Machine Learning Books

Exellent study guide for beginner and journeyman Accompanying CD is articulate and the best I have seen.The author has done a fine job,both technically and artistically.

Very competently written text that covers all areas of traditional motor control. The author demonstrates his field knowledge by including typical characteristics of device failure and indications where inspection and possible maintenance would be prudent. Circuits and schematics are shown in standard ladder logic convention and in some instances even shows analogy to digital logic for those who are formerly trained electronics technicians.

My one criticism is the sequencing of some of the chapters. The author seems to have weaved in and out of some areas where I expected more complete coverage only to find the topics further toward the back of the text. This detracts from an otherwise well written and authoritative text for the 1st time and intermediate level learner.

For those who want to know the fundamentals of Automated Industrial Systems under the heading of Electrical Motor Control... This text of ISBN 0-8273-8640-0, I can not recommend. There maybe better text that I am not aware of, But ISBN 0-8269-1663-5 from 1987 is a text you can grow with. I'll be upgrading to ISBN 0-8269-1675-9 by Gary Rockis & Glen Mazur. Sincerely, Industrial Maintenance Electrician of 12+ years. I will post an additional review once I have viewed the upgraded edition. (I've wondered about these reviews... Are they posted by a PR / sales people???) Good Luck & Be Safe.

Well written electronic/electrical books are like finding gold nuggets. I have an electrical background and used this book as a refresher to study for a professional test. I found it was well written, accurate and to the point. The authors did not get bogged down in endless calculations, but did present what was necessary. The authors covered each subject completely with well selected words, illustrations, and pictures that conveyed the idea in a simple to understand format. I would recommend this book to anyone, beginner, or craftsman seeking review of subjects they have not worked with for some time. I liked the short unit concept of covering each subject. I will look to these authors first for other electrical titles in the future.

This book is loaded with informative and valuable essential
information that anyone in the field would find enteresting, and would be of extreme benefit to those entering the field.

I found each chapter well written, easy to understand, and not cluttered with useless information.

The Unit on "Synchronous Automatic Motor Starter" was of particular interest to me as I have operated these machines before.

The section on "Developing Control Circuits" was also exciting for me as I have worked on storm drain pumping stations, and sewage lift stations which utilized the two pump alternating scheme shown in Figs. 2 thru 6.

The solid state theory and circuits was excellent.

This book makes for great reading, and I recommend it for all hands.

Suchman's book is a classic (and about to be updated!), but that doesn't mean there aren't any caveats. Suchman's analysis is deep, her writing thick (incredibly terse, dense prose that may require a good dictionary), and her perspective is still controversial.

This book doesn't tell you how to "do" very much - it's not a step-by-step method book. This is a mix of theory and method that will force the engaged reader to reflect on his/her own work.

This book stands as perhaps the best example of a socio-cognitive analysis of technology, and is therefore correctly treated as fundamental in HCI and related fields. For a researcher who is interested in the relationship between technology and people, or technology and the world, this is a must-read. AI and HCI stumble into each other frequently, but this is a book for both audiences.

As for the debate of plans vs. situated action, well, to some extent I find it irrelevant. Suchman never claims that plans don't exist or are unimportant. Even if your work is completely plan-oriented - say, AI planning (e.g. path planning), you should read this book - it will challenge some of your assumptions, and force you to grapple with problems that exist when technology interacts with the world.

That having been said, this is not an introductory reader on HCI, AI, or any other topic. Suchman's terse language frustrates even some very intelligent grad students and PhD's, and again, this book is deep. It's a book that has challenged me as I've read and re-read it over the years, and I treasure it.

This book is not for everyone. Suchman makes connections between AI, HCI and the sociological areas of ethnomethodology and conversation analysis (EM/CA) - connections that have been very visible and influential in subsequent HCI and CSCW research. If you don't have any background in these sociological areas, it will take some work to read it.

That said, I think this book is reasonably accessible, and certainly more so than has been suggested by some reviewers. Suchman was writing to counter a prevalent mindset in the AI community of the time. Basically, Chapters 2 and 3 set up a technical and philosophical strawman (human action as the execution of plans), Chapters 4 and 5 provide an explanation of some necessary theoretical background, and the rest is an analysis of interaction in the context of these theories that serves to knock down the strawman. It's fairly hard to have a more clear and logical organization than that. There's no part of that organization that could be left out and still have the book make sense.

Furthermore, by comparison, the theoretical parts of this book should be easier for the uninitiated to read than are Garfinkel's writings on ethnomethodology (or most CA writings by almost anyone). They may or may not do justice to those ideas, but that's a separate question. And for someone with any background at all in these areas (though as suggested by other reviewers, this does not include a huge number of people), this book should be a very straightforward read.

