hobbit@SHUM.HUJI.AC.IL (yoav gonen) (05/07/91)
Gregory Rawlins wrote, refering to Manber's book (Introduction to Algorithms): > Why the hysteria? Looking at the list of point you give leads me to > suspect that you don't want a textbook, but a handbook or reference book. > Manber's book is an attempt to teach students to think for themselves. I was looking for a book which will help me will the course : Algorithms that I learn in the university. So I bought Manber's. Unfortunately, the book covers only 25% of the material that I learned in class. Diane Musicant wrote: > Manber's book was extremely helpful for me! I was> > told by one of my profs that the book was written for just such a > purpose - to make the learning less intensive. He highly recommended it, > and now, so do I. That was my problem too: my lecturer in the course alse recommended that book. But never mind. I want to show you several examples from the book itself: 1. Page 242: It says: "If the capacities of all edges in the network are INTEGERS, then there is a maximum-flow whose value is an integer..." Now, a man who reads that will soon ask himself: So, what about non-integers? Why can't it work then??? So here is challenge to all of you: Try to find an example of a network, with edges that are NOT integers - that whatever you'll do: there will always be an augmenting path. It is very hard to find such an example. That should have been Manber's job. HE should have shown us the example. But he didn't. 2. Page 189: Manber talks about the problem of Eulerian Circuit. Instead of showing us the complexity of his algorithm (which is O(n^3) - but this is not so trivial to prove it) he goes straightly to the next problem. 3. Page 307: "This algorithm is known as the FOUR-RUSSIANS-ALGORITHM." This is not true. The complexity of the algorithm which is described in the book is O(n^3/logn), while the F-R-A take only O(n^3/(logn)^2). 4. Page 309: The fast fourier transform. Although FTF is the best way to solve the problem of p(x)*q(x), a polynomial product can easily be solved by an algorithm of O(n^1.5) without needing FTF. FTF is used by more complicate problems. ________________________________________ | Yoav Gonen | | The Hebrew University, Givat Ram | | Jerusalem, Israel. | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
gln@cs.arizona.edu (GaRY NEweLl) (05/08/91)
In article <hobbit.673609163@shum>, hobbit@SHUM.HUJI.AC.IL (yoav gonen) writes: > I was looking for a book which will help me will the course : Algorithms that >I learn in the university. So I bought Manber's. Unfortunately, the book covers > only 25% of the material that I learned in class. Strange class you took then. Clearly the book covers much more than 25% of the material that I have seen in any algorithms text. It seems clear to me now that you simply do not understand the purpose of the text - it was not designed to replicate Sedgewick or Aho - it was designed (I believe) to fill a gap in the area of creative design to algorithms. The text teaches algorithm design through the use of induction and attempts to repeatedly show how problems can be solved in this way. It sounds like you are simply attempting to design a course around the presentation of existing algorithms along with analysis of each and every one of them. This may be effective for some, but I know that I would just as soon have someone point out the source book and let me read it for myself - but what have I learned from such a task? very little in my opinion. On the other hand, if you provide a student with a logical, effective means of approaching algorithm design, then you have given him a head-start that will go far beyond the one semester that you lecture - he/she will be able to take with them an effective technique to approaching problems that may not be in any of your reference books or may require some variation of existing work. Don't get me wrong - exposure to many algorithms is essential but in my opinion is not sufficient to a course on algorithms....
rawlins@iuvax.cs.indiana.edu (Gregory J. E. Rawlins) (05/08/91)
In article <hobbit.673609163@shum> hobbit@SHUM.HUJI.AC.IL (yoav gonen) writes: >I was looking for a book which will help me will the course : Algorithms that >I learn in the university. So I bought Manber's. Unfortunately, the book covers >only 25% of the material that I learned in class. [further comments deleted] I quite agree that Manber's book does not cover as much material as, say, Cormen, Lieserson, and Rivest (a 1000+ page book). Nor does he analyse the material to the depth of, say, Knuth (a three volume book). But that is not his purpose. Before buying another book i suggest you read the preface to the book carefully. I think Manber states fairly clearly his book's niche. Since he does not misrepresent his purpose i cannot see that you have a case. gregory.
hobbit@SHUM.HUJI.AC.IL (yoav gonen) (05/08/91)
Oops! I had a mistake: EUler circuit is O(n^2) (not O(n^3))