structured/0707.1176.structured.json

{"paper_meta":{"paper_id":"arxiv:0707.1176","title":"0707.1176","authors":[],"year":null,"published":null,"updated":null,"venue_or_source":"arxiv-papers-shard","primary_category":"cs","secondary_categories":[],"doi":null,"license":"unknown","source_type":"pdf+shard","source_url":null,"pdf_url":null},"abstract":{"raw":"","cleaned":"","offsets":[]},"full_paper_text":{"raw_ordered_text":"Expressing an NP-Complete Problem as the Solvability of a Polynomial Equation\nDeepak Ponvel Chermakani, IEEE Member\ndeepakc@pmail.ntu.edu.sg deepakc@usa.com deepakc@myfastmail.com deepak.chermakani@ustri.com\nAbstract: We demonstrate a polynomial approach to express the decision version of the directed Hamiltonian Cycle \nProblem (HCP), which is NP-Complete, as the Solvability of a Polynomial Equation with a constant number of \nvariables, within a bounded real space. We first introduce four new Theorems for a set of periodic Functions with \nirrational periods, based on which we then use a trigonometric substitution, to show how the HCP can be expressed as \nthe Solvability of a single polynomial Equation with a constant number of variables. The feasible solution of each of \nthese variables is bounded within two real numbers. We point out what future work is necessary to prove that P=NP.\n1. Introduction\nGiven a graph G (V, E), where V is the set of N vertices, and E is the set of edge-costs that are either 1 or 0, the decision \nversion of the HCP is to find out whether there exists a round trip tour to visit each vertex exactly once, such that the total \ncost of the tour equals N. If the edge-cost in traveling from one edge to another, is the same as the reverse cost, then the HCP \nis said to be undirected, while if it is either same or different than the reverse cost, then the HCP is said to be directed. Both \nthe directed, as well as the undirected forms of the decision version of the HCP, are known to be NP-Complete.\nIn the rest of this paper, we shall refer to “decision version of the directed HCP”, simply as the “HCP”. Our paper \ndescribes a new polynomial Approach for expressing the HCP as the solvability of an Equation. Our Approach is discussed in \nSection-3 and is based on new Theorems, which we will introduce in the next section (Section-2) of this paper.\n2. Theorems on a Set of Periodic Functions with Irrational Periods\n2.1 Definition of the Periodic Functions\nLet us define n periodic Functions to have the following three properties.\nProperty-1: Period of Function i, Ti = (1 / (1+ iR) ), where R is irrational non-zero real, and i represents integers 1 to n\nProperty-2: All the n Functions are binary functions and have a binary high value for the same duration x, where x is non-\nzero, real, positive, and lesser than the smallest period, which is 1 / (1+ nR).\nProperty-3: Function i is allowed to start to become high from time t=0, with real delay Di , where 0 ≤ Di < Ti\nFor better clarity, we attempt to show these three properties of the n Functions, in Figure-1 below. Note that the \nvariable t along the t-Axis, varies from 0 to positive infinity.\nFig 1. Definition of the Periodic Functions\n\n2.2 Definition of E(t)i\nWe define E(t)i as the closest positive displacement from t when Function i again starts to become high. For the purpose of \nbetter clarity, we show this definition in Figure 2 below.\nFig 2. Definition of E(t)i , for Function i\n2.3 Theorem-1:\nIf we consider any two Functions i and j from the n Functions defined in Section 2.1, then every real value of t in [0, ∞[ has a \nunique set {E(t)i , E(t)j }. In other words, {E(tA)i , E(tA)j } ≠ {E(tB)i , E(tB)j }, if tA ≠ tB.\nProof: If {E(tA)i , E(tA)j } = {E(tB)i , E(tB)j }, for say tA < tB, then it would mean that it is possible to identify another \npoint tC where {E(tB)i , E(tB)j } = {E(tC)i , E(tC)j }, and where tC – tB = tB – tA. It would also mean that an infinite \nnumber of similar points like tC can be periodically found along the t-Axis. This, in turn, would imply that N1 Ti = N2 Tj, \nfor some natural numbers N1 and N2, where N1 and N2 ≠ 0. But we know that (Ti / Tj ) = (1+ jR) / (1+ iR), which is \nirrational and therefore cannot be expressed as (N2 / N1). Hence Proved Theorem-1.