数学家传记
理查德·贝尔曼是一位美国应用数学家,在动态规划方面取得了重要进展。
理查德·贝尔曼的父亲是弗瑞兹·约翰 James Bellman,母亲是Pearl Saffian。双方家族都来自犹太血统,John Bellman的父亲从俄罗斯移民,Pearl Saffian的父亲从波兰移民。尽管有犹太血统,贝尔曼出生的家庭是不可知论者。
大萧条始于1929年,当时贝尔曼九岁,到1932年,美国四分之一的工人失业。20世纪30年代的大萧条时期工资低下,反犹主义盛行。John Bellman被大萧条毁掉了,但尽管困难重重,他还是设法让贝尔曼获得了良好的教育。贝尔曼十一岁时第一次接触数学,当时他学习了Schultze的Elementary and advanced algebra。他不仅对这次早期的数学接触感到兴奋,而且作为一个男孩,他还喜欢其他活动,如如饥似渴地阅读、逛纽约的博物馆,以及在42街公共图书馆度过快乐的时光。
贝尔曼 在布鲁克林的 马克斯·亚伯拉罕 林肯高中就读,在那里他代表学校参加数学队,并在最后一年取得了全纽约学生中排名第一的成绩。高中毕业后,贝尔曼 于1937年1月进入纽约市立学院。在这个阶段,他已下定决心成为一名理论物理学家,并为此在该学院选修课程。1938年,他从市立学院转到布鲁克林学院,在那里他决定将数学作为主要学习领域。在布鲁克林学院的最后两年里,他代表布鲁克林学院参加了洛厄尔·普特南数学竞赛的三人团队。他于1941年获得数学学士学位,并于同年9月进入巴尔的摩的约翰斯 威廉·霍普金斯 大学攻读研究生课程。
1941年12月7日日本袭击珍珠港的美国舰队后,美国加入了第二次世界大战。到12月11日,美国与德国处于战争状态。贝尔曼 于1942年初离开约翰斯 霍普金斯 大学,前往威斯康星大学担任军事电子学讲师。在作为战争努力的一部分教授电子学的同时,他进行了自己的数学研究,并于1943年获得威斯康星大学的硕士学位。在继续从事战争工作的同时,贝尔曼 接下来去了普林斯顿大学,在那里他在陆军专业训练计划中任教。他得以继续从事数学研究生工作,但在1944年12月,他被征召入伍,并被分配到洛斯阿拉莫斯的曼哈顿计划。在那里,他从事理论物理问题的研究,直到1946年退伍。
贝尔曼立即回到普林斯顿,在所罗门·莱夫谢茨的指导下完成了博士研究。他的学位论文研究微分方程的稳定性,关注实微分方程的解当自变量t趋于无穷时的行为。该论文于当年晚些时候提交给普林斯顿,他获得了博士学位。学位论文的成果出现在他1953年出版的著作Stability theory of differential equations中。一位评论者称赞该书:-
……清晰而吸引人的表述方式。
获得博士学位后,他留在普林斯顿任数学助理教授,但1948年他离开那里,前往斯坦福大学担任数学副教授。在接下来的夏天,他首次在兰德公司工作。他写道[1]:-
1949年夏天我非常渴望去兰德……我与Ed Paxson成了朋友,问他兰德对什么感兴趣。他建议我研究多阶段决策过程。我开始按照这个建议去做。
在兰德公司度过第二段时间后,贝尔曼从斯坦福请假一年,在普林斯顿从事氢弹研究。在那里期间,他开始思考自己应该留在斯坦福——他在那里研究自己最热爱的课题,即解析数论——还是应该在洛杉矶的兰德公司任职。他写道[1]:-
我对动态规划很感兴趣。我很清楚那里有大量出色的分析。此外,我能看到许多应用。这是一个明确的选择。我要么做一个传统知识分子,要么做一个现代知识分子,用自己的研究成果来解决当代社会的问题。这是一条危险的道路。我要么研究太多而应用太少,要么研究太少而应用太多。我有信心能做好这种微妙的平衡,锦上添花。
决定做出后,贝尔曼于1952年离开斯坦福,接受了兰德公司研究数学家的职位。正如Dreyfus在[6]中所写:-
在确信能成功走上传统学术道路的情况下,贝尔曼……却选择投身于后来被称为运筹学的应用数学。在那个年代,应用实践者被视为数学界中明显低人一等的公民。一向喜欢争议的贝尔曼,当受邀在各大学数学系的研讨会上演讲时,乐于为其选择应用数学而非纯数学辩护,称其动机是现实世界更大的挑战和数学需求。
贝尔曼关于动态规划的第一篇论文发表于1952年,他的第一本关于该主题的书An introduction to the theory of dynamic programming由兰德公司于1953年出版。为了了解该主题的内容,我们引用书中研究的一个典型问题。贝尔曼写道:-
我们被告知一个粒子处于状态0或状态1,并且初始给定它处于状态1的概率为x。使用操作A会将该概率降低到ax,其中a是某个小于1的正的常数,而操作L,即观察粒子,将明确告诉我们它处于哪个状态。如果希望在最短时间内将粒子转变为状态0,最优程序是什么?
贝尔曼对这些工具的多样化应用继续解决了一系列问题。正如他所写[1]:-
……大约在1954年,我偶然闯入了一些重要类型的问题,并被迫,不管愿不愿意,去回答一些重要类型的问题。我能在一定程度上处理确定性控制过程,也能处理经济学和运筹学中的随机决策过程。
此时,他在研究方向方面做出了一些非常明确的决定,回过头去审视其数学模型背后的原始假设,并试图引入更符合实际的假设。他接着在1957年引入了马尔可夫决策问题,并在1958年发表了他的第一篇关于随机控制过程的论文,其中他引入了今天所称的贝尔曼方程。
