Home How To Participate Users Recent Changes All Pages RSS

PnC Chapter 2

Chapter 2: Natural Deduction for Conditionals

This chapter currently has three sections. (It used to have four, but now it has three.) I've decided to rearrange the discussion to freeform rough chronological order so I don't have to move things to a different page when I rearrange the draft. If you have comments about any section, solutions to any of the exercises, or requests for something to add, please them to this page.

Outline

The chapter contains three sections:

Exercise Solutions

(Add Comments/Solutions Here. See the Advice for Typing Proofs.)

Q1.1

                       -----(3)   --- (1)
                       A -> B      A
        ------(2)     --------------- [->E]
        B -> A             B             
       ---------------------- [->E]   
                A
            -------- [->I,1]
             A -> A
      ------------------ [->I,2]
        (B->A) -> (A->A)
  ------------------------------- [->I,3]
   (A->B) -> ((B->A) -> (A->A))

The formula has a linear proof. — Greg

It also has a non-linear proof, using two vacuous discharges.

                 --- (1)
                  A
              -------- [->I,1]
               A -> A
           ---------------- [->,2] (vacuous discharge)
           (B->A) -> (A->A)
      ---------------------------- [->,3] (vacuous discharge)
      (A->B) -> ((B->A) -> (A->A))

This seems to me to be a very different kind of proof. Can you see why I think that? — Greg 2005-03-02 05:25 UTC

Requests

(Add any requests for this chapter here, if they don't fit in the other sections. If you start a paragraph with 'TODO' followed by a colon, then it will appear on my search for things I need to do with the text.)

You can go back to the Outline of PnC, or to the main page for the book: Proof and Counterexample.

Discussion

When you add yours, please date your entry, and — if you like — sign your entry so we know to whom we owe thanks.

The easiest way to sign is to simply write a string of four "plus" symbols — like this ++++ — and make sure that the "Username" field of the form contains your name. The string will be converted to your name and a datestamp.


Mar 1, 2005

Page 8, end of first paragraph. Missing "not"? "I will feel free to use expressions such as […] even though they are, strictly speaking, in FORMULA."

Thanks! That'll be fixed in the next update. — Greg

Mar 2, 2005

Page 8, first paragraph in the margin, it says "If you like you can think of them has…" — Olaf

Thanks! That'll be fixed in the next update. — Greg 2005-03-02 03:42 UTC

(I notice that we're coming from significantly different timezones. I've added the signature extension, so we can have timestamped comments with a uniform timestamp. Insert ‘++++’ (without the curly quotes) and it will be replaced by your username and a UTC timestamp.)


Mar 3, 2005

Page 11, the paragraph beginning "I have been intentionally unspecific …" has two instances of [→ E] that should be [→ I], and the second offset proof on the page also has [→ E,1] instead of [→ I,1]. — David Scarratt 2005-03-03 02:44 UTC

Hi, David! Thanks for that. It'll be fixed. — Greg 2005-03-03 05:06 UTC

Mar 3, 2005

Page 12, line -8, another [→ E] for [→ I]. — David Scarratt 2005-03-03 06:08 UTC

Got it. — Greg 2005-03-03 06:46 UTC

Mar 4, 2005

Page 13, line -4, "An counterexample". — David Scarratt 2005-03-04 04:38 UTC

Fixed. — Greg 2005-03-04 08:36 UTC

Mar 5, 2005

Page 14, paragraph 2, line 5 thereof; should 'deleted' be 'discharged'? — Olaf 2005-03-05 01:48 UTC

It would make sense with 'discharged', or with 'deleted.' The way that you get the new premise set from the original one is by deleting some copies of a premise that was discharged in the proof step. Either word makes sense, I think. — Greg 2005-03-05 10:14 UTC

Mar 7, 2005

Ch. 2, Exercises, Q.1, iii. Formula is missing ")" parenthesis, on the end.

