The P vs. NP Problem

in #steemstem7 years ago (edited)

In 2000, Clay Mathematics Institute selected seven unsolved problems in mathematics and put Million dollar prize for solving each problem. These seven problems are known as the Millennium Prize Problems. As of today, only the Poincare conjecture has been solved ( solved by Grigori Perelman in 2003). In this article, I want to talk about my favourite one: "The P vs. NP Problem"[1].
800px-Complexity_classes.svg.png

Image credit: Wikimedia,Public Domain

P Problems

Let's begin with a simple mathematical addition : 1.png

If I give you the value A=2 & B= 3, you can instantly calculate the value of C, which is 5

What if I give you the value A=5,B=2 and C=6 and ask you to check the whether the given value satisfies the equation? Again this is as simple as calculating the value of C in the previous step. We can derive a simple conclusion from this:

  1. The addition problem is easy to solve
  2. The addition problem is easy to check

We categorise this class of problem as "P" problems. ( polynomial time)

NP Problems

Now think about sudoku. Suppose I give you an unsolved Sudoku, you can't come up with a solution instantly but if I give you a solved Sudoku then you can easily check the validity of the solution.

364px-Sudoku-by-L2G-20050714.svg.png

Image credit: Wikimedia, Public Domain

This leads to following conclusions:

  1. Sodoku is difficult to solve but
  2. Easy to check

Now we have a different class of problem: "NP" problems (Non-deterministic polynomial time).Another simple example of NP problem is a jigsaw puzzle. Jigsaw puzzles are difficult to solve since there are thousands of possible combinations but checking the validity of a jigsaw puzzle is easy, even children can do that.

We have two sets of problems "P" and "NP" , and every mathematical(or computational) problem can be categoried into the two sets. Its either "P" or "NP" , and finally the P vs NP problem:

3.png or 2.png

The question is: Whether all NP problems are just P problems (P=NP) or fundamentally the NP problems are different from the P problems(N does not equal to P) ?

We have stated the fact that Sudoku problems are NP problems but can we come up a better Sudoku solving algorithm and render sudoku as a P problem? we know that every P problem is an NP problem, i.e. P is a subset of NP. What we want to know is if the reverse is also true?

We have been classifying problems as "easy" and "difficult" but these words don't mean anything to a computer. To compare the time consumed by an algorithm we have to use the Big O concept.

Big-O[2]

Now you may argue that though Sudoku problems are difficult for us humans, a computer can solve them pretty fast. Yes, even our smartphones can solve 9x9 Sudokus in less than a second but what happens when the scale of the problem goes up? Can our computers catch up with the problem scaling? Give a 1000x1000 soduku to a computer and the time taken by the computer will grow exponentially but this doesn't happen for P problems, even large multiplication problems are easily solvable by computers.We are concerned is with how the runtime of an algorithm increases with its input? Back to Big O.

Big O is a concept in computer science to quantify the performance of algorithms. Let us compare two ways of transferring data:

  1. A pigeon carrying a 1 TB flash drive
  2. Downloading data from the internet.

Up to 1 TB, the time taken by the pigeon to transfer the data from one place to another doesn't depend on the size of the data, the pigeon would take the same time to transfer 1 KB or 200 MB. We call this the Big O of 1, O(1)

An example of O(1): Checking whether a number is even. No matter how big the input is, the time taken by the algorithm is constant.

But in the case of downloading data from the internet, the time required to download 1 MB and 200 MB isn't the same. This is the case of O(n), where n is the size of the data.This means that the time taken to download the data depends linearly on the size of the input.

There are Big Os with higher power too; O(n2), O(n3) and even exponentials exist; O(2N). The graph below shows some of the typical Big Os. On the X-axis we have the input variable as n(size of the input) and on the Y-axis we have the output variable as N(time required for computation).

Let's look at the two most impotant ones:

  1. Polynomial: O(nK), K = 0,1,2,3,.. <<<<<<<<<< Good Algorithms
  2. Exponential: O(2n), O(100n),... <<<<<<<<<<<< Bad Algorithms

As you can see, the polynomial(n and n2) increases slowly in comparision with the exponential. We always want to avoid the exponential and factorials for problem solving algorithm but sadly the NP problem algorithms fall into this "bad category".