\n2.4 Theorem-2:\nIf we consider all the n Functions defined in Section 2.1, then every real value of t in [0, ∞[ has a unique \nset {E(t)1 , E(t)2 , E(t)3 ,......., E(t)n-1 , E(t)n }\nProof: This is straightforward from Theorem-1, and it is easy to see that all the periods of the n Functions are irrational with \nrespect to each other. Hence Proved Theorem-2.\n2.5 Theorem-3:\nIf we consider any two Functions i and j from the n Functions defined in Section 2.1, then the following two statements are \nTrue (also see Figure 3):\ni) There exists an infinite number of Intervals along t-Axis, where these two functions are simultaneously high for a \nduration of atleast (x - μ), where μ is positive non-zero real and (μ/x) 0.\nii)The starting points of such Intervals (e.g. tA , tB in Figure 3) do not have a constant period, but their occurrence \nalong the t-Axis can be expressed by some relationship involving Ti , Tj, Di , Dj, μ, and x.\nFig 3. There are infinite Intervals along t-Axis, when the 2 Functions are simultaneously high for duration of atleast x - μ\n\nProof: We will first prove this by assuming zero-delays for Functions i and j. We will then incorporate delays, and show that \nthe proof holds. We define a ‘Token’, as a movable point used to indicate the current location of a process along the t-Axis.\nLemma-3.1: If Di = Dj = 0, there exists some real t > 0, where the start of the highs of Functions i and j, come closer to \neach other upto a small positive distance μ, where (μ/x) 0.\nProof: Consider two tokens, Tokens A and B, to be placed at t=0. If Ti > Tj , then let TA = Ti and TB = Tj , else let TA = Tj\nand TB = Ti . Now follow the procedure below (note that INT (argument) returns Integer portion of that argument):\nStep 1: Set Position_of_Token_A = TA\nStep 2: Set Position_of_Token_B = Position_of_Token_B + TB\nStep 3: Go back to Step 2, if Token B has not yet overtaken Token A\nStep 4: Set Difference_in_token_positions = ( Position_of_Token_B - Position_of_Token_A )\nStep 5: Set Integral_Multiple = 1 + INT ( (Position_of_Token_A) / (Difference_in_token_positions) )\nStep 6: Set Position_of_Token_A = ( 2 x Position_of_Token_A ) x ( Integral_Multiple )\nStep 7: Set Position_of_Token_B = Position_of_Token_B x ( Integral_Multiple )\nStep 8: Go back to Step 4\nIt will be observed that the variable Difference_in_token_positions will continue to decrease to a very small value\n(though it will never become zero), with each pass through Step 4 of the procedure. There are two reasons why this happen. \nFirst, the remainder of division is less than the divisor during the division process (see Step 5), and in the case of our \nProcedure, the kth time that Token B overtakes Token A, then the divisor for the (k+1)th overtaking is being obtained from the \nremainder (i.e. difference in Token positions) of the kth overtaking. Second, a zero remainder is prevented from occurring \nduring the division process, because it is impossible for ( (Position_of_Token_A) / (Difference_in_token_positions ) ) to be an \ninteger. The key factor which enables this happen is the fact that the ratio of periods of Functions i and j, is irrational.\nNext, as it is obvious that the positions of both Tokens A and B, represent the start of highs of Functions i or j\n(because both Tokens start from t=0, and because both Tokens are iteratively shifted forward by integral multiples of Ti and\nTj), we can conclude that at some real value of t > 0, the start of highs of Functions i and j, come closer to each other, upto a \nsmall positive distance μ, where (μ/x) 0. It is also important to note that the start of highs will never touch each other (i.e. μ\n≠ 0), because if they touch, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, where N1 and N2 ≠ 0, \nwhich is false as we proved earlier in Theorem-1. Hence Proved Lemma-3.1\nLemma-3.2: If Di = Dj = 0, then there exist an infinite number of Intervals along t-Axis, where Functions i and j, are\nsimultaneously high for a duration of atleast (x - μ), where μ is positive non-zero real and (μ/x) 0. Further, the starting \npoints of such Intervals (e.g. tA , tB in Figure 3) do not have a constant period, but their occurrence along the t-Axis can be \nexpressed by some relationship involving t, Ti , Tj, μ, and x.