The bottom line for me is that this book (like Paul Dourish's "Where the Action Is") is an interdisciplinary gem that has the potential to change how you think about how people approach technology. There aren't that many books for which that can be said.

If you do read it, read only the last chapter and the conclusion.

Summary:
Keep in mind that the title of the book is Plans and Situated Actions: The Problem of Human Machine Communication. The majority of the book is the 'plans and situated actions' part.

The basic idea of the book is that humans don't really function using plans. Plans, as the author defines them, are something akin to diagrams for behavior, explicating specific activities. Instead, the author argues that humans behave based on 'situated actions'. Situated actions are, "the view that every course of action depends in essential ways upon its material and social circumstances. Rather than attempting to abstract action away from its circumstances and represent it as a rational plan, the approach is to study how people use their circumstances to achieve intelligent action." (p. 50).

In other words, people have a goal in mind. To achieve their goal, people may or may not set up a plan (the author discusses how this could be culturally relative, but I think this is a weak point in her argument because she doesn't really do a good job of distinguishing one type of plan from another), but what is important is that in trying to achieve their goal they are placed in situations that determine their actions. This could also be said: people behave in specific situations based upon the factors that affect the situation.

Let me give an example... Let's say your goal is to get to the dentist. You set up a 'plan' for getting to the dentist prior to leaving. Your plan would include a calculation of the time and the route and your mode of transportation. The situated action approach would say that you can only understand the individual's behavior in terms of their actions in specific situations. So you get in your car and on the way to the dentist's office you run into a detour due to construction. If you had to follow your plan, you couldn't make it to the dentist. But when you leave the road and find an alternate route, this behavior is only understood in terms of situated action. Does that explain it? Wow, and it only took me a few paragraphs.

The author discusses plans and situated actions in terms of conversations, cognitive science, ethnomethodology, and a whole bunch of other theoretical perspectives and technical jargon. In the end she finally gets to the human and machine communication. This is also where the book begins to get interesting. She studied how people interacted with copy machines that were trying to give people instructions. Her studies, undoubtedly helped the people at Xerox figure out ways to improve their copy machines and instructions for them. Like I said above, the last chapter and the conclusion are the most interesting parts of the book. Skip the rest and read them.

My Comments:
For someone so concerned with understanding how people communicate this book is horribly written and nearly unintelligible. The first six chapters are theory and examples of the theory that are completely unrelated to machines. The book finally gets to human and machine interaction after nearly one hundred pages of inchoate theory. And the human and machine interaction stuff isn't really all that interesting - especially since it predates the 1990s, is talking about interaction with copying machines, and has nothing to do with computers.

The author should have chosen a specific approach and then stuck to it. Perhaps she could have tripled the length of the book and gave clear and understandable explanations of the theories (though I am pretty much convinced after having read the book that this would be impossible because of the author's writing style) and used examples that applied only to human and machine interaction. Or she could have just jumped into her findings that dealt with human and machine interaction. The first approach could have been 'dumbed down' to make the book readable by the general public. The second approach could have served a more academic market.

The book reads something like a doctoral dissertation (it very well may be one, I don't know) in that she gives some information on each theory, but not really enough to give someone a good understanding of it - something like a literature review - and cites examples of research that are completely unrelated to the topic of the book to illustrate the theories . The she presents her methods, results, and conclusion.

I guess my problem is that I was expecting a book that would actually be enjoyable to read, interesting, and would focus on human and machine communication. If that is what you are looking for, look somewhere else. This book is nearly impossible to understand. I read the book for a graduate level course in Ethnomethodology and I didn't really understand it very well. By no means am I an expert in Ethnomethodology, but I'm pretty sure I know more about it than probably 95% of the world's population (keep in mind I don't know very much at all), so I'm pretty confident most people would find this book nearly impossible to decipher.

This is an outstanding book. The insight that showed the power of the idea of `situated action' goes far beyond the realm of interactive design or even human computer interaction in its entirety. It is a fundamental solution to the problem of facing complexity and contingency. Its implications are widespread. This book was published in the 1987 when during the last days of classical AI. This is one of the seminal books that showed the inadequacies of the classical formulation. Indeed it showed a new and much more way of achieving the goals that classical AI set for itself and failed. Despite its age the ides in this book are still fresh and important.

. Absolute certainty is impossible and the quest for it is costly and futile. Instead of trying to overcome the uncertainty that is in the world, the system designer should embrace it and use it as a tool to solve the problems that it creates.