Not too sure about the next one, but p.15, Example 2.9 [A simple valuation], last line thereof: "0 + 0 = 0 …". Should it not be "0 * 0 = 0", as we have not set a * b = a + b ? — Olaf 2005-03-07 08:52 UTC

I think the formula is OK. The version I have here is (A → B) → ((A → (B → C)) → (A → C)), which has balanced parentheses. The antecedent is A → B, the consequent is (A → (B → C)) → (A → C). Its antecedent is A → (B → C), and its consequent is A → C. On Example 2.9, you're right. It should be "0 * 0 = 0." Using "+" is just confusing here. (And it's a leftover from when I used "+" in the zeroth draft, before I changed over to "*".) Thanks. — Greg 2005-03-07 09:19 UTC

Mar 9, 2005

page 16, proof of lemma 2.11, lines 6-7 of proof " Take any c E P where c |⊢ A. We wish to show that ( a * b ) * c |⊢ A → B … " The way that I understood the proof, we want to show ( a * b ) * c |⊢ B , and not the conditional formula… I hope I am not making a monkey of myself

No, that's not being a monkey. That's right. I stuffed up. It should be "We wish to show that "(a * b) * c |⊢ B". — Greg 2005-03-09 05:48 UTC

also, one that Kate picked up on p.15, line before Example 2.9. Is the turnstyle-y thingo meant to be ' |⊢ ' rather than ' |= ' ? … sorry Kate if you think I'm trying to steal your brownie points :) — Olaf 2005-03-09 03:45 UTC and Kate by proxy

Yep. Any |= occuring in Chapter 2 is an error. — Greg 2005-03-09 05:48 UTC

Greg, I think I would find a slightly longer or more detailed explanation of the notion of "points" quite helpful in this section; or perhaps a reference to another text that deals with it more explicitly? — Olaf 2005-03-09 03:48 UTC

I think that's right. I feld awfully rushed when teaching the materials on models. There's more material there than I suspected. Maybe it's better excised out of the chapter and left for a little bit later. Do people think that the flow would be better if I did that? (I only noticed this comment today. I'm used to checking for typos.) — Greg 2005-03-11 04:32 UTC

Mar 11, 2005

page 13, definition of "finite multiset". "the members may be present any finite number of times" You mean to say that the number of times each member is present is a natural number (or counting number, in your words). Otherwise I would like to have a multiset with an element that is present π (or \sqrt{-1}) times! No one would be confused with the present wording, but it doesn't seem (to me) quite right to drop to such a level of informality here.

After that you start omitting the word "finite" and talk about "multisets". Perhaps you had better make it explicit that when you say "multiset" you always mean "finite multiset".

The definition in the margin as a function is my preferred definition (being a computer scientist), but here we come up against this ambiguity - you say "you can also define a multiset" but do you mean to give an alternative definition of "finite multiset"? If so, you have to include an extra clause of the form "where f(A) is non-zero for only a finite number of A \in FORMULA". — Richard Walker 2005-03-11 04:14 UTC

Hi, Richard! Nice to see you here. TODO: Yes, I'm conflicted about how best to define multisets. Maybe I should use the definition in terms of functions, but my intended reader is not familiar with functions considered as first-class objects. So I'm not sure of the best pedagogy at this point. — Greg 2005-03-11 04:29 UTC

Mar 12, 2005

This is definitely the clearest treatment of discharge I've seen yet: the formal notation gives excellent intuition without cluttering the proof-trees. (I'm hoping it carries over to proofs by contradiction, too.) However, I did find the margin-note next to definition 2.2 ("Now is a good time to try some of the exercises") a bit jarring; because I eagerly turned to do so, only to find that the material covered up to that point only allowed half of Question 1 to be done (ignoring the request for counterexamples). Tentative suggestion: why not split Q1 into two questions, the first giving the list of theorems to be proven, and the second asking for counterexamples? Or put the suggestion a bit later. — Benet 2005-03-12 04:42 UTC

Thanks for the positive feedback on the discussion of discharge. I agree that I must change that marginal note. I'm not sure whether to go for a radical restructuring of the material (Chapter 1: Natural Deduction for Conditionals (no models), Chapter 2: Sequents for Conjunction & Disjunction (no algebras) and a new Chapter 3: Models, in which we do the discussion of models in a slightly more leisurely way.) If I do that, then the exercises will have to be untangled. If I don't, I'll untangle the exercises anyway. — Greg 2005-03-12 22:23 UTC

Mar 14, 2005

As I work through this section, I'm wondering how much mathematical maturity is expected/assumed from the reader and/or student. If this is a first exposure to logic, it may also be a first exposure to the concept of induction on structure: so, particularly in Lemma 2.11 (Preservation for Conditionals), I was thinking that making some of the steps more explicit might be easier on the student, especially since this is the first non-obvious result in the book, and quite algebraic. For instance, both times when Definition 2.8 is used (before "Take any c \in P" and after "since a |⊢ A->B"), I think a brief reminder to look back to the definition would help. As to the placement, I actually think having models and counterexamples this early – immediately on the heels of defining proof – is good, unless it means having to slow down the presentation too much.