\nProof: We just proved in Lemma-3.1 that the start of the highs of Functions i and j, do come closer to each other, upto a small \npositive distance μ, where (μ/x) 0. Let Li denote the distance from t=0 to the start of that high of Function i, and let Lj\ndenote the distance from t=0 to the start of that high of Function j, so that abs (Li - Lj) = μ. Now starting from t=0, let us \nplace infinite Tokens of type A along the t-Axis so that two neighboring tokens of type A are separated by distance Li. \nSimilarly, starting from t=0, let us place infinite Tokens of type B along the t-Axis so that two neighboring tokens of type B \nare separated by distance Lj. Now imagine a single Token C, starting from t=0, and iteratively taking leaps of Li every unit of \ntime. It is obvious that Token C will observe a discrete relative-speed (i.e. discrete steps of μ per unit time) of an infinite\nseries of Tokens of type B. This can also be imagined as if stationary Token C observes a moving Train having infinite \ncarriages, and where the start of each carriages is represented by a Token of type B, moving in the opposite direction (if Li > \nLj) or moving in the same direction (if Li < Lj), with a discrete relative-speed (i.e. discrete steps of μ per unit time).\nAs (μ/x) 0 (which means μ is very much greater than x), it can be concluded that there exists an infinite number of \nIntervals along t-Axis, where Functions i and j, are simultaneously high for a duration of atleast (x - μ), where μ is positive \nnon-zero real and (μ/x) 0. It is important to note that the starting points of such Intervals cannot have a constant period. It is \neasy to see that if they do have a constant period, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, \nwhere N1 and N2 ≠ 0, which is false as we earlier proved. Instead, the occurrence of such Intervals along the t-Axis is \ngoverned by a very sophisticated relationship among t, Ti , Tj, μ, and x. It is beyond our scope to go into details of that \nrelationship. The existence of such a relationship is verified since the occurrence of the starting points of such Intervals is \nmore frequent, if the values of μ and x are increased. Hence Proved Lemma-3.2\nFinally, introducing delays Di and Dj, simply shifts the initial position of the infinite ‘Train’ of Tokens of type B, \neither forward or backward by abs (Di - Dj). Token C continues to observe the ‘Train’ moving with the same discrete \nrelative-speed. So the Lemma-3.1 and Lemma-3.2 continue to remain true. Hence Proved Theorem 3.\n\n2.6 Theorem-4:\nIf we consider all n Functions defined in Section 2.1, there are an infinite number of Intervals along the t-Axis, when all n\nFunctions are simultaneously high for duration greater than (x - μ), where (μ/x) 0, and μ 0. Please see figure 4 below.\nFig 4. There are an infinite number of Intervals, when all n Functions are simultaneously high for duration greater than x - μ\nProof: We shall give two angles of argument, to support the validity of this Theorem.\nAngle-1: As discussed under the proof for Theorem-3, that for any two Functions i and j, there is a sophisticated relationship\ngoverning the occurrences of the start of Intervals along the t-Axis, where during such Intervals, both Functions are \nsimultaneously high for duration greater than x - μ1, where μ1 > 0 and (μ1/x) 0. This sophisticated relationship involves \nTi , Tj, Di , Dj , μ1 , and x. Let this relationship be expressed as Gij = 0. Next, as there are infinite such Intervals, whose \nstarting points have no constant period, the degree of Gij is also infinite. Therefore when we start considering some another \nFunction k defined in Section 2.1 (apart from already considering Functions i and j), then it is impossible to express Function \nk as a rational multiple of Gij =0. Thus, the Function Gij=0, and Function k, should again have an infinite number of \nIntervals, where x - μ2, where μ2 > 0 and (μ2/x) 0. Of course, the Intervals while considering Functions i, j, and k, will\nbe less frequent along the t-Axis, than the Intervals when only the Functions i and j were considered, but there will still be an \ninfinite number of such Intervals. Similarly, as we go on considering more Functions from Section 2.1, one by one, we \nshould find Infinite such Intervals along the t-Axis, though their occurrence is with lesser and lesser frequency.