Algorithms Books

This is a book that is hard to pinpoint who it was written for.
On very few pages, it tries to cover a huge variety of topics, from modelling and animation to rendering.
Unfortunately, the results are usually not very successful, since sometimes important areas of research are not even mentioned and others, the solution provided by the author is not a great one at all.
The book can be best seen as a tutorial covering the algorithms used in R. Stuart Fergusson's now freely available OpenFX (with source code).
Albeit the title implies that this is a book aimed at experienced people in computer graphics, don't be deceived. This is a book for newbies. However as such it is hard to recommend it. At best, this book can help students understand the complications involved in going from a theoretical algorithm to an actual implementations of it. And the code provided should be seen by students as an example of how NOT to code for CG graphics: a lot of it in spaghetti C.
The text is clear and the first pages covering most of the basic concepts of 3D graphics is a good introduction. However, even then, some important concepts of CG are hardly ever mentioned, like NURBs curves and surfaces.
Albeit each chapter covers a specific area, the book is disorganized jumping from modeling to rendering and back again for no reason.
Pages are devoted to the most trivial storage of polygons in memory, and the author then provides a very inefficient way to store polygon connectivity for subdivision (never mentioning that edge structures such as half-edge or quad-edges have been known for quiet some time and are way more practical than his ad-hoc methods).
Following are chapters devoted to scanline rendering and raytracing. The information provided is also simple and mostly just a description of the techniques implemented in his 3D package.
Then jumping to animation... keyframing is described briefly and mostly in terms of the very bad TCB spline approach. Hierarchies and bones come next, with a decent introduction to IK systems. Physical simulations are discussed later, but so badly that it is not worth the few pages there are.
The Polygonal modelling chapter is okay, given the basics of several common algorithms that anyone dealing with polys will need all the time.
The coverage of image processing is laughable.
And then comes a chapter devoted to procedural textures. This is probably the best thing in the book, assuming you already read "Texturing and Modelling: A procedural approach". Besides covering the obvious textures based around Perlin noise, the author goes beyond that by giving an introduction to the more powerful crystal-based textures, which afaik have not been well documented in the past and are usually not described in most courses, since they are impossible to do just with Prman's SL. The explanation and theory is not amazing but it does give an introduction that can help understand his code and is indeed more practical than the famous "T&M" book, which these days seems pretty outdated. As is the case in the rest of the text, the author gives you a little peek at his code and structures he has used, which students that have a solid coding experience may find helpful. Unfortunately, these snippets of structs are really pretty bad and I would not want to encourage anyone to follow those constructions, beyond a learning guide.
The final chapters are devoted to Win32 specifics, which are likely already out of date: some DirectX code and a Windows player.
Overall this will be a book useful for only for someone that has just started computer graphics or maybe a shader writer interested in procedurals. The explanations are certainly more accessible than other more popular texts.
At the same time, it would be unfortunate if this book was their only reference.

In Practical Algorithms For 3D Computer Graphics, computer animation expert Stuart Ferguson has compiled a superbly presented compendium of tools for creating a complete suite of programs for 3D computer animation, modeling, and image synthesis. After a comprehensive introduction to basic principles and key concepts regarding 3D computer graphics, Practical Algorithms For 3D Computer Graphics goes on to provide useful, practical algorithms (and code) for plugin and game engine developers. Featured are real-time 3D graphics for Windows including sample programs in OpenGI and Direct 3D, as well as an accompanying CD-ROM with sample programs. Practical Algorithms For 3D Computer Graphics is a superb introduction to, and continuingly useful reference work for, anyone involved with the development of computer games, movie animation, and graphics programming.