贝尔曼撰写的论文和书籍数量相当惊人。例如,论文[3]列出了贝尔曼撰写(或合著)的621篇论文、41本书和21本书的翻译。尽管在这样的文章中不可能给出比一瞥更多的东西,我们通过列出他几本书的标题来指示其范围。这些包括,除了已经提到的那些:A Survey of the Theory of the Boundedness, Stability, and Asymptotic Behavior of Solutions of Linear and Nonlinear Differential and Difference Equations(1949);A survey of the mathematical theory of time-lag, retarded control, and hereditary processes(1954);Dynamic programming of continuous processes(1954);Dynamic programming(1957);Some aspects of the mathematical theory of control processes(1958);Introduction to matrix analysis(1960);A brief introduction to theta functions(1961);An introduction to inequalities(1961);Adaptive control processes: A guided tour(1961);Inequalities(1961);Applied dynamic programming(1962);Differential-difference equations(1963);Perturbation techniques in mathematics, physics, and engineering(1964);和Dynamic programming and modern control theory(1965)。
你可以在THIS LINK看到更多关于这些书以及后来一些书的信息。
我们在列出贝尔曼的著作时停在了1965年,因为正是在这个时候,他离开了兰德公司,接受了南加州大学数学、电气工程和医学教授的职位。他继续着他卓越的研究和出版记录。他的方法在他1966年在肯塔基大学所做的一系列讲座中得到了总结,他在这些讲座中旨在:-
……描述现代世界的问题如何提供有趣的数学问题,并开辟全新的数学领域。……[我相信]重要数学的发展关键取决于与真实世界的持续互动。
他的兴趣转向将计算机作为数学研究的工具,并撰写了诸如Algorithms, graphs and computers(1970年)等书籍。1973年,悲剧降临,他被诊断出患有脑瘤。切除肿瘤的手术很成功,但术后出现了并发症,使他几乎完全瘫痪。然而,值得注意的是,尽管身体有问题,他在数学研究方面仍然极为活跃,在他生命的最后10年里,他写了大约100篇论文。在这些年间出版的书籍中,我们应该提到特别重要的与G M Wing合著的An introduction to invariant imbedding(1975年)。然而,他还写了Analytic number theory(1980年)、Mathematical methods in medicine(1983年)和The Laplace transform(1984年)。
1984年他去世后,他的书籍继续出版,如Partial differential equations(1985年)、Selective computation(1985年)、Methods in approximation(1986年)和Wave propagation: An invariant imbedding approach(1986年)。
贝尔曼因其对数学应用的杰出贡献而获得了许多荣誉。1970年,他被授予首届诺伯特·维纳应用数学奖,这是美国数学会和工业与应用数学学会的联合奖项。同样在1970年,卡内基梅隆大学授予贝尔曼首届伦纳德·尤金·迪克森奖,三年后,他被生物医学工程学会任命为ALZA杰出讲座。
他于1975年当选为美国艺术与科学院会士,并在随后的几年里获得了冯·诺伊曼理论奖,这是由管理科学研究所和美国运筹学会联合颁发的另一个奖项。1977年,他当选为美国国家工程院院士。他于1979年被授予IEEE荣誉金奖:-
因对决策过程和控制论,特别是动态规划的创建和应用做出的贡献。
贝尔曼于1980年当选为数学生物学学会会士,1983年当选为国家科学院(美国)。1983年,他获得了美国控制委员会颁发的传统奖章。
1979年IEEE颁奖招待会手册写道:-
贝尔曼是现代控制理论和系统分析贡献者中的一位杰出人物。他发明的动态规划标志着大规模系统分析与优化新时代的开始,并为从航天器制导系统设计到害虫控制和网络优化等广泛问题领域中应用复杂的计算机导向技术开辟了道路。
在像我们上面那样列举了他的众多荣誉之后,手册接着对他作了这样的个人致敬:-
但最重要的是,贝尔曼赢得的不只是名声——他以其非凡的勇气和作为人的伟大,赢得了所有认识他的人的钦佩和爱戴。
Richard Bellman's father was John James Bellman and his mother was Pearl Saffian. Both sides of the family came from Jewish descent, with both John Bellman's father having emigrated from Russia and Pearl Saffian's father having emigrated from Poland. Despite the Jewish descent, the family that Richard was born into were agnostics.