This is a book that should be read by anyone who has set the task for themselves of developing any system that must function in an uncertain environment. In short this is a book that should be read by anyone who is developing a system that will have to function within the real world

This is THE book to start with if anybody is interested in studying interaction design. In a time everybody calls themselves an interaction designer, it's a highly recommended reading to learn there's more to interaction than simply large colourful buttons... Based on an ethnomethodological perspective, Suchman does a brilliant job in analysing users' interactions with an advanced Xerox machine, and putting forth an interesting critique of classical AI concepts. It's highly recommended for anybody interested in Human-Computer Communication and interaction design.

provide a good and easy to understand introduction to the subject

This book gives a good introduction to the mathematical modeling of cognition and does so with a level of mathematics that is very accessible to a typical graduate student in computer science or psychology. The book could have been written using tools from measure theory but luckily it was not for a book at an introductory level. The concept of probably approximately correct is introduced early on in the third chapter of the book with efficient learning given later on in Chapter 5. Chapter 7, the best chapter of the book, discusses the idea of VC dimension, which has had many applications, such as network stability and optimization. VC dimension plays the pre-dominant theme in the rest of the book, with the book ending with an application to neural networks. There are short problem sets at the end of the chapters, and these are useful for more understanding of the concepts in the book. A very interesting book and worth the price.

This book is provides a good overview to the theory behind MDL. Not for the faint of heart, however.

This book provides an excellent, in-depth summary of current MDL theory. It is arguably the best text on the topic of MDL currently written, and is one of the best expositions of a general approach to statistical inference more broadly.

One of the greatest strengths of the text is its ability to start from basics of probability and work toward profound insights into statistical inference, without sacrificing important mathematical detail. Grunwald does a good job of explaining the MDL paradigm and how it is related to more traditional paradigms--especially the Bayesian paradigm--while acknowledging current limitations in our understanding of MDL inference.

This text has much to offer those already familiar with MDL, however, and serves as much more than an introduction to the topic. Grunwald provides an excellent summary of important but lesser-known information-theoretic results that are critical to MDL and statistical inference more broadly. He also provides an excellent summary of current problems facing the field, which can be helpful even among those who are already aware of them.

Although excellent, the text is not perfect of course. Grunwald's tone is sometimes overly antagonistic towards critics, and he has a tendency to overstate the seriousness of relatively trivial problems, some of which apply to any inference paradigm. I also would have preferred more discussion of relationships between MDL and the frequentist paradigm, as there is much there to explore, especially in the domain of exact methods.

The text can be mathematically intense at times, but not inappropriately so, being a book on a major theory of statistical inference. I was actually impressed at the extent to which major theorems were gradually built up from extraordinarily basic mathematical principles.

In general, I would highly recommend this text to anyone interested in general principles of statistical inference or information theory. Although the book is on a particular approach to statistical inference, using a particular branch of information theory (i.e., optimal inference is based on choosing models that result in the most parsimonious information-theoretic description of the data and model), it has implications and utility in a number of fields.

It may come as a surprise to some, but computer games are a serious study of academic study, as showng by this book. Here, the games are strategy games, as opposed to the twitch games like Quake or Doom.

The book accurately reflects trends in our understanding of how to program strategies for the various games. Chess may be falling off simply because current chess algorithms are quite good. Ever since 97, when Deep Blue defeated world champion Gary Gasparov. Now, a chess program that plays at grandmaster level elicits little surprise. So it has been a good few decades for chess research, but diminishing returns in understanding may be setting in.

Whereas the book has articles on other games like Go, which are far harder to a machine to devise strong winning methods. A puzzling aspect about the book are the articles on draughts (checkers). It's a relatively simple game. Less depth than chess. While everything might not be known about machine strategies for draughts, how interesting are these?

Perhaps a future version of this book might have articles on various war games. These use far larger boards than Go, and have terrain information at each square or hex, and many different types of pieces.

Just in case you thought the field was getting fully known.

This is a collection of papers presented at a NIPS workshop held in 1997. So it provides a good entry point for access to forefronts of this rapidly developing field. Many leading researchers have contributed to this volume including V. vapnik who wrote a very succinct and readable survey. The introduction (Chapter 1) is also very useful. Though all chapters are written by leading experts in their areas and are enjoy to read. Personally I like particularly Part II on implementation in large data sets. G. Wahba provides some background on RKHS theory and a statistical perspective from GACV, for which she is mainly responsible for its popularity in statistics. I recommend this book for researchers and practitioners who may want more details and update recent developments.

