-
Machine Learning Editor
This entry is compiled and applied to the scientific entry of "Science China" encyclopedia. Machine Learning (machines learning, ML) is a multidisciplinary interdisciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithmic complexity theory and many other disciplines. Specialized in computer simulation or realization of human learning behavior, in order to acquire new knowledge or skills, reorganize the existing knowledge structure to continuously improve their performance. It is the core of artificial intelligence, is the fundamental way to make the computer intelligent, its application throughout the field of artificial intelligence, it mainly uses induction, synthesis rather than deduction.
-
Chinese name
-
Machine learning
-
Foreign names
Machine
-
Learning, ML
-
sex quality
-
multidisciplinary interdisciplinary
-
collar domain
-
probability theory, statistics, approximation theory
- Research significance
Learning is an important intelligent behavior of human beings, but what is learning is a long-standing controversy. Sociologists, logics and psychologists all have their own different views. For example, Langley (1996) Defines machine learning as "machine learning is a science of artificial intelligence, and the main research object in this field is AI, especially how to improve the performance of specific algorithms in experiential learning." (Machine learning are a science of the artificial. The field ' s main objects of study is artifacts, specifically algorithms that improve their performance with experience. ' Tom Mitchell's Machine Learning (1997) explains some of the concepts in information theory in detail, in which machine learning is defined as "machine learning is the study of computer algorithms that can be automatically improved through experience." (Machine learning are the study of computer algorithms that improve automatically through experience.) Alpaydin (2004) at the same time put forward their own The definition of machine learning, "machine learning is to use data or previous experience to optimize the performance standards of computer programs." "(Machine learning are programming computers to optimize a performance criterion using example data or past experience.) although In order to facilitate discussion and to estimate the progress of the subject, it is necessary to define machine learning, even if the definition is incomplete and inadequate. As the name implies, machine learning is a discipline that studies how machines are used to simulate human learning activities. A slightly stricter formulation is that machine learning is a learning machine to acquire new knowledge and skills, and to identify existing knowledge. The "machine" here refers to computers, electronic computers, neutron computers, photon computers, or neural computers, and so on. Can machines have the ability to learn as humans do? In 1959 Samuel (Samuel) designed a chess program that had the ability to learn and to improve his chess skills in a constant game of chess. 4 years later, the program defeated the designer himself. Over the past 3 years, the program has beaten the United States to a 8-year-old undefeated champion. This program shows people the ability of machine learning and puts forward many thoughtful social problems and philosophicalLearning problems. The ability of the machine to exceed the human, a lot of negative opinion of the person's main argument is: The machine is man-made, its performance and action is entirely prescribed by the designer, so in any case its ability will not exceed the designer himself. This opinion is true for machines that do not have the ability to learn, but the ability to learn is worth considering, because the machine's capabilities are constantly improving in application, and after a while, the designer himself does not know what level it is capable of. There are several definitions of machine learning: "Machine learning is a science of artificial intelligence, and the main research object in this field is artificial intelligence, especially how to improve the performance of specific algorithms in experiential learning." "Machine learning is the study of computer algorithms that can be automatically improved through experience." "Machine learning is using data or previous experience to optimize the performance standards of computer programs. A frequently cited English definition is: A computer program is said to learn from experience E with respect to some class of tasks T and performance Measure P, if its performance at the tasks in T, as measured by P, improves with experience E. Machine learning has a very wide range of applications, such as: Data mining, computer vision, self- such as language processing, biometric identification, search engines, medical diagnostics, detection of credit card fraud, stock market analysis, DNA sequencing, speech and handwriting recognition, strategy games, and robotic applications. History machine learning is a relatively young branch of artificial intelligence research, its development process can be divided into 4 periods. The first stage was in the the mid 1950s leaves to the 60 mid-leaf, belongs to the warm period. The second stage was in the the mid 1960s leaf to the mid-70, known as the cool-down period of machine learning. The third stage is from the mid 1970s to the mid-80, called the revival period. The latest phase of machine learning began in 1986. The important performance of machine learning in the new stage is as follows: (1) machine learning has become a new frontier subject and form a course in colleges and universities. It combines applied psychology, biology and neurophysiology as well as mathematics, automation, and computer science to form the basis of machine learning theory. (2) Combining various learning methods, the study of various forms of integrated learning system is emerging. In particular, the coupling of learning symbolic learning can better solve the problem of acquisition and refinement of knowledge and skills in