The Great Depression began in 1929, when Richard was nine years old, and by 1932 one quarter of the workers in the United States were unemployed. The Depression of the 1930s saw low wages and there was much anti-Semitism. John Bellman was ruined by the Depression but, despite great hardship, managed to see Richard obtain a good education. Richard first met mathematics at the age of eleven when he studied Schultze's Elementary and advanced algebra. He was not only thrilled with this early encounter with mathematics but as a boy he enjoyed other pursuits such as reading avidly, going round the museums of New York, and spending happy hours in the 42nd Street Public Library.
Richard attended Abraham Lincoln High School in Brooklyn where he represented his school on the mathematics team and in his final year was rewarded with achieving the top rank among all New York school pupils. After High School Bellman entered the City College of New York in January 1937. At this stage he had made up his mind to become a theoretical physicist and he took courses at the College with this in mind. In 1938 he moved from City College to Brooklyn College where he now decided to make mathematics his main area of study. He represented Brooklyn College in the three man team in the Lowell Putman mathematics competition in his final two years at Brooklyn College. He graduated with a B.A. in mathematics in 1941 and in September of that year he entered Johns Hopkins University in Baltimore to undertake postgraduate studies.
The United States entered World War II after the Japanese attacked the American fleet at Pearl Harbour on 7 December 1941. By 11 December the United States was at war with Germany. Bellman left Johns Hopkins University early in 1942 to take up a position as Instructor in Military Electronics at the University of Wisconsin. While teaching electronics as part of the war effort, he undertook his own studies in mathematics and was awarded a Master's Degree from Wisconsin in 1943. Continuing to undertake war work with his teaching, Bellman next went to Princeton University where he taught in the Army Specialized Training Program. He was able to continue the undertake graduate work in mathematics but in December 1944 he was drafted into the army and assigned o the Manhattan Project in Los Alamos. There he worked on problems in theoretical physics until his discharge in 1946.