\nAngle-2: Let n Trains stand next to each other on parallel tracks, with each Train having an infinite number of carriages. Let \nthe length of all the carriages of Train i is set to Ti , for all i in [1,n], where the irrational R is very small (e.g. 10-n √2).\nEach carriage has exactly one window situated at a distance of Di from the start of that carriage.\nNow imagine a window-cleaner-man on Train 1, who, after cleaning the window of a carriage, moves to clean the \nwindow of the next forward carriage of the same Train 1. Let the man keep moving forward in this way, cleaning windows of \nthe carriages of Train 1. It is obvious from Theorems 1 and 2, that every time the man looks out from a different window of\nTrain 1, the man will always perceive a different permutation of arrangements of carriages of the parallel Trains 2 to n. This \nalso means that if, for every Train, if we were to divide each carriage into M Equal parts (where M is an extremely large \nnatural number), and color the parts with M different colors (same coloring sequence is followed for the M parts of all \ncarriages of all Trains), then the following statement is true: - Irrespective of how large M is, it is always possible to draw \nMN lines orthogonal to the Trains, where each line has a different permutation of colors.\nThus, the fact that the periods of all n Periodic Functions of Section 2.1, are irrational with respect to each \nother is the most important factor, which enables us conclude that there exist infinite Intervals along the t-Axis, where all n \nFunctions are simultaneously high for a duration greater than x - μ, where μ > 0 and (μ/x) 0. Hence Proved Theorem-4.\n\n3. Our New Approach\n3.1 Basic Idea of our Model\nThe basic idea of our Model is to build a set of Equations describing the HCP, such that this set of Equations has a constant \nnumber of variables, irrespective of N, the number of nodes in the HCP. The feasible solution of each variable is bounded \nwithin the space of 2 real numbers.\n3.2 A Crucial substitution, based on Theorem-4, to keep a Constant number of variables in our Model\nFor each of the N(N-1) edge-variables in the directed Graph G, we make the following substitution for xi , j (edge-variable for \nthe edge from node i to node j, which in our Model, is allowed to have a value that is either approx 2 or approx 0):\n \n \n \n \n \n \nj\ni\nj\ni\nC\nj\nN\ni\nx\nhave\nwe\nj\ni\nand\nN\nin\nj\ni\n,\n,\n)1\n(\ncos\n1\n,\n],\n,1[\n}\n,\n{\n \n \n \n \n \n \n \n \n \nCi, j is the cost from Node i and Node j, which can be either 0 or 1, and is obtained from the specific instance of the HCP.\nThe above substitution of xi , j is performed with the trigonometric cosine function with period = 1/(1+((i-1)N + j)(β/α)),\nbecause from Theorem-4 (replacing R with β/α), 2N(N-1) permutations are possible, as explained by the following statement:\nThe highest-point/lowest-point of x1 ,1 can come ‘extremely close’ to the \n(statement starts and is contd in next line)\nthe highest-point/lowest-point of x1 ,2 , which can come ‘extremely close’ to the\n(statement contd in next line)\n...................\n(statement contd in next line)\nthe highest-point/lowest-point of xi ,j , which can come ‘extremely close’ to the\n(statement contd in next line)\n...................\n(statement contd in next line)\nthe highest-point/lowest-point of xn ,(n-1) , which can come ‘extremely close’ to the\n(statement contd in next line)\nthe highest-point/lowest-point of xn ,n .\n(statement ends)\nBy ‘extremely close’, we mean ‘within a distance that tends to 0, and which still may not yet be equal to 0’.