Two more very small points: these are things that might be pitfalls for readers who are more familiar with mathematics and mathematical logic. 1) It is a little bit jarring that little 'p' ranges over atoms, not over elements of big 'P'. I can't think of an ideal solution: P is the best name for the set of points, but p,q,r,.. are the traditional symbols for propositional atoms. But thought I'd make a note of it. 2) It's easy for someone used to classical model theory to keep forgetting that |⊢ is a relation on atoms and points, not a function from atoms to points. Maybe emphasize this a little bit more? — Benet 2005-03-14 00:49 UTC

Yes, I'm convinced! I've got to cater for the kind of student who isn't yet comfortable by a proof by induction. I've got to restructure this chapter. I'll be doing some rewriting and expanding the section on counterexamples, including more gently looking at induction, and flagging how the notation works. It might end up in a new chapter, it might end up as a separate chapter. I'm not sure about that. I've got some writing to do tonight! (I really appreciate the feedback, everyone.) — Greg 2005-03-14 05:00 UTC

Added Mar 19th, 2005 — I guess fewer things are obvious to me; I didn't think that theorem 2.7 was all that obvious. It's also a proof by induction, (isn't it?) I think novice readers may balk at "Since pi' is normal, we may assume that every formula in pi' is in sf(x, B)." I was wondering if you could change the structure of the proof so that what you are proving is that a first step containing a formula not in sf(X, A) can't be introduced through [->I] or [->E] (and of course trivially it won't be an premise.) Since every step is a premise, or the result of one of those rules, that means there's no first step containing a formula not in sf(X,A), and of course, if there were any, there would be a first one, so it follows that all the steps are such that the formulas are in sf(X,A). Maybe this is still inductive in a sense, but I think it's much more intutive for people who haven't been taught induction explicitly. (An alternative might be Graham's strategy from An Introduction to Non-classical Logic; you could explain induction in an appendix (or the introduction) and refer beginners to that when you give the first few proofs by induction.) I think it's really difficult (maybe too difficult, I don't know) and admirable to try to make the book accessible to beginners; there are so many books that start out well and then give up around page 10. — Gillian Russell 2005-03-19 19:34 UTC

I'm going to slog at it. (Where it is making the thing accessible.) My strategy in the writing is to try to get the order of the examples/definitions/theorems/lemmas/corollarys/etc. correct, and fill out as much explanatory text as is required. Having my wonderful readers giving me advice on what is needed is just the ticket. — Greg 2005-03-20 09:52 UTC

Mar 16, 2005

At this point I restructured the chapter, cutting out the section on models and moving that into [PnC Chapter 4]?. — Greg 2005-03-15 14:10 UTC


Mar 16, 2005

2.3 (not sure what it is in the new version) Normal Proofs, first paragraph: "Some proofs are quite strange. Take a proof concludes with…"

and also, paragraph before Def. 2.22 [Complexity], (p.23 Old version of text): "we must will some other measure that decreases…"

paragraph after Def. 2.23 [Non-normality] (p.24 old version): "Non-normality measures are satisfy the…"

Greg, are you still interested in typo checks, seeing as you are restructuring and rewriting the chapter? — Olaf 2005-03-16 04:09 UTC

Yes, the more typos we find, the fewer will be there later. (The restructure is a restructure, not a complete rewrite. I'm just moving stuff around an expanding things. I'm deleting little.) Some of these remained in the text. In general, you could check if the typo is present in the latest version before posting here, but everything helps. — Greg 2005-03-16 06:59 UTC

Mar 18, 2005

Hi Greg, LOVE the aims for the book, and LOVE what I've read so far. Here are a couple of comments to add to the mix:

Thanks! The praise means a lot to me. — Greg 2005-03-19 09:37 UTC

P.8 "logic is an account of general principles of valid reasoning." If this is what you think, then it's what the book should say. But it's not what I think, and so I was wondering, is it really what you think? (For the reasons I gave at the end of the Goldfarb review.)