continuous signal processing. (3) The unity of knowledge of machine learning and artificial intelligence is forming. For example, learning is combined with problem solving, and knowledge expression facilitates learning, resulting in a universal intelligence system soAR block Learning. The case-based approach combining analogical learning with problem solving has become an important direction of experiential learning. (4) The scope of application of various learning methods has been expanded, and part of the product has been formed. The Knowledge acquisition tool of inductive learning has been widely used in diagnostic classification expert system. Connection learning prevails in the recognition of sound and text. Analytical learning has been used in the design of comprehensive expert system. Genetic algorithm and reinforcement learning have a good application prospect in engineering control. The neural network connection learning which is coupled with the symbolic system will play a role in the intelligent management of the enterprise and the motion planning of the intelligent robot. (5) Academic activities related to machine learning are more active than ever. In addition to the annual machine learning Workshop, there are also computer learning theory conferences and genetic algorithms conferences. The main strategy learning is a complex intelligent activity, the learning process and the reasoning process is closely linked, according to the use of reasoning in learning, machine learning strategy can be broadly divided into 4 kinds of-mechanical learning, through teaching learning, analogy learning and through case learning. The more reasoning used in learning, the more powerful the system. The basic structure represents the basic structure of the learning system. The environment provides some information to the learning portion of the system, and the Learning section uses this information to modify the knowledge base to improve the performance of the system's execution of the tasks, performing the tasks according to the knowledge Base, and giving feedback to the learning section. In the specific application, the environment, the knowledge base and the implementation part decide the specific work content, the study part needs to solve the question completely from above 3 parts to determine. Here we describe the impact of these 3 sections on the design learning system. The most important factor affecting the design of the learning system is the information provided by the environment to the system. Or, more specifically, the quality of the information. The Knowledge Base stores the general principles that govern the execution of some actions, but the information provided by the environment to the learning system is varied. If the quality of information is relatively high, and the difference between the general principle is relatively small, the learning section is easier to deal with. If the learning system is provided with a haphazard guide to perform specific actions specific information, then the learning system needs to get enough data, delete unnecessary details, to summarize the promotion, the general principle of guiding action, into the knowledge base, so that the task of learning part is more onerous, design is more difficult. Because the information that the learning system obtains is often incomplete, so the reasoning of the learning system is not entirely reliable, the rules it summarizes may be correct, or may not be correct. This is to be tested by the effect of execution. The correct rules can improve the efficiency of the system and should be retained; Incorrect rules should be modified or deleted from the database. Knowledge Base is the second factor that influences the design of learning system. There are many forms of knowledge representation, such as eigenvectors, first-order logic statements, production rules, semantic networks and frameworks, and so on. These representations each have their own characteristics, in the choice of presentation to take into account the following 4 aspects: (1) strong expression ability. (2) Easy reasoning. (3) Easy to modify the knowledge base. (4) Knowledge representation is easy to expand.One of the last things to be explained in the knowledge base is that the learning system cannot acquire knowledge without any knowledge, and each learning system requires some knowledge to understand the information provided by the environment, to analyze comparisons, to make assumptions, to examine and modify these assumptions. Thus, more precisely, the learning system is the extension and improvement of existing knowledge. The executive part is the core of the whole learning system, because the part of the action is to learn some of the actions to improve. There are 3 issues related to the implementation segment: complexity, feedback and transparency. code example
| 1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556 |
//在gcc-4.7.2下编译通过。//命令行：g++-Wall-ansi-O2test.cpp-otest#include<iostream>usingnamespacestd;voidinput(int&oper,constboolmeth){//meth为true则只判断1，为false则判断1或0while(true){cin>>oper;if(meth&&oper==1)break;elseif(oper==0||oper==1)break;cout<<"输入错误，请重新输入。"<<endl;//判断参数cin.sync();//避免极端输入导致死循环cin.clear();}}intmain(void){cout<<"1+1=2吗？那要看您怎么教我了，不要惊讶我会学习的"<<endl;intladd,radd,aprs,rcnt(0),wcnt(0);//定义输入与结果，正确次数与错误次数cout<<"开始学习……"<<endl;for(inti(0);i!=10;++i){cout<<"参数1（必须是1）："<<flush;//提示输入参数input(ladd,true);cout<<"参数2（必须是1）："<<flush;input(radd,true);cout<<"结果："<<(ladd+radd)<<endl;//输出结果cout<<"您对这满意吗（满意输入1，不满意输入0）："<<flush;//评价等级input(aprs,false);if(aprs)//判断用户评价++rcnt;else++wcnt;cout<<"正确次数："<<rcnt<<"错误次数："<<wcnt<<endl;//错误次数}if(rcnt>wcnt)//判断学习结果cout<<"主人告诉我1+1=2。"<<endl;elseif(rcnt<wcnt)cout<<"主人告诉我1+1!=2。"<<endl;elsecout<<"我不明白主人是什么意思。"<<endl;intterm;//退出部分cout<<"您对我的表现满意吗？满意请输入1不满意请输入0："<<flush;input(term,false);if(term)cout<<"谢谢我会继续努力学习"<<endl;elsecout<<"谢谢我会继续努力学习D"<<endl;//cin>>term;//在Windows上测试时启用return0;} |
This procedure will judge the execution result "1+1=2" according to your evaluation.