\nin\npolynomial\na\nas\nx\nress\nto\ntry\nshall\nwe\nN\nin\nj\ni\nNow\nj\ni,\nexp\n],\n,1[\n}\n,\n{\n, \n \n \n \n \n \n \n \n \nsin\ncos\n,\nsin\n,\ncos\nand\n. \nWe know from standard Trigonometry that the following 3 Equations are true:\n \n \n \n \n \n \n \n \n \n \n \n \n \n \nk\nk\nk\nsin\nsin\ncos\ncos\ncos\n \n \n \n,and\n \n \n \n \n \n \n \n \n \n \n2\ncos\ncos\n)1\n(\ncos\n2\ncos\n \n \n \n \nk\nk\nk\n, and\n \n \n \n \n \n \n \n \n \n \n2\nsin\ncos\n)1\n(\nsin\n2\nsin\n \n \n \n \nk\nk\nk\nIteratively follow the above 3 Equations, so as to express\n \n \n \n \n \n \nk\nand\nk\nsin\ncos\n, as polynomials in \n \n \n \n \nsin\ncos\nand\n. Note that this has been discussed in further detail in [1][2]. It is important, and worthwhile to note \nthat though the co-efficients of the above polynomials have values of the form 2P(N) (where P(N) is some polynomial\nfunction of N), still these co-efficients occupy polynomial size proportional to log (2P(N)), which is P(N).\ns\npolynomial\nas\nx\nress\nto\nable\nare\nwe\nN\nin\nj\ni\nThus\nj\ni,\nexp\n],\n,1[\n}\n,\n{\n, \n \n \n \n \n \n \n \n \nsin\ncos\n,\nsin\n,\ncos\nand\nin\n.\nThe purpose of performing this substitution is to ensure that evaluating the expressions of our Model are maintained within \npolynomial complexity. For example, it is obvious that, without our substitution, the process of bringing the expression \n \n \n \n \n \n \n \n \n \n \n \n \n \nN\ni\nN\ni\nj\nj\nj\ni\nx\n1\n,\n1\n,\ninto simplest form (i.e. where all terms of the expression have only ‘+’, ‘-’, ‘x’ or ‘/’ \noperators) needs exponential space and time. Whereas, our crucial substitution enables this expression to be brought into \nsimplest form within polynomial space and time.\n\n3.3 Set of Constraints for our Algorithm (note that before bringing any of these Constraint Sets to Simplest form, \nwe should first Substitute all xi, j by the trigonometric polynomials in 4 variables, as described in Section 3.2)\n \n \n \n \n \n \n1\n1\n,1\n2\n,\n2\n,\n)1\n(\n4\n0\n2\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \nN\nN\nx\nx\nN\ni\nN\ni\nj\nj\nj\ni\nj\ni\nCONSTRAINT-SET-1\n2\n1\n2\n,1\n,\n3\n)1\n(\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \nN\ni\nN\ni\nj\nj\nj\nix\nCONSTRAINT-SET-2\n \n \n \n \n3\n2\n2\n1\n,\n2\n1\n,1\n1\n2\n \n \n \n \n \n \n \n \nN\ni\nN\ni\nN\nY\nY\nCONSTRAINT-SET-3\nWhere\n]\n,2\n[\n0\n,1\n1,\n1,1\nN\ni\nY\nY\ni\n \n \n \n \n, and\n \n \n \n \n \n \n \n \n \n \n \n \nN\ni\nj\nj\nj\ni\nt\nj\nt\ni\nt\ni\nx\nY\nY\nY\nhave\nwe\nN\nin\ni\n,1\n,\n,\n,\n1\n,\n2\n],\n,1[\n \n \n2\n2\n0\n1\nN\n \n \n \n \n \n \n2\n2\n0\n2\nN\n \n \n \n \nCONSTRAINT-SET-4\n \n \n2\n2\n0\n3\nN\n \n \n \n \n \n1\ncos\n1\n \n \n \n \n \n1\nsin\n1\n \n \n \n \n \n1\ncos\n1\n \n \n \n \n \n1\ncos\n1\n \n \n \n \nCONSTRAINT-SET-5\n \n \n1\ncos\nsin\n2\n2\n \n \n \n \n \n \n1\ncos\nsin\n2\n2\n \n \n \n \n3.4 Explanation of the Constraint Sets\nIn sub-Section 3.3, Constraint Set 5 ensures that each edge variable is within [0,2], and also validates the trigonometric \nsubstitution. Constraint Set 1 ensures that the value of each edge-variable is either approximately 2 or approximately 0 (by \nthe word ‘approximately’ we mean plus/minus a small quantity much less than 2-(NxN)). Constraint Set 2 then ensures that\n\nthere is exactly one edge-variable connected to node i, whose value is approximately 2, while all other edge-variables \nconnected to node i have values approximately 0. Once this is done, the resulting feasible solutions are either a Hamiltonian \nCycle (i.e. Cycle of length = N), or isolated rings (i.e. isolated Cycles each of length < (N-2) ). \nNext, Constraint Set 3 constrains the previously discussed solutions to only Hamiltonian Cycles (here we assume \nN>2). Constraint Set 3 is actually the simulation of a signal propagating among all the Nodes, from time or t = 1 to t = N, \nwhere t increases discretely in steps of 1. At time =1, we start with Unit signal strength at Node 1, and zero signal strength at \nall the other Nodes. At the next Unit of time, each Node receives signals from all other Nodes, along the edges, and adds it to \nwhatever signal strength it already had. Finally at t = (N+1), if Node 1 has signal strength of approximately 2, and every other \nNode has a signal strength of approximately 1, then it signifies a Hamiltonian Cycle.\nLastly, Constraint Set 4 constrains all approximation to be within 2-(NxN), so that Constraint Sets 1-3 remain valid.