You're right. I don't think I agree with that. (I'm not sure why it came outh that way.) — Greg 2005-03-19 09:37 UTC

P.13 I found the finite multiset definition confusing. The first sentence is "A finite multiset of objects is a finite collection of those objects in which the members may be present any finite number of times." One natural way to read that is as saying that an element might be a member at one time (say, 5pm) but not be a member at a later time, but maybe be a member at a later time again (in which case it would be present twice). But at any one time they are either a member or not a member. On that reading I am a member of a multiset a finite number of times in the same sense as I visit your website a finite number of times. But on the intended reading I am a member of a multiset a finite number of times in the same sense as the letter "l" appears in "Philadelphia" a finite number of times. I know what a multiset is and I still read the first sentence the first way; I wondered whether this might be some new kind of multiset … A reader who doesn't have any idea what one is could be mislead here and that would be the first really frustrating experience in the book…I think it should be either somehow rephrased, or you could put in one of those little side notes pointing out the two possible readings and identifying the right one. — Gillian Russell 2005-03-19 06:35 UTC

Thanks for clarifying the way that the definition could confuse. I certainly didn't intend "time" to be used in the temporal sense (except perhaps loosely when one enumerates the elements of the multiset and you list them one-at-a-time). So, I'll clarify this in the next draft.
(Oh, and one other thing. I use the "major/minor edit" distinction to go roughly along the distinction between edits that bring new content and edits that fix typos etc. I don't think the comments here are minor edits.) — Greg 2005-03-19 09:37 UTC

OK, sorry, I'll remember the distiction in future. — Gillian Russell 2005-03-19 19:00 UTC


Mar 19, 2005

P. 15, final complete paragraph, "conjuring new arguments out of thin air" — of course we get new arguments underwritten by proofs when we introduce the conditional rules; we didn't have any rules before, so we couldn't have any proofs with steps in them. Isn't this getting new arguments out of thin air? But the conditional rules are ok, right? I'm being flatfooted here, but I also don't know what the strictly correct way to put this is.

P. 15, "if the conditional is defined by way of these rules, …" — one thing I've learned from the comments on my weblog is that not everyone gets the intuitive pull to think of the rules as defining the connective, or at least, they don't get it straight away without any explanation. I think this is the first time you've mentioned it; my hunch is that some readers will be puzzling over this antecedent unless you say more.

P. 16, definition of subformulas is very clear.

Gillian Russell 2005-03-19 19:00 UTC

Super comments. I was reading the comments on your weblog with interest, and I though that I would need to spend more time clarifying the reasons why one would take inference rules to constitute the meaning of logical particles. Actually telling the tale can help clarify what's going on when people take the rules to be definitions. — Greg 2005-03-20 09:52 UTC

Mar 19, 2005

I took a quick peek at this chapter, and besides being quite intrigued and impressed, I had one small comment to make. On page 15 of your March 16th, 2005 draft, you use the term "critical pair" to refer to a detour in a proof, where a use of an introduction rule for a connective is followed by a use of the connective's elimination rule. In term rewriting theory, "critical pair" has a special meaning. Since term rewriting theory overlaps a bit with applied proof theory, I was wondering if you were intending to use "critical pair" here with the same meaning, or whether perhaps it was an accidental namespace clash. A quick description of the meaning in term rewriting is here (or see something like page 562 and following of the Rewriting Chapter of the Handbook of Automated Reasoning for a more detailed discussion). — Aaron Stump 2005-03-19 18:36 UTC

Thanks, Aaron. I've completely mistaken my terminology here. It's a stupid slip, and I'll fix it up. Thanks for spotting it! — Greg 2005-03-20 09:52 UTC

Further to Gillian's comment about P.8 "logic is an account of general principles of valid reasoning.", maybe a footnote summarising a range of suggested definitions would be in order to put your approach in context? FYI, the first chapter of my DPhil begins Logic is the study of arguments correct in virtue of its form. which serves as a sort of announcement of hostilities against the Tarskian view of logic. — Charles Stewart


Mar 20, 2005

Greg, I'm wondering about the relation between premises and assumptions. Section 2.1 seems to draw the distinction that proofs have assumptions, and arguments have premises; the relation being that X (therefore) A is a valid argument iff there exists some valid proof from X to A. Section 2.1 also seems to state (or at least suggest strongly) that whether in an argument or proof, the set X (resp. premises and assumptions), contains only un-discharged formulas. However, in Section 2.2 (top of p.15 in the current draft), the text speaks of a proof having "two instances of A in the premise list", and in the following figure both instances appear discharged. Furthermore, the illustration is a proof rather than an argument. (To make an even tinier quibble, if you plan later to distinguish between multisets of premises and lists, probably it's a little misleading to use list here.)