Bellman returned immediately to Princeton where he completed his doctoral studies under Lefschetz's supervision. His doctoral dissertation on the stability of differential equations was concerned with the behaviour of the solutions of real differential equations as the independent variable t tends to infinity. It was submitted to Princeton later that year and he was awarded his Ph.D. Results from his dissertation appeared in the book Stability theory of differential equations which he published in 1953. A reviewer praised its:-
... lucid and attractive manner of presentation.
He remained at Princeton as Assistant Professor of Mathematics after the award of his doctorate but in 1948 he left to take up the position of Associate Professor of Mathematics at Stanford University. During the following summer he first worked at the RAND Corporation. He wrote [1]:-
I was very eager to go to RAND in the summer of 1949 ... I became friendly with Ed Paxson and asked him what RAND was interested in. He suggested that I work on multistage decision processes. I started following that suggestion.
After a second period of time at RAND, Bellman spent a year on leave from Stanford, working at Princeton on H-bomb research. While there he began to ponder whether he should remain at Stanford where he was working on the topic he loved most, namely analytic number theory, or whether he should take up a position at RAND in Los Angeles. He wrote [1]:-
I was intrigued by dynamic programming. It was clear to me that there was a good deal of good analysis there. Furthermore, I could see many applications. It was a clear choice. I could either be a traditional intellectual, or a modern intellectual using the results of my research for the problems of contemporary society. This was a dangerous path. Either I could do too much research and too little application, or too little research and too much application. I had confidence that I could do this delicate activity, pie a la mode.
The decision made, Bellman left Stanford in 1952 and took up the position of Research Mathematician at RAND. As Dreyfus writes in [6]:-
Assured of a successful conventional academic career, Bellman ... cast his lot instead with the kind of applied mathematics later to be known as operations research. In those days applied practitioners were regarded as distinctly second-class citizens of the mathematical fraternity. Always one to enjoy controversy, when invited to speak at various university mathematics department seminars, Bellman delighted in justifying his choice of applied over pure mathematics as being motivated by the real world's greater challenges and mathematical demands.
Bellman's first publication on dynamic programming appeared in 1952 and his first book on the topic An introduction to the theory of dynamic programming was published by the RAND Corporation in 1953. To get an idea of what the topic was about we quote a typical problem studied in the book. Bellman writes:-
We are informed that a particle is in either state 0 or 1, and we are given initially the probability x that it is in state 1. Use of the operation A will reduce this probability to ax, where a is some positive constant less than 1, whereas operation L, which consists in observing the particle, will tell us definitely which state it is in. If it is desired to transform the particle into state 0 in a minimum time, what is the optimal procedure?
Bellman's diverse applications of these tools continued to solve a whole range of problems. As he wrote [1]:-
... as of 1954 or so I had stumbled into some important types of problems and had been pushed, willy-nilly, into answering some significant kinds of questions. I could handle deterministic control processes to some extent and stochastic decision process in economics and operations research as well.
At this point he made some very definite decisions on the direction his research should take, going back at looking at the original assumptions underlying his mathematical model and trying to introduce more realistic assumptions. He went on to introduce Markovian decision problems in 1957 and in 1958 he published his first paper on stochastic control processes where he introduced what is today called the Bellman equation.