In fact, only the simplest if else for statement
This is an example of machine learning that can be learned through environmental impact.
In this case, it is not difficult to see that, under the guidance of manual error, the machine will give the wrong answer is not equal to 2.
So this kind of learning method, must be in the right guidance under the practice, otherwise you will get the worst results.
After the completion of the study, the computer will record the results of this study, stored in the database, the next time the corresponding task, then the results are paged out. Classification learning strategy based on learning strategy refers to the inference strategy adopted by the system in the course of learning. A learning system is always composed of two parts: Learning and environment. Information is provided by the environment (such as a book or teacher), and the learning section realizes the conversion of information, remembers it in a form that can be understood, and obtains useful information from it. In the course of learning, the less reasoning students use, the greater their reliance on the teacher (environment), and the heavier the burden on teachers. The classification criteria of learning strategies are classified according to how much reasoning and difficulty the students need to realize the information conversion, and the following six basic types are divided into two categories: 1) Learners of mechanical learning (Rote learning) without any inference or other knowledge conversion, Direct absorption of the information provided by the environment. Like Samuel's Checkers program, Newell and Simon's LT system. This kind of learning system mainly considers how to index the knowledge of storage and make use of it. The learning method of the system is to learn directly through pre-programmed and well-structured programs, learners do not do any work, or by directly receiving the established facts and data to learn, the input information does not make any inference. 2) Teach learning (learning from instruction or learning by being told) students obtain information from the environment (teachers or other sources of information such as textbooks, etc.) and convert the knowledge into an internal available representation, and the new knowledge and the original knowledge organically integrated into one. Therefore, students are required to have a certain degree of reasoning ability, but the environment still has to do a lot of work. Teachers present and organize knowledge in some form, so that the knowledge that students possess can be continuously increased. This method of learning is similar to that of a human society, and the task of learning is to establish a system that allows it to receive teaching and advice and to effectively store and apply the knowledge learned. Many expert systems use this method to achieve knowledge acquisition when building a knowledge base. A typical example of teaching learning is the Foo program. 3) Deductive learning (learning by deduction) The form of reasoning used by students is deductive reasoning. Reasoning from the axiom, through the logical transformation to deduce the conclusion. This reasoning is the process of "fidelity" transformation and Special (specialization), which enables students to acquire useful knowledge in the process of reasoning. This learning approach includes macro manipulation (macro-operation) learning, knowledge editing, and block (Chunking) techniques. The inverse process of deductive inference is inductive reasoning. 4) Analogical Learning (learning by analogy) utilizes knowledge similarity in two different domains (source domain, target domain), which can be derived from the knowledge of the source domain (including similar characteristics and other properties) by analogy.The corresponding knowledge of the target domain is deduced to realize the learning. The analogy learning system can transform an existing computer application system into a new field to accomplish a similar function that was not originally designed. Analogical learning requires more reasoning than the above three ways of learning. It is generally required to first retrieve the available knowledge from the knowledge source (source domain), then convert it into a new form and use it in the new state (target domain). Analogy learning plays an important role in the history of human science and technology, and many scientific discoveries are obtained by analogy. The famous Rutherford analogy, for example, reveals the mysteries of atomic structures by analogy of atomic structures (target domains) with the Solar system (source domain). 5) Interpretation-based learning (explanation-based learning, EBL) students based on the concept of goals provided by teachers, an example of this concept, domain theory and operational guidelines, first constructs an explanation to explain why the example satisfies the concept of the target, It then interprets the generalization as a sufficient condition for the objective concept to satisfy the operational criteria. EBL has been widely used in the knowledge base to refine and improve the performance of the system. The famous EBL system has the T.mitchell of Dichoen (G.dejong) Genesis, Mitchell (LEXII) Minton and Leap, and S.minton (prodigy). 