Thanks for pointing out the difficulty in interpretation here. At this stage, I intendent to take π1 to be the proof up to the deduction of B, and hence to have two instances of A as (undischarged) assumptions. However the notation here is unclear. That interpretation is actually needed in order to get the before-and-after picture right. (The partly normalised proof contains the proofs π1 and π2, so it cannot contain the step of the original proof at which those instances of A are discharged.) So I must clarify this. The quibbles are well-made, and I will fix them up too. — Greg 2005-03-21 12:03 UTC

Now that I look back at the text, I note that the proof on page 11 of (A->B) states that A->(A->B) is the premise/assumption of the proof; but later, on page 13, it says ".. we have a linear proof from A->(A->B),A,A to B, but.. [no linear proof] from A->(A->B),A to B." If only undischarged assumptions ever appear in either the premise multiset or assumption multiset, I'm not quite sure I see how this distinction can be drawn, because if all the A's are discharged none can actually appear in the premise multiset under the given definition. I am quite sure that I'm missing some subtle – or, indeed, some basic – point, and my only justification for going on about it at such length is that I imagine it would be very possible for a student new to logic to miss it too. — Benet 2005-03-21 02:15 UTC

I'm not sure I get this point, but I agree that my explanation is not yet crystal clear. The proof on page 11 here is a proof from the assumption A → (AB) to AB, in which two assumptions of A are made and later discharged (in the last inference). This proof isn't a linear one. Inside it, before the discharge of A is made, we have a linear proof from A → (AB), A, A to B. (This proof consists of two instances of [ → E].) A linear proof from A → (AB), A to B would be a different kettle of fish. It would have to be a proof in which the only undischarged assumptions are A → (AB) and A (one instance each) and with the conclusion B. The proof we have from A → (AB) to AB won't help here, since it used a non-linear (duplicate) discharge. As a matter of fact, there is no linear proof from A → (AB), A to B, as a counterexample for that inference can be constructed. Does that make a bit more sense now? — Greg 2005-03-21 12:05 UTC
Thanks for the thoughtful response. I understand the distinction between linear, relevant, and standard proofs based on number (1, more than 0, any) of discharges of an assumption. Let me try to restate my question: take the non-linear proof on page 11 from A → (AB) to AB, the one where two assumptions are made and then discharged. I'm pretty confident in saying that this underwrites a (standardly valid) argument where X is the singleton multiset A → (AB). The only thing which puzzles me is the paragraph on page 13 preceding Def. 2.4: what is the difference between the linear proof taking A → (AB) and two instances of A (which later get discharged), and the standard proof where one copy of A gets discharged twice? Does this mean that the argument from X ( = A → (AB)) can be called linearly valid? I think the main point of the puzzlement is this distinction between multiple copies of a discharged assumption (being discharged, so not appearing in the premise set), and multiple discharges of a single assumption. The confusion isn't sufficient to keep me from pressing on with the succeeding material, either the model theory or the proof theory, but it is there. — Benet 2005-03-21 23:58 UTC
The linear proof taking A → (AB) and two instances of A that are later discharged will be a proof from A → (AB) to the conclusion A → (AB), and not AB (which is the conclusion in the non-linear proof). So,
  A -> (A -> B)   A
  -----------------
        A -> B       A
       ---------------
               B
is a linear proof, with two assumptions of A. Now discharge one. You get this.
                 --- 1
  A -> (A -> B)   A
  -----------------
        A -> B       A
       ---------------
               B
            ------- [->I,1]
             A -> B
It's still a linear proof. But only one A is discharged. (We can't discharge both at once here.) So, we have a linear proof underwriting the argument A → (AB), A to AB. Now discharge A again and we get a linear proof
                 --- 1
  A -> (A -> B)   A
  -----------------  --- 2
        A -> B        A
       -----------------
               B
            ------- [->I,1]
             A -> B
         ------------- [->I,2]
         A -> (A -> B)
but it's one in which the conclusion isn't the desired AB. Does that help, or am I answering a question you're not asking, and not answering the question you are asking?
To everyone else, are you getting Benet's question? Can you answer it? What can I explain better? — Greg 2005-03-23 13:41 UTC

Greg, thanks for the response. It clarified a lot of my puzzlement. I think I still have some questions related to discharge and book-keeping; but I understand now how you are distinguishing between single and multiple discharge. I'm going to let these questions simmer on a back-burner for a couple more drafts and see if they are either resolved by the continuing development, or if I can put them in a concise form. — Benet 2005-03-27 05:54 UTC