This book overviews the status of research in automated reasoning as it stood in 1997, but with emphasis on the work of Larry Wos, and one year after the announcement of the proof of the Robbins conjecture by the EQP reasoning machine. The book is still worth perusing, even though it is seven years out of date, and there has been a considerable amount of work done in the field of automated reasoning since 1997. Much of this work has involved dealing with the informal nature of mathematical proofs as they are currently done in most of the mathematical research literature, and thus involves research both in natural language processing and in automated reasoning. The applications of automated reasoning have to this date involved deductive databases and formal checking of computer systems and circuit designs. The field of automated reasoning is of course of great importance to those involved in the construction of intelligent machines. Throughout this book proof systems such as OTTER (for Organized Techniques for Theorem proving and Effective Research) are referred to quite frequently. OTTER is now one of the most popular systems for the exploration of automated reasoning, and is publicly available. Due to constraints of space, only selected articles of the book will be reviewed here.

The notion of linked resolution is discussed in the second article of the book, which is a strategy used to deal with "uninteresting" clauses in resolution-based provers. Linked resolution will perform proof steps until an "interesting" clause is derived, which it will retain and then eliminate the intermediate results. It does this by identifying a collection of clauses called `linking clauses', which are used as a nucleus only in certain steps, like a hyperresolution clash. The author of the article proves a completeness theorem for predicate logic without equality that incorporates linked resolution, and considers this result a generalization of the completeness of hyperresolution. He emphasizes though that adding equality and paramodulation will eliminate the completeness result.

The third article in the book discusses Isabelle, which is a theorem prover that is based upon resolution, and searches for proofs using a tableau approach. It is, according to the author of the article, a `generic' reasoner, in that it can work in a variety of domains without reducing them to first-order logic. Isabelle is also `interactive' meaning that its user can direct each step of the proof. Interactive provers are to be distinguished from the `resolution' systems such as OTTER, but Isabelle is also based upon a form of resolution, and therefore represents a "synthesis" between the these two traditions in automated reasoning. The author thus describes Isabelle as an interactive prover based on the typed lambda-calculus, and uses as its primary inference rule a generalization of Horn clause resolution. The author outlines in detail the tableau approach, and uses Isabelle to prove classical results in set theory and first-order logic and proves the Church-Rosser theorem for combinators. When reading the article one sees immediately the connection of Isabelle with higher-order logic, it being a `meta-logic' that supports `object-logics' (the meta-logic supporting connectives, implication, and equality), making use of functions called `tacticals' that operate on proof states called `tactics', with concepts from logic programming used to support automation on the tactics and tacticals.

The fifth article in the book, written by William McCune, who is responsible for the use of EQP and its consequent proof of the Robbins conjecture, is an overview of how to evaluate the `paramodulation' strategies using EQP. A test collection of 33 equational theorems is used for this evaluation, and the performance of OTTER and EQP are compared on this collection. The proof of the Robbins conjecture, because of the timing in publishing this book, is not included, as it was solved by EQP, as the author indicates, after this article was finished. The author though describes in some detail the Robbins algebra, and in a footnote briefly discusses how the Robbins conjecture was proven.

The most interesting of the articles in the book is the last one, which concerns what the author of article calls "metalevel reasoning", and its role in controlling automated reasoning programs. He defines metalevel reasoning as being the reasoning with facts and knowledge beyond the symbols of the formula given to the basic reasoning system. This notion encapsulates perfectly the process by which human mathematicians have generated proofs for the last few centuries. Unlike the case of the formal and automated reasoning systems that have been developed over the last five decades, human mathematicians use ordinary language and extraneous knowledge that is typically outside of the scope or domain of the problem they are analyzing. Automated reasoning systems developed up until now do not do this, says the author, but instead manipulate the symbols of formulas according to a collection of inference rules. A reasoning system that incorporates metareasoning would be an enormous benefit argues the author, and could provide the "proving" expertise that is now introduced by fiat by the developer or user of the reasoning system. He gives a few examples of metalevel reasoning, and gives an overview of just how a reasoning system that incorporated it would behave. The author also discusses how metalevel control has been studied in the context of general research in artificial intelligence. The research that he describes in this article is now occupying the time of many researchers in automated reasoning and automated theorem proving, and if it comes to fruition will represent a major advance in machine intelligence. For a machine to be able to generate and prove theorems in the manner that human mathematicians are able to, would be an example of machine creativity and has enormous potential in real-world applications such as bioinformatics, automated scientific discovery, network management, and financial engineering. It is for this reason that this reviewer regards this type of research into automated reasoning as being the most important of all activities to date.

It is a book that helps our students of doctorate.

There seem to be more and more quality books coming to the market as of late. I recently received a complimetary copy of this book and was pleasantly surprised by its' accuracy, layout, and compact format. Mr. Mattson deserves credit for including very accurate programs and realistic fixturing and manufacturing processes.

This book is the best value for the CNC student's money today.

it was my term homework which i prefer then the other units.And i m very glad for choseing this book cause it s out of boring classic managment units. It gives me diffrent management dimension. sorry about my english