The book is pretty good, and I got a lot out of it. It contains lots of interesting ideas about pruning. I suppose I got out of it what I wanted to get out of it. The reason I don't rate the book higher is because most of the book was pretty worthless. Nearly 50 pages of the book is devoted to the move lists of all the games they played! I don't need to see 50 pages worth of computer chess game moves. That's something that could be better served from a website. Sure it's probably needed for the thesis to verify that they actually did the studies they say they did, but I don't think it belongs in the book. Also, a large portion of the book is devoted to end game databases and recognizing when to look in them and optimal ways to index, etc. I wasn't very interested in that. My last complaint is that there isn't really any source code included with the book. All of the snippets of code that are shown are greatly cut up to the point where they're completely illegal. This book would be a 5 start book if they had actual pieces of code - or even the whole source - instead of the stupid moves. Code like if(try_null) { // do null move here } is the kind of code you see. He should have just printed the full alpha-beta algorithms. He tries to describe them as clearly as possible, using as little as possible real code. Like he wants to explain it all, yet keep it a secret! Putting in the code would be self-descriptive.

This book will provide the most benefit to someone who has already written a computer chess program and is looking to improve it. The chapters on extended forward-pruning were especially informative and a number of high-level programs currently use the adaptive null-move pruning presented here.

I also found the chapter on interior-node recognizers unique and eye opening. A number of clever enhancements are provided that coupled with chapter 6, on knowledgable tablebase encoding, are sure to provide a idea or two even to hardened computer programmers.

This book is not for the faint of heart, but if you spend the time delving through it, you will more than likely find a number of ideas to apply to your chess engine. It is written with an academic perspective and is rigourous in the ideas that it presents (unlike Levy and Monroe's books which are geared more towards the general public).

This book is unique in that it is both low-priced and provides an overview of four popular languages - C, C++, Java and VisualBasic (I couldn't find anything on Pascal and Fortran, unlike what the Publisher's review implies). It is sometimes heavy going and or lacks detail in places, and will need to be supplemented by other books and courses. Still, it's good value for the price.

This is a pretty good introduction to basic undergraduate ideas in computer science including different number systems (hex, binary), computing structures (components, architecture), and the concept of program layout and algorithms. This book does not go deep into these concepts, but at least it makes the student aware that these issues exist and gives a very brief idea of how they work. In other words, don't expect any in-depth examination of computational complexity in the section on algorithms. The lion's share of the book is dedicated to simple programming concepts. First the ideas of variables, constants, operators, operator precedence, and input/output are introduced. Then these concepts are used to examine the constructs of control structures, functions, subroutines, arrays, strings, files, and data structures. There is a very basic chapter on object oriented programming that includes the advantages of object orientation as well as an introduction to terminology and program structure. Simple programs are shown as examples and also given as exercises in four popular languages - C, C++, Visual Basic, and Java.

This book was written in 1999, so the properties of the four featured programming languages have changed some, but not enough to nullify the usefulness of this book. I would say that this book is suitable for high school and freshman college students getting their first look at the basics of computing and of programming in any language. This book does not substitute as an introductory textbook for learning the four programming languages used, but it would make a nice supplement.

When I was a graduate student at Stanford I labored with tedious algebraic computations while trying to prove limit theorems for maxima of stationary sequences. Often it would be very frustrating as I would spend hours checking calculations only to repeatedly make minor mistakes that would cause results not to match. When properly programmed, this is ideal work for a computer. The computer can do the algebra flawlessly and much more quickly freeing the researcher to think about the harder problems. Mathematica was just coming out in the late 1970s but it was too late for my use. Now symbolic computing has advanced to the stage where we all should be using it. Andrews and Stafford make that point in this book. They cover many of the important mathematical results that are needed to solved advanced statistical problems. Edgeworth expansions, moment calculations and Taylor series expansions can all involve a great deal of tedious algebra. All these methods and their applications are exposed in this marvelous book. The authors also introduce us to an analytic bootstrap, Bartlett corrections, saddlepoint approximations, Cornish-Fisher expansions and Edgeworth expansions and applications to survey sampling designs. Chapter 3 goes into the fundamentals of symbolic computing and is the foundation needed for the applications that follow. This is a unique book that is valuable to any research statistician who may need to develop asymptotic results.