480px-Comparison_computational_complexity.svg.png

image credit: wikimedia, CC BY-SA 4.0

Back to P vs. NP


Hope you got the underlining principle of Big O. This will help us understand P and NP problems in detail ;)

Checking the solution of a Sudoku is in Polynomial time ( O(NK), but do we have an algorithm for solving sudoku that operates in poly time too( just like checking it)? So far we don't have any algorithms for that. This leads us to categorize the Sudoku in NP class.

The time required to solve sudoku increases exponentially as its size increases but the time required to check the solution of solved sudoku increases linearly

In the case of P problems, both the checking and solving fall into the Polynomial-time category. Hence they are easier to solve and check, even if the input size is large. If we can find algorithms for NP problems that operate in Polynomial time too, then the P vs. NP problem will be solved but sadly we don't have those algorithms.

There are more!


The subclasses of problems aren't just limited to P and NP, there are more. We have only touched the tip of the iceberg. We have something even harder than NP: The NP-hard problem, they are at least as hard as NP problems but not all NP-hard are NP, so the intersection of NP-hard and NP is the NP-complete. There is another interesting one, which category is best suited for chess? Solving chess is difficult and checking a solution isn't easy either, it doesn't satisfy the criteria for either P or NP, and this results in another class of problem; "EXPTIME-hard". I can go all night on this but let's focus only on P and NP.

640px-P_np_np-complete_np-hard.svg.png

Image credit: wikimedia, CC BY-SA 3.0

Euler diagram for P, NP, NP-Complete, and NP-Hard set of problems

Does it even matter?


Let's suppose that I proved P=NP and won the Million dollar. The implication of this would be enormous. The internet would take the first blow. The whole of internet relies on public key cryptography (prime factorization) but now theoretically we could crack the entire security system. The entire cryptocurrency market is based on the assumption that P!=NP. Basically crypto would die. There is a problem though, even if we manage to prove that P=NP, we still have to develop constructive algorithms for NP problems. Just proving P=NP won't do much. Polynomial algorithms are faster than exponentials but it doesn't necessarily mean that polynomial time will be a reasonable time. Even if someone proves P=NP by developing a poly-time algorithm with O(N100), our public key cryptography would still be safe.

On the other hand, the positive impact would be enormous too. The implications would be big in mathematics. If you manage to prove P=NP, then you might take home Six million dollars instead of one, since solving the other 5 millennium problems would be a lot easier. Another benefit would be in the field of optimization. Multivariable optimization problems pop up in most fields of science, we rely on giant supercomputers and slow algorithms, but with P being equal to NP, these optimization problems could be solved as easily as they can be verified. Protein-Folding is also an important NP-Complete problem. Just like an object falls under the action of gravity to reduce its potential energy, protein molecules fold themselves to minimize their chemical energy, and it turns out that the behaviour of a protein molecule depends on its geometrical shape. Prediction of protein structure correctly would open new frontiers in drugs development.

The implications of the theorem being right, either way, are too much but don't get your hopes too high on P=NP. Most of the computer scientist and mathematicians don't believe P to be equal to NP. In fact, a poll created in the University of Maryland back in 2002 revealed that 69 % believed P!=NP, 9 % supported the opposite and the rest weren't sure about the answer.

Finally, I want to quote Scott Anderson, a complexity researcher at MIT:

If P = NP, Everyone who could appreciate a symphony would be Mozart; everyone who could follow a step-by-step argument would be Gauss; everyone who could recognize a good investment strategy would be Warren Buffett.

My opinion: I am a firm supporter of P!=NP, I know opinions don't matter in mathematics, and this problem might remain unsolved for decades or centuries but until then we have to assume P!=NP. Cheers!