Mar 21, 2005

"modest super multi-sets" — here's the most conservative suggestion I can think of: just abbreviate "multiset" to "m-set" or "mset". Admittedly, "modest super m-set" is only one syllable shorter, but "proper superset" is only one syllable shorter again and that standard piece of terminology is fine because in practice we normally talk in terms of the subset relation rather than the superset one. In practice we might end up doing the same here: y is a modest submset of x" (read: "y is a modest sub-m-set of x") rather than "x is a modest super multiset of y." (If you don't mind the idea of slightly less conservative you might decide to write "sumset" for "sub-multiset" (wheras the less-used "supermset" wouldn't be shortened.) — Gillian Russell 2005-03-21 23:51 UTC

I don't like the idea of abbreviating multiset, as that's a completely standard usage. I'm thinking, instead, of talking of the ground of a multiset (the set of its all of its elements-to-any-degree), and then we can see that one multiset might cover the same ground as another, but cover the same ground to a lesser degree (x has the same ground as y but things are be members of x less often than they are members of y) or to a greater degree (more often). So, x is a modest supermultiset of y (the old terminology) iff x and y cover the same ground, but x covers that ground to a greater degree. Does that sound less clunky? — Greg 2005-03-23 13:41 UTC

I like it - at least as a definition of modest super multiset; but as replacement terminology (which is what you seemed to be asking for on P. 18) isn't it a bit, erm,…long? (Of course I understand about not wanting to abbreviate 'multiset'—fair enough.) — Gillian Russell 2005-03-24 10:29 UTC

Do all multisets have the same ground? (Everything's a member to some degree, possibly zero?) — Gillian Russell 2005-03-24 22:47 UTC

I'm not sure whether it would reduce or promote confusion, but 'bag' is another standard term for multiset. 'Modest super-bag' has its own problems, but has fewer syllables at least. Is 'ground' standard? I'm more familiar with 'support' as the term for those elements of the universe of discourse that have non-zero presence in a multiset (or other set generalization). (Speaking of multisets, I just wanted to introduce another teensy notational question: if you are using \box for the empty multiset, is there a possibility that will introduce confusion when you get to the later part of the book and use \box as a modal operator?) — Benet 2005-03-25 21:12 UTC


April 3, 2005

Hello, I've been reading the printing of Mar 23, and had some comments. It's very interesting to track the progress of a book; thanks for making this visible. After Chapters 2 and 3, I think I'm finally beginning to get to grips with normalization and the notion of cut; the chapters are exceptionally clear. Here's some feedback on which portions I got confused at, and some typos.

In the first pages 7 and 8 of Chapter 2, the relationship between a formula and a "conditional judgement" (the latter being described in the motivating para) is not explicitly stated, but left implicit. That is, it might make things a little clearer if there is a line somewhere stating, "Now, after setting up our formal language, we can express our notion of conditional judgement using a formula as follows …".

On page 17, in the proof of Thm 2.8, in the Introduction case, it seems that there's an implicit acceptance of the fact that a subproof of a normal proof is also normal. Perhaps that fact could be made more explicit, or even proved, somewhere close to where the notion of a normal proof is defined. Also, in the Elimination case, I think the variables X and Y were swapped midway through the para. Perhaps a marginal diagram might clarify the variables here.

In a few places, proofs of lemmas and theorems follow a page or more after they were stated. It might help searching for proofs if they were explicitly labelled, e.g. "Proof (Thm 2.10, relevant and standard case)".

In the descending sequence of non-normality measures in Example 2.14, the measure (1,3088) is not smaller than (138, 47).

Prashanth Mundkur 2005-04-03 23:54 UTC

Thanks for catching these. It looks like I've completely botched the non-normality measure discussion. By the time I got to the example, I was reading the order of the list in reverse! Thanks for picking that up. — Greg 2005-04-07 04:42 UTC
Yikes, I'm not sure what you mean. On re-reading the non-normality measure stuff, it seemed clear enough to me, unless I missed some crucial detail. BTW, could we start a page on Chapter 4, which I've started reading, but in contrast to Chapters 2 and 3, I'm finding fairly heavy going. I'm still in 4.1. I can see the trees, but not the forest. Is there any way the basic intuitions could be illustrated pictorially? — Prashanth Mundkur 2005-04-09 20:11 UTC