Scribd
Upload
894 views

Jaynes - Prob Theory Lectures

Ed Jaynes - Lectures at the University of Washington on Probability Theory and Bayesian Inference.

Uploaded by

Michael Roberts
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF or read online on Scribd
894 views

Jaynes - Prob Theory Lectures

Ed Jaynes - Lectures at the University of Washington on Probability Theory and Bayesian Inference.

Uploaded by

Michael Roberts
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 446
PROBABILITY TEKORY with Applications in Science and Engineering A Series of Informal Lectures by : BE. T. Jaynes Professor of Physics Washington University St. Louis, Missouri 63130 Fregmentary Edition of February, 1974 *The following notes represent what is completed to date of a projected ook manuscript. All Lectures after No. 9 are incomplete; Lectures 11 and 12 are missing entirely, although their content is already largely published in E. T. Jaynes, "Prior Probabilities," IEEE Trans, Syst. Sci. and Cybern. SSC-4, Sept. 1968, pp. 227-241; and "The Well-Posed Problem," in Foundations of Physics, 3, 477 (1973). The projected work will contain epproximately 30 Lectures; in the meantime, comments are solicited on the present material. SUMMARY OF BASIC RULES AND NOTATION Deductive Logic (Boolean Algebra): Denote propositions by Rod, eter, their denials by as oA is false,” etee Define the jogical product and logical sum by AB = "Both A and 8 are true.” A+B = "At least one of the propositions, A, B is true.” Deductive reasoning then consists of applying relations such as AA = A; A(BHC) = AB + AC; AD+a = ab+B; if D = ab, then d= A+B, etc., in which the = sign denotes equal "truth value Inductive Logic (probability theory): This is an extension oF deductive Togie, describing the reasoning of an idealized being (our robot"), who represents degrees of plausibility by real numbers: (A|B) = probability of A, given B. Elementary requirements of common sense and consistency, such as: (a) if a conclusion can be reasoned out in more than one way, every possible way must lead to the same result; and (>) in two problems where the robot has the same state of knowledge, he must assign the same probabilities, then uniquely determine these basic rules of reasoning (Lect. 3) Rule 1: (A8{C) = (A|BC)(B|C) = (BIAC)(A|C) Rule (A[B) + (a|B) = Rule (AtB|C) = (Alc) + (BIC) - (AB|C Rule 4: If {Ay.,.An} are mutually exclusive and exhaustive, and information B is indifferent to them; i.e., if B gives no preference to one over any other, then (A, [B) = V/n, i= 1, 2, 0000 From Rule 1 we obtain Bayes’ theorem (alge) = cafe) {Blac Corolla From Rule 3, if {Ay...A,} are mutually exclusive, (Ay + oe. + AL [BD = Byte) If in addition the A, are exhaustive, we obtain the chain rule: E (alageytay |e) (ale) = Fa, |o) = isl These are the relations most often used in practical calculations. (continued on inside back cover) PREFACE This book has grown over several years from a nucleus consisting of transcripts of tape recordings of a series of lectures given at the Pield Research Laboratories of the Socony~Mobil Oil Company in Dallas, Texas, during March, 1958 and dune, 1963. ‘he lectures were given also, with gzadually increasing content, at Stanford University in 2958, at the University of Minnesota in 1959, at the University of California, Los Angeles in 1960 and 1961, at Purdue University in 1962, at Dartmouth College in 1962 and 1963, at the Standard Oil Company Research Laboratories, Tulsa, in 1963, at the University of Colorado in 1964, at the University of Maryland in 1968, and at Washington University in 1966, 1969, 1970 and 1972. The material of lectures 1-10 and 16-1? was issued by the Socony-Mobil 641 Company as Number 4 in their series, "Colloquium Lectures in Pure and applied Science", and is reproduced here, with permission, in considerably expanded forn, In editing and adding new material, the informal style of the original presentation has been retained. ‘his and the general format axe intended to emphasize that the book is in no sense a textbook or complete treatise, but only a series of informal conversations (necessarily rather one- sided), concerning the foundations of probability theory and how to use it for current applications in physics, chemistry, and engineering. The speaker is simply sharing his views with the audience, and trying to give some more or less convincing arguments in support of them, often, the trend of a lecture was determined by questions raised from the audience. The material is addressed primarily to scientists and engineers who ave already familiar with applied mathematics and perhaps with certain special uses of probability theory, such as statistical mechanics, communication theory, or data analysis, but who may not have had the time to make an ex- tensive study of modern statistics. Such persons may be appalled, as I was when I commenced serious study of the field in 1950, by the enormous volume of literature dealing with statistical problems, and may despair of ever mastering it--not because it is too advanced, but simply because the field is too large. There is so much diverse and intricate detail that it is almost impossible to locate the underlying

You might also like