When I was a graduate student at Stanford I labored with tedious algebraic computations while trying to prove limit theorems for maxima of stationary sequences. Often it would be very frustrating as I would spend hours checking calculations only to repeatedly make minor mistakes that would cause results not to match. When properly programmed, this is ideal work for a computer. The computer can do the algebra flawlessly and much more quickly freeing the researcher to think about the harder problems. Mathematica was just coming out in the late 1970s but it was too late for my use. Now symbolic computing has advanced to the stage where we all should be using it. Andrews and Stafford make that point in this book. They cover many of the important mathematical results that are needed to solved advanced statistical problems. Edgeworth expansions, moment calculations and Taylor series expansions can all involve a great deal of tedious algebra. All these methods and their applications are exposed in this marvelous book. The authors also introduce us to an analytic bootstrap, Bartlett corrections, saddlepoint approximations, Cornish-Fisher expansions and Edgeworth expansions and applications to survey sampling designs. Chapter 3 goes into the fundamentals of symbolic computing and is the foundation needed for the applications that follow. This is a unique book that is valuable to any research statistician who may need to develop asymptotic results.

This is the first book that I am aware of that addresses the foundations of evolutionary and genetic algorithms, evolution strategies, and evolutionary programming from a rigorous mathematical point of view. The book is designed for an audience of mathematicians and computer scientists who are curious about evolutionary strategies and need a formal treatment of its foundations. Readers currently involved in designing and writing genetic programs will find this book helpful in the optimizing of their algorithms, even though at times they might find the presentation a little heavy-handed.

Evolutionary strategies are thought of as dynamical systems in the book, but these are not in general deterministic, but probabilistic in nature. The state space of the dynamical system consists of the direct product of an object parameter space, an endogenous strategy parameter set, and a collection of fitness functions. Evolution takes place in this state space via the "genetic operators", i.e. the selection, mutation, reproduction, and recombination operators. The goal of course is to find an optimum solution to the problem, and so a consideration of the convergence of the evolution strategy to this optimum must be addressed.

These issues and others, such as the differentiation between evolutionary strategies and ordinary Monte Carlo methods, are discussed in great detail in the book. The author emphasizes that the mechanism of evolutionary strategies lies in the local properties of state space, the evolutionary process being obtained by small steps in this space. He also suggests three prerequisites for the working of evolutionary algorithms, namely the evolutionary progress principle, the genetic repair hypothesis, and mutation-induced operation by recombination. The first is the statement that each change of the individuals in the state space can result in fitness gain as well as fitness loss. The second is a device employed for statistical estimation, and attempts to answer why recombinant evolution strategies are better than nonrecombinant strategies. The third is the statement that dominant recombination causes cohesion of a population and is represented by a local operator which transforms the mutations by a random sampling process.

The author makes use of differential geometry in the book to establish a theoretical framework to predict the local performance of evolution strategies. The hypersurface model is constructed as a fitness model for the calculation of progress measures, and for an elementary model of evolution dynamics. Tensor calculus is employed to study deformations of the sphere model, with the goal of obtaining useful formulae for the progress rate. A mean radius of this deformation is calculated, to serve as a substitute radius in the progress rate formulae for the sphere model.

For the case of (1+1)-selection, i.e. one parent and one offspring, where both parents and offspring are contained in the selection pool, the author derives exact integral representations for the progress rate. The quality gain for one parent and any member of offspring is also considered, and the author derives an integral expression for it using an approximation of the distribution function of the mutation-induced fitness distribution. He argues that the progress rate and the quality gain are progress measures that describe totally different aspects of the performance of evolution strategies.

The general problem of an evolution strategy with arbitrary numbers of parents and offspring is also considered. Since the distribution of parents in the parameter space is unknow, and since it changes in successive generations, this makes the analysis of the progress rate extremely difficult. The author does however derive the relations for this model in terms of a formal expression for the progress rate which is given as an integral over the distribution of a single descendant, which is generation-dependent and unknown. This distribution is approximated using Hermite polynomials and the determination of this function is then reduced to the finding of a collection of coefficients. These coefficients are functions of moments of the offspring and are estimated by the random selection process of the evolution strategy.

Recombinative evolution strategies are also studied by the author, and two special recombination types considered, namely the intermediate and dominant cases. Intermediate recombination is shown to lead to higher performance compared to nonrecombinativie strategies. The dominant case is shown to lead to mutation-induced speciation by recombination.

