On Fri, Apr 3, 2015 at 1:46 PM, Keean Schupke <[email protected]> wrote:
> On 3 April 2015 at 18:42, Matt Oliveri <[email protected]> wrote:
>> On Fri, Apr 3, 2015 at 1:39 PM, Matt Oliveri <[email protected]> wrote:
>> > On Fri, Apr 3, 2015 at 8:02 AM, Keean Schupke <[email protected]> wrote:
>> >> On 3 April 2015 at 12:21, Matt Oliveri <[email protected]> wrote:
>> >>> This is due to the HOAS semantics. Again, you cannot expect to go
>> >>> about all of math in this manner.
>> >>
>> >> Please provide an example. I think you can derive the whole of maths
>> >> :-)
>> >
>> > Well, I'm worried you'll just misunderstand the task, but that would
>> > hopefully clarify the misunderstanding, at least. I'll put it in
>> > another email.
>>
>> Define two predicates on natural numbers, even and odd:
>> even : nat->type
>> even_z : even z
>> even_ss : even N->even (s (s N))
>>
>> odd : nat->type
>> odd_sz : odd (s z)
>> odd_ss : odd N->odd (s (s N))
>>
>> (I'm not sure those are valid Twelf, but you know what I mean.)
>>
>> Prove that any natural number that isn't even is odd.
>>
>> For starters how would you _express_ that proposition, seeing that LF
>> doesn't have negation or falsehood?
>>
>> No, proving that every number is either even or odd is not good
>> enough. The point is you're missing negation.
>
> Indeed, but we an add negation, using negation elimination:
>
> http://www.cs.cmu.edu/afs/cs/user/mobile/thesis/bella/thesis.ps.gz

OK. I'll give you the benefit of the doubt that with additions like
this, a logic programming language can be made strong enough for
formalizing math. (It still doesn't seem like a very convenient way to
go to me.) But it seems we agree that regular Prolog or Twelf aren't
strong/expressive enough, which was what I was arguing.

One of the things I wasn't clear on before was whether you were
talking about using Twelf as a logical framework. It seems the answer
is "no". You were talking about Twelf as a logic programming language
with inductive types, basically.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to