6) Inductive learning (learning from induction) inductive learning is an example or counter-example of a concept provided by a teacher or environment, allowing students to derive a general description of the concept through inductive reasoning. This kind of learning is much more than teaching learning and deductive learning, because the environment does not provide generic conceptual descriptions (such as axioms). To some extent, inductive learning is much more speculative than analogy, since no similar concept can be used as a "source concept". Inductive learning is the most basic and mature learning method, which has been widely researched and applied in the field of artificial intelligence. The knowledge obtained from the Representation classification learning system based on the acquired knowledge may include: Behavioral rules, description of physical objects, problem solving strategies, various classifications and other types of knowledge used in task implementations. For the knowledge acquired in learning, there are the following expressions: 1 The object of algebraic expression parameter learning is to adjust the algebraic expression parameters or coefficients of a fixed function form to achieve an ideal performance. 2) Decision trees Use decision trees to divide the genus of objects, each inner node of the tree corresponds to an object attribute, and each side corresponds to an optional value for those attributes, and the leaf node of the tree corresponds to each basic classification of the object. 3) Formal grammar in the recognition of a particular language of learning, through the language of a series of expressions to form a formal grammar of the language. 4) The production rules are expressed as conditional-action pairs, which have been used very extensively. The learning behaviors in the learning system are: generation, generalization, specialization (speciAlization) or synthetic production rules. 5) The basic components of formal logic expressions are propositions, predicates, variables, constrained variable range statements, and embedded logical expressions. 6) Graph and network systems use graph matching and graph conversion schemes to effectively compare and index knowledge. 7) Framework and schema (schema) each frame contains a set of slots that describe various aspects of a thing (concept and individual). 8) computer programs and other process codes acquiring this form of knowledge is intended to achieve a capability to achieve a particular process, rather than to infer the internal structure of the process. 9) Neural network This is mainly used in connection learning. Learning the acquired knowledge, finally summed up as a neural network. 10) A combination of multiple representations sometimes the knowledge acquired in a learning system needs to be comprehensively applied to the above mentioned several knowledge representations. The knowledge representation can be divided into two categories based on the degree of refinement represented: a coarse-grained symbol representing a high degree of generalization 、?? The fine-grained sub-symbol (sub-symbolic), which has a low generalization level, is indicated. such as decision tree, formal grammar, production rules, formal logic expressions, frames and patterns are symbolic representation classes, whereas algebraic expression parameters, graphs and networks, neural networks, etc. are sub-symbolic representations. The most important applications by application domain are: Expert system, cognitive simulation, programming and problem solving, data mining, network Information Service, image recognition, fault diagnosis, natural language understanding, robot and game. From the task type reflected in the execution part of machine learning, most of the applied research areas are mainly focused on the following two categories: Classification and problem solving. (1) The classification task requires the system to analyze the unknown pattern (description of the pattern) based on known classification knowledge to determine the genus of the input patterns. The corresponding learning goal is to learn the criteria for classification (such as classification rules). (2) The Problem Solver task requires,?? for a given target State Look for an action sequence that transforms the current state into a target state; machine learning in this field is mostly focused on learning to acquire knowledge (such as search control knowledge, heuristic knowledge, etc.) that can improve the efficiency of problem solving. Comprehensive classification takes into account the historical origins, knowledge representation, inference strategies, similarity of results evaluation, relative centralization of researchers ' communication and application fields. The machine learning method [1] is divided into the following six categories: 1) Empirical inductive learning (empirical inductive learning) empirical inductive learning using some data-intensive empirical methods (such as version space method, ID3 method, Law discovery method) to induce the study of examples. The examples and learning results are usually represented by attributes, predicates, relationships, and other symbols. It is equivalent to inductive learning based on the classification of learning strategies, but it deducts the part of join learning, genetic algorithm and reinforcement learning. 2) Analytical learning (analyticLearning) The analytical learning method is based on one or a few instances, using domain knowledge for analysis. Its main characteristics are: • Inference strategy is mainly deductive, rather than inductive; • Use past problem solving experiences (examples) to guide new problem solving, or to generate search control rules that can be used more effectively in domain knowledge. The goal of analytical learning is to improve the performance of the system, not the new concept description. Analytical learning includes application of interpretive learning, deductive learning, multi-level structure blocks, and macro-manipulation learning techniques. 3) Analogy learning it is equivalent to learning based on the analogy of Learning strategy classification. A more compelling study in this type of study is learning by analogy with specific examples of past experiences, known as paradigm-based learning (case_based learning), or example learning. 