\nNote that we could have chosen any other number such as 2-(NxNxNxN) to constrain the approximation. The reason why we \nstuck with 2-(NxN), is because it is sufficiently small to ensure that the Model is valid, and still can be expressed within \npolynomial space of (NxN) bits. Note that a too small number like 2-(NxNxNxN.....N times) will take up exponential space.\nIn summary, our Constraint Equations are basically looking for the possibility that N edge-variables possess\nvalues approximately equal to and less than 2, whose corresponding edges form a Hamiltonian Cycle, while all other \nedge-variables possess values approximately equal to and greater than 0.\n3.5 Converting Inequations into Equations\nThe Inequations described under Section 3.3 (Constraint Sets 4 and 5) can be converted into Equations, because an \nInequation of the form \n \n \nx\n can be converted into an Equation (by the introduction of an extra variable z) of the form:\n \n \n0\n2\n \n \n \n \nx\nz\n, and then checking for the existence of solutions for both variables x and z. It is to be noted, that even the \nextra variable will also be bounded within the space of 2 real numbers.\n3.6 Final Set of Equations, and their conversion into a single Polynomial Equation\nAfter converting the Inequations to Equations, we finally have a constant number of Polynomial Equations, with a constant \nnumber of variables. It is easy to convert them into a single Polynomial Equation, by squaring and adding the Polynomial \nportions of the Equations, and then equating the result to zero.\n4. Conclusion, and Future Work\nIn this paper, we presented a polynomial Approach for expressing the decision version of the directed Hamiltonian Cycle \nProblem (NP-Complete) or HCP, in terms of the Solvability of an Equation. Using a crucial trigonometric substitution that \nenables the N(N-1) edge-variables to be expressed as polynomials in 4 variables, our Approach describes the HCP with a \nsingle Polynomial Equation with a constant number of variables, where the feasible solution of each variable is bounded \nwithin a set of 2 real numbers. It is hoped that our approach will be of use, in mankind’s P? =NP Question.\nWe hope someone in the world can develop EITHER:\nTechnique-1: A Technique that determines whether or not a Polynomial Equation (with a constant number of variables) is \nsolvable within the Infinite unbounded Real Set, OR\nTechnique-2: A Technique that determines whether or not a Polynomial Equation (with a constant number of variables) is \nsolvable within a bounded Real Set (i.e. where the feasible set of each variable is bounded within a space of 2 real numbers)\nWe could use either of these Techniques for our HCP’s Polynomial Equation (Technique-1 will also work for our \nModel, because we have explicit constraints limiting our variables within the feasible bounded Real Set). So we could use \neither of the Techniques to determine whether, for a particular HCP instance, if there does exist atleast one solution point for \nour variables (i.e. sin(α), cos(α), sin(β), cos(β), ε1 , ε2 , ε3, and the other variables introduced by converting Inequations \nto Equations), then we can conclude that the HCP instance does have a Hamiltonian Cycle. And if their Technique finds that \nno such solution is possible, then we can conclude that the HCP instance does not have a Hamiltonian Cycle.\nSo mankind will be able to prove that P=NP, if either Technique-1 or Technique-2 is developed.\nReferences\n[1] http://mcraefamily.com/MathHelp/GeometryTrigEquivCos3xEtcB.htm\n[2] http://mcraefamily.com/MathHelp/GeometryTrigEquivSin3xEtcC.htm\nAcknowledgement\nI am grateful to Dr. David Moews (djm.cc/dmoews.html), who is an IMO Gold Medallist, and three time Putnam Prize Winner with a \nperfect score in the 1987 difficult one, for his enlightening discussions, and also for pointing out a crucial error in a previous version of my \npaper. I asked Dr. Moews whether I could co-author him in this paper, but he has not responded yet.","paragraphs":[{"paragraph_id":"p1","order":1,"text":"Expressing an NP-Complete Problem as the Solvability of a Polynomial Equation\nDeepak Ponvel Chermakani, IEEE Member\ndeepakc@pmail.ntu.edu.sg deepakc@usa.com deepakc@myfastmail.com deepak.chermakani@ustri.com\nAbstract: We demonstrate a polynomial approach to express the decision version of