The number of papers and books which Bellman wrote is quite amazing. For example the paper [3] list 621 papers, 41 books and 21 translations of books authored (or co-authored) by Bellman. Although it is impossible to give more than a flavour in an article such as this we give an indication of their scope by listing the titles of a few of his books. These include, in addition to those already mentioned: A Survey of the Theory of the Boundedness, Stability, and Asymptotic Behavior of Solutions of Linear and Nonlinear Differential and Difference Equations (1949); A survey of the mathematical theory of time-lag, retarded control, and hereditary processes (1954); Dynamic programming of continuous processes (1954); Dynamic programming (1957); Some aspects of the mathematical theory of control processes (1958); Introduction to matrix analysis (1960); A brief introduction to theta functions (1961); An introduction to inequalities (1961); Adaptive control processes: A guided tour (1961); Inequalities (1961); Applied dynamic programming (1962); Differential-difference equations (1963); Perturbation techniques in mathematics, physics, and engineering (1964); and Dynamic programming and modern control theory (1965).
You can see more about these books as well as some later ones at THIS LINK.
We have stopped at 1965 in giving a list of Bellman's books since it was at this time that he left RAND and accepted an appointment as Professor of Mathematics, Electrical Engineering, and Medicine at the University of Southern California. He continued his remarkable research and publication record. His approach was summed up in a course of lectures he gave at the University of Kentucky in 1966 in which he aimed:-
... to describe some of the ways that the problems of the modern world provide interesting mathematical questions and open up entirely new domains of mathematics. ... [I believe that] the growth of vital mathematics depends crucially on continuing interaction with the real world.
His interests turned towards computers as a tool in mathematical research and he wrote such books as Algorithms, graphs and computers (1970). In 1973 tragedy struck when he was diagnosed with a brain tumour. An operation to remove the tumour was successful but following the operation complications set in which left him an almost total cripple. Remarkably, however, he remained extremely active in his mathematical research despite his physical problems and in the remaining 10 years of his life he wrote around 100 papers. Among the books he published during these years we should mention the particularly important An introduction to invariant imbedding (1975) written jointly with G M Wing. However he also wrote Analytic number theory (1980), Mathematical methods in medicine (1983), and The Laplace transform (1984).
After his death in 1984 his books continued to be published such as Partial differential equations (1985), Selective computation (1985), Methods in approximation (1986), and Wave propagation: An invariant imbedding approach (1986).
Bellman received a great many honours for his outstanding contributions to the applications of mathematics. In 1970 he was awarded the first Norbert Wiener Prize in Applied Mathematics, this being a joint prize of the American Mathematical Society and the Society for Industrial and Applied Mathematics. Also in 1970 Carnegie-Mellon University awarded Bellman the first Dickson Prize and three years later he was appointed to the ALZA Distinguished Lectureship by the Biomedical Engineering Society.
He was elected to Fellowship in the American Academy of Arts and Sciences in 1975 and, in the following years, he received the John von Neumann Theory Award, another joint award this time by the Institute of Management Sciences and the Operations Research Society of America. In 1977 be was elected to the National Academy of Engineering. He was awarded the IEEE Gold Medal of Honor in 1979:-
For contributions to decision processes and control system theory, particularly the creation and application of dynamic programming.
Bellman was elected a Fellow of the Society for Mathematical Biology in 1980 and in 1983 he was elected to the National Academy of Sciences (United States). In 1983 he received the Heritage Medal from the American Council for Control.
The 1979 IEEE Awards Reception Brochure states:-
Richard Bellman is a towering figure among the contributors to modern control theory and systems analysis. His invention of dynamic programming marked the beginning of a new era in the analysis and optimization of large-scale systems and opened a way for the application of sophisticated computer-oriented techniques in a wide variety of problem-areas ranging from the design of guidance systems for space vehicles to pest control and network optimization.
After listing his many honours, as we have above, the brochure then pays him this personal tribute:-
But what matters most is that Richard Bellman won more than fame - he won the admiration and affection of all who know him for his outstanding courage and greatness as a human being.
正文里的方括号编号指向这里,悬停即可直接看到条目。书目保留原文——译了书名反而查不到文献。
原站列出的延伸阅读与外部数据库,照原样保留,目标多为英文页面。
关于理查德·贝尔曼的其它页面:
关于理查德·贝尔曼的其它网站:
原站的交叉引用。指向本站已镜像专题的留在站内,其余仍指回原站。