4) Genetic algorithm (genetic algorithm) genetic algorithm simulates mutation, exchange of biological reproduction and Darwin's natural selection (survival of the fittest in every ecological environment). It encodes the possible solution of the problem into a vector, called an individual, each element of the vector is called a gene, and uses the objective function (corresponding to the natural selection criteria) to evaluate each individual in a group (a collection of individuals), to select, Exchange, and mutate the individual according to the evaluation Value (fitness), So as to get new groups. Genetic algorithms are suitable for very complex and difficult environments, for example, with a lot of noise and irrelevant data, things are constantly updated, problem targets cannot be clearly and precisely defined, and the value of current behavior can be determined through a lengthy execution process. As with neural networks, the research of genetic algorithms has developed into an independent branch of artificial intelligence, whose representative character is Hollede (J.h.holland). 5) Connection Learning A typical connection model is implemented as an artificial neural network, which consists of some simple computational units called neurons and weighted joins between cells. 6) Enhanced Learning (reinforcement learning) enhancement learning is characterized by the determination and optimization of action choices through interaction with the environment's exploratory (trial and error) to achieve so-called sequence decision tasks. In this task, the learning mechanism interacts with the environment by selecting and executing actions that cause changes in the state of the system and the possibility of getting some kind of reinforcement signal (immediate return). Strengthening the signal is the reward and punishment of the system behavior. The goal of the system learning is to find a suitable action selection strategy, that is, in any given state to choose which action method, so that the resulting action sequence can obtain some of the best results (such as cumulative immediate return maximum). In the comprehensive classification, empirical inductive learning, genetic algorithm, connected learning and reinforcement learning are all inductive learning, in which the empirical inductive learning adopts symbolic representation, while the genetic algorithm, the connecting learning and the reinforcement learning are represented by the sub-symbol, and the analytical learning belongs to the deductive learning. In fact, the analogy strategy can be regarded as the synthesis of inductive and deductive strategies. Thus the most basicLearning strategies are only inductive and deductive. From the perspective of learning content, the use of inductive strategy learning is based on the induction of input, the knowledge is clearly more than the original system can contain the scope of knowledge, the results of the study changed the system of knowledge deduction closure, so this type of learning can also be called knowledge-level learning; While the use of deductive strategy learning, although the knowledge can improve the efficiency of the system, but still can be the original system of the knowledge base, that is, the knowledge has not changed the deductive closure of the system, so this type of learning is also known as symbol-level learning. Learning Form Category 1) supervised learning (supervised learning) supervised learning, i.e. providing right and wrong instruction during mechanical learning. It is generally true that the data set contains the final result (0,1). The algorithm lets the machine self-reduce the error. This type of learning is primarily used for classification and prediction (regression & classify). Supervised learning learns a function from a given set of training data, and when new data arrives, the result can be predicted based on this function. The training set requirements for supervised learning are both input and output, as well as features and goals. The goal of the training set is to be marked by people. Common supervised learning algorithms include regression analysis and statistical classification. 2) Unsupervised learning (unsupervised learning) non-supervised learning is also known as inductive learning (clustering) using K-mode (Kmeans), establishing the Center (centriole), through cyclic and decrement operations (iteration& Descent) to reduce the error to achieve the purpose of classification. Research in the field of machine learning focuses on the following three areas: (1) Task-oriented research and research and analysis of learning systems that improve the performance of a set of scheduled tasks. (2) Cognitive models study human learning processes and perform computer simulations. (3) Theoretical analysis theoretically explores various possible learning methods and algorithms that are independent of the field of application machine learning is another important field of artificial intelligence application after expert system, and it is also one of the core research topics in artificial intelligence and neural computing. Existing computer systems and artificial intelligence systems do not have the ability to learn, at most, only a very limited ability to learn, and therefore can not meet the new requirements of technology and production. The discussion of machine learning and the progress of machine learning research will promote the further development of AI and the whole science and technology.
Machine Learning: Expert systems, cognitive simulations, programming and problem solving, data mining, network information Services, image recognition, fault diagnosis, natural language understanding, robotics and gaming.