The author also analyzes the dynamic adaptation of the mutation strength to the local topology of the fitness landscape. Self-adaptation, which is the method for applying evolution to the adjustment of optimal strategy parameter values, is given detailed treatment for the case of one parent in terms of mean value dynamics.

This monograph is a detailed treatment of a strain of evolutionary computing called "evolution strategies" (ES), which comes out of Germany and follows from the work of Ingo Rechenberg, Hans-Paul Schwefel, Günter Rudolph, Beyer, and a few others. It is distinct from Goldberg's genetic algorithms (GA), Fogel's evolutionary programming (EP), Koza's genetic programming (GP), and simulated annealing.

It is quite a dense book, making heavy use of differential geometry. You can find a more brief treatment of ES along with a comparison to EP and GA in Bäck's "Evolutionary Algorithms in Theory and Practice".

I believe this book is well intended and cover a lot of useful information; but I have to say, sometimes itfs more like a puzzle than a book. There is a lot to learn from this book, but you better prepare yourself to spend a lot of time deciphering it.

This book contains the most complete and comprehensive topics on trellises and trellis-based decoding algorithms, the material is presented in a simple and unified form, and it is explained in an easily understood manner. It is the most readable coding book.

This book includes the latest research results as well as the material which is essential and useful for practical applicaitons. It is a very good text book for a course on specail topics in coding or a supplemental text book for a course on error correcting codes, and it is a very informative coding book for communication systems engineers.

Being one of the authors of the book I have the following observations: - the book is the first monograph presenting the ABS (Abaffy-Broyden-Spedicato) methods, that unify the field of algorithms for linear algebraic equations - the book can still be considered a fine introduction to the ABS field; it was translated also in Russian and Chinese; it was based upon about 60 papers on ABS methods; now, end of 2000, there are close to 400, many of them now applied to optimization; there is a new monograph by Zhang, Xia and Feng, of Dalian University of Technology, that presents the new developments in optimization (only in Chinese; English translation under way) - ABS methods have led to the most efficient general linear solver (the implicit LX algorithm), to a more efficient reformulation of the linear algebra for the LP problem in standard form, to a new solution algorithm for linear Diophantine systems, implying first generalization of the Diophantus-Euler theorem from a single equation to a system - ABSPACK, a package for linear and nonlinear systems based upon ABS methods, is now being developed

Topics are carefully selected. Good coverage of both traditional and cutting-edge new theories and algorithms. Less errors than a normal technique book. Only key derivations are outlined with others just omitted. This makes it easy to follow for an engineer like me but i suppose an academic researcher may dislike it somewhat. Unfortunately, too few results are given for each algorithm. Readers don't know if the algorithms have the potential to meet the various requirements of EC, i.e., residue error, convergence rate, complexity, numerical stability. (By the way, even a simple touch of industry standard of AEC might be helpful for this book) Anyway, for anyone who is interested in either accademic research or implementation of EC, i guess this book gives him/her a solid starting point.

Interesting mixture of topics. It touches on audio systems, psychoacoustics, 3D sound, adaptive filters, active sound control, multi-resolution spectral analysis. The most important of all that it also includes practical guidelines for implementation of those techniques.

Congratulations to the authors! It is a really good book on Adaptive Resonance Theory (ART) Neural Networks. It worth it each dollar you pay for and might be read by every student or professional interested in applications aimed to datamining with unsupervised ANN. Well written and very practical, this book presents all the knowledge the computing professional needs to bring the ART models down to Earth and enjoy a lot with examples of applications on MATLAB code. After almost 1.000 trials on ART1 and Fuzzy-ART codes presented by the authors, the euclidian space of the resulting domain and image parameters comparison which I came to was the same one presented in the book, proving the solid author's background and the code's value. In fact, a lot of MATLAB background is required to test the codes (unfortunatelly, the Fuzzy implementation performs two dimensions input vectors only). The book do not get a "5" degree, in my opinion, because it do not present ART2 algorithm. I have made some modifications on the original codes (ART1 and Fuzzy-ART) to run input vectors with unlimited dimensions, which proved its value when tested with a lot of data (in fact, I am using the codes for my IT Master thesis).