References:


1. http://www.claymath.org/millennium-problems/p-vs-np-problem
2. https://brilliant.org/wiki/big-o-notation/
3. https://math.berkeley.edu/~kpmann/encryption.pdf
4. Guyeux, C., Côté, N. M., Bahi, J. M., & Bienia, W. (2014). Is Protein Folding Problem Really A Np-Complete One? First Investigations. Journal of Bioinformatics and Computational Biology, 12(01), 1350017. doi:10.1142/s0219720013500170
https://arxiv.org/abs/1306.1372

5. https://www.newscientist.com/article/dn21578-poll-consensus-on-million-dollar-logic-proble
6. https://brilliant.org/wiki/p-versus-np/
7. https://cacm.acm.org/magazines/2009/9/38904-the-status-of-the-p-versus-np-problem/fulltext

darwin.gif

DQmeJgsbM5K3pUC8kPBToDKRxE2gijUXvgvX6oUiBgaaiyk.gif

If you are new on steemSTEM or just want to be a part of the community, the please join us on our Discord channel and please check out the posts on #steemstem!

Sort:  

Interesting mathematical stuff. Of more interest is to me because of the implication of proving P=NP in unravelling the structure of proteins. I am a biologist, that would interest me so much.

Well done.

Thanks. Most people view mathematics as an isolated and cold field of study but that's not the case. If we examine the nature of reality closely, everything looks connected. Even the distant looking fields such as biology and mathematics have a lot to share. The protein folding problem is a perfect example of this!

Much as I would have love to be involved in interdisciplinary research, the environment here isn't favourable for such. Most of the leading reseachers work in isolation.

How would this proof help structure biologist? A mathematical proof would be nonsensical ?

I am a bio engineer so I work to improve our ways to find protein structures via many different technologies , MMR, XMR, you name it, and fail to see how the N=NP proof would be of any interest ?

If I run out of money, I guess I could solve a math problem!


The last section is great, it really brought home the importance of the P/NP problem, and all its ramifications, and cryptos too of all things, and protein folding which interests me cos I'm interested in biology in general, and then the Anderson quote took it even further! So you made a pretty convincing case for why you chose this as your favorite problem.

This statement is a bit incomplete: There is one misconception though, even if we managed to prove that P=NP, we still have to develop the algorithms for NP problems. The point of NP-completeness is that if you can show that there exists a NP-complete problem which is also P then you have shown that NP=P. By doing this you will have found a method to solve all NP problems in polynomial time (by the definition of NP completeness).


DQmWPdXSwgcwQV5uuTRybUZH3vUZ6pTw4ESZmDT7Ey7DXbW_1680x8400.png

Yes indeed. That's the whole point of NP-completeness. I was thinking of non-constructive proofs. I mean Polynomial time doesn't necessarily mean it will be quick. If someone comes up with, say O(n100), that wouldn't be practical. But I clearly messed up.

I made the correction. Thanks for the feedback!

I didn't know about the NP-subclasses! But you never defined them? Grrrr... I need to go to sleep, not to google that :D

Sorry for leaving you on a cliffhanger :(
The post was already getting too complex and I didn't want to bring the whole NP-hard stuff. I may write another post entirely about complexity classes :D

Please do so! :)

Trying to wrap my head around it.

I tried my best to keep it as simple as possible, hope you got the basic outline of the problem :)

You got me lost somewhere. You went from P=NP to P!=NP. What’s does P!=NP stand for and what does it imply too ?

Great post btw :)

P not equal NP. ! is the not operator in many logical notations. It is commonly used in computer programming.

Oh thank you. I understand now

This is a very interesting post I am going to have to look more into this. This problem reminds me of Fermat's last theorem. It states that a^n+b^n=c^n when greater then two will have no possible positive integers for n that can be true. It took centuries in order to solve this problem but eventually in 1994 a successful proof showing how the statement above was true by a man named Andrew Wiles. I just found this pretty similar to that math equation discussed in this post. If you want to read about Fermat's Last Theorem here is the Wikipedia link: https://en.wikipedia.org/wiki/Fermat%27s_Last_Theorem

Ah, Fermat's last theorem is a classic. The result looks too simple but took us around 400 years to prove it!

until then we have to assume P!=NP.

it is a safe bet!