, side value 1,..., side value 1p Side name 12, side value 121, side value 122,..., side value 127 Side name 1m Side value 1mi, side value 1m2,..., side value 1mpm Slot name 2: side name 21 side value 211, side value 212,..., side value 21p Side name 22, side value 221, side value 222,..., side value 2202 Side name 2m Side value 2ml, side value 2m2,..., side value 2mpm ... Slot name n: side name m side value n11, side value m2,..., side value mipl Side name n2 side value n21, side value m22,..., side value m2p2 Side name nm, side value nmi, side value nm2,..., side value nmpm Constraints: Constraints:

Reasoning refers to starting from known facts and using the knowledge you have mastered according to a certain strategy. The process of deriving the factual conclusions contained therein or deducing some new conclusions.

Natural deductive reasoning refers to starting from a set of facts known to be true and directly using propositional logic or predicate The process of deriving conclusions from the inference rules in word logic.

Inductive deductive reasoning is a machine reasoning technology based on the principle of induction

How to find usable knowledge, that is, how to determine the line of reasoning so that we can pay as little as possible The problem is solved satisfactorily at a reasonable price. The search is based on the actual situation of the problem, According to certain strategies or rules, find available knowledge from the knowledge base, so as to Search is a core technology in artificial intelligence and an integral part of reasoning It is directly related to the performance and operating efficiency of the intelligent system. in search questions The main job is to find the right search strategy. search strategy) the state space or The method by which the problem space is expanded also determines the order in which states or problems are accessed.

Blind search strategy is also called information-free search strategy, that is to say, during the search process , only searches according to the pre-specified search strategy without any intermediate information to change these strategies. Commonly used blind search strategies include breadth-first search Depth first search and equal cost search, etc.

Subtopic breadth-first search is also called breadth-first search. Its basic idea is to start from the starting node, expand (or search) the node layer by layer, and at the same time check whether it is the target node. For example, for the search tree shown in Figure 4-2, the search order should be A→B→C→D→E→F→G→H.

The basic idea of ​​depth-first search is to start from the starting node and select a node among its sub-nodes for investigation. If it is not the target node, select a node among the sub-nodes of the sub-node for investigation, and continue searching downwards like this. If it is found that the target node cannot be reached, return to the previous node, then select another child node of the node to search downwards, and so on until the target node is searched or the search is complete. to the local node.

Heuristic search strategy, also known as informative search strategy, refers to the Use information relevant to the problem to guide the search in the most beneficial direction, Speed ​​up search and improve search efficiency. Heuristic search strategies involved The important contents include heuristic information and valuation functions. Commonly used heuristic searches Search strategies include A search and A search.

Computational intelligence (CI) is the ability of people to Inspired by the laws and biological intelligence mechanisms, a set of algorithms imitated and designed based on their principles , for solving complex real-world problems. At present, computational intelligence has not yet unified The definition of one is as follows: some scholars’ different descriptions of computational intelligence are listed below.

Computing classification is divided into evolutionary computing, swarm intelligence, neural computing, fuzzy computing, immune computing, artificial life

Evolutionary computation, a mature model of biological evolution mechanism design with high robustness and An extensive global optimization method with self-organizing, adaptive, and self-learning characteristics. It is not limited by the nature of the problem and can effectively handle complex problems that are difficult to solve with traditional optimization algorithms. question.

Swarm intelligence is an intelligent optimization method inspired by the intelligence phenomenon of biological groups in nature. ization method is one of the key technologies in the field of computational intelligence

Find the shortest path that consumes the least energy. best path

Machine learning (machinelearnng) is to learn how to complete tasks from data through various algorithms. A discipline that accomplishes tasks and acquires methods for accomplishing tasks. It can automatically process data Analyze and obtain rules or models from them, and then use the rules or models to analyze unknown data Make predictions. It is the core of artificial intelligence and an important way to make computers intelligent. At present, there is no generally accepted and accurate definition of machine learning. Some of them are listed below. Scholars’ description of machine learning (1) Machine learning is the study of how to use machines to simulate anthropology A discipline of learning activities.

Supervised learning uses labeled data sets to train the learning model, and then obtains Prediction model, and finally a learning method that uses the test set to evaluate the performance of the prediction model.

Unsupervised learning is a learning method that discovers potential relationships between data in unlabeled data sets.