Post me your hardest maths question you know

As I'm sure you can show (with your fingers?!), given any finite sequence A,B,C,D,...,E of numbers, there are infinitely many different functions such that f(1)=A, f(2)=B, f(3)=C and so on. So "spotting a pattern" will never constitute a proof, unless you can then go on to prove that your guess is correct for all numbers.

For instance, if I worked out the problem in hand and got upper bounds of

4, 6, 8, 10, 12

for n = 4, 5, 6, 7, 8 gossiping women, then I might guess an upper bound was

f(n) = 2n - 4 + (n - 4)(n - 5)(n - 6)(n - 7)(n - 8)

which agrees with my list for n = 4, 5, 6, 7, 8. But of course we now know this guess (which is but one of infinitely many) is wrong. :)

You didn't ask for proof of the formula, just the formula! I never said my fingers could prove it :p

Edit: Computers were made for proving stuff like this :D

Edit2: Oh, and my fingers weren't used to count ;)
 
You didn't ask for proof of the formula, just the formula! I never said my fingers could prove it :p
Apologies - I should have been more explicit. If you guess a result, you should prove it is correct! In fact, I think the original problem gives the result and simply asks the reader to prove it. And I stress again: it is very hard, and I don't realistically expect anyone to solve it. The spaghetti problem is much more approachable.
The spag. answer must be zero. A loop does not have a knot in it. All spag. will end up knotted. Therefore no loops.
Out-of-the-box thinking is recommended, but not so much that it trivialises the problem!
 
Apologies - I should have been more explicit. If you guess a result, you should prove it is correct! In fact, I think the original problem gives the result and simply asks the reader to prove it. And I stress again: it is very hard, and I don't realistically expect anyone to solve it. The spaghetti problem is much more approachable.

Out-of-the-box thinking is recommended, but not so much that it trivialises the problem!

Oh, I don't even begin to think I can mathematically solve it. However my fingers help me visualise search spaces and optimum search patterns. CompSci training I guess :D
 
Here is a very hard problem that doesn't require any advanced mathematical techniques to solve, just a huuuuge brain!

There are N>3 gossiping women and each of them knows some item of gossip not known to the others. They talk to each other via the telephone and in each conversation they tell each other all they know at that point in time. What is the minimum number of phone calls required before each woman knows all the gossip.

Hmm, interesting question.
I've got something that shows you can do it in 2N - 4 for an even number of women. I imagine you could extend this to do it for an odd number too, but it doesn't really show that this is the minimum number of conversations required.
I'd be interested in knowing the proper proof as I'm sure it's a lot more elegant than what I have below!

Ignoring women gossiping and thinking about it in terms of connected nodes.
Basically, it is:

Let there be C nodes and let N(i) represent the ith node.

Step 1
Join the nodes into C/2 disconnected pairs (i.e. 1-2, 3-4, etc.)
Thus each node has knowledge of one other node.

This step takes C/2 connections, one connection for each pair.

Step 2
Now join N(2) to N(j) where j is even and <= C/2
i.e. N(2) -> N(4), N(2) -> (N6) etc.
if C/2 is even this needs (C/4)-1 connections.
if C/2 is odd it needs ((C-2)/4)-1 connections.

Now, we do the same for N(k) to N(C) where k is even and > C/2

if C/2 is even this needs (C/4)-1 connections.
if C/2 is odd it needs ((C+2)/4)-1 connections.

And we see that to complete to this step requires (C/2)-2 connections whether C/2 is even or odd.

Step 3

We now have two nodes, N(a) & N(b) where a,b <= C/2 that know about all other nodes from N(1) to N(l) where l is the largest even integer that is <= C/2

Similarly we have two nodes N(c) & N(d) where c,d > C/2 that know about all nodes from N(l+1) to N(C)

Now, join N(a) with N(c) and N(b) with N(d) to give four nodes that know about all nodes from N(1) to N(C)

Step 3 takes two connections.

It has taken C/2 + (C/2) - 2 + 2 = C connections to get this far.
As there are four nodes that know about all C nodes there are C-4 nodes that don't, so we join from one of the nodes in the know to the other nodes, which takes C-4 connections.

Total is C + C - 4 = 2C - 4 connections.
 
If this is the first time you've ever seen this problem or anything like it, then that's very good attempt. Sadly, the difficulty of the question lies in the proof that 2n-4 is truly the minimum number of calls. A hint can be found in Bollobás' "Extremal Graph Theory", and a complete solution can be found in "The art of Mathematics", by the same author.

A simpler proof for what you've attempted (for all n) is this: make a split n = (n - 4) + 4, and have the (n - 4) women all call the remaining 4. Have these 4 women then speak to one another so they each know everything. Then get the other n - 4 women to give one of them a call. Count up the calls and you will find this takes 2n - 4.
 
If this is the first time you've ever seen this problem or anything like it, then that's very good attempt. Sadly, the difficulty of the question lies in the proof that 2n-4 is truly the minimum number of calls. A hint can be found in Bollobás' "Extremal Graph Theory", and a complete solution can be found in "The art of Mathematics", by the same author.

A simpler proof for what you've attempted (for all n) is this: make a split n = (n - 4) + 4, and have the (n - 4) women all call the remaining 4. Have these 4 women then speak to one another so they each know everything. Then get the other n - 4 women to give one of them a call. Count up the calls and you will find this takes 2n - 4.

Nope, never seen this until last night.
I have a degree in mathematical physics, but graduated nearly 10 years ago and have pretty much forgotten all of it now! Hell, I'd struggle with A-Level maths these days.

In terms of proving it's the minimum I was thinking along the lines of the fact that (using my terminology) for C nodes, it takes at least C-1 connections for one of the nodes to know everything as the other C-1 nodes must have at least one connection, otherwise they will be isolated and not be able to share their knowledge, but couldn't really get anything rigorous with it.

I always hated the maths stuff in my degree, thought about switching to straight physics numerous times. Things like this justify my dislike :p
 
I have a bowl containing 1000 strands of spaghetti. I pick two ends of spaghetti at random and tie them together, and I keep doing this until there are no more ends to grab. What is the expected number of loops in the bowl when I'm done?

The first time you connect two ends, you have a 1/1999 chance of making a loop, and a 1998/1999 chance of not making a loop.
The next time you connect 2 ends, you have a 1/1997 chance of making a loop, and a 1996/1997 chance of not making a loop.
etc etc until you add all of the chances of making a loop up
1/1999+1/1997+1/1995...........
This can be done quickly using excel ( :p ) finding that you will have on average 4.44 loops
(This might be completely wrong)
 
Awesome - nice job! Here's another, which is probably a little easier but still nice.

Dave and Jim play a coin tossing game: a coin is tossed until either the sequence HHH occurs (in which case Dave wins), or the sequence THH occurs (in which case Jim wins). What is the probability that Jim wins?
 
Dave and Jim play a coin tossing game: a coin is tossed until either the sequence HHH occurs (in which case Dave wins), or the sequence THH occurs (in which case Jim wins). What is the probability that Jim wins?

Jim wins every time (lucky chap) to get an HHH it needs to be preceeded by a THH.

EDIT: actually Dave can win if (and only if) the first three tosses are all heads. Simple probability i cba to do :p
 
Awesome, although I'm baffled as to how you got that and were unable do the 1/2 x 1/2 x 1/2 = 1/8 in your head! Here's another:

I have to randomly distribute 10,000 prizes among 10,000 Christmas crackers (multiple prizes can go into a single Christmas cracker). What is the probability that after I've distributed all the prizes, exactly one Christmas cracker is left without a prize?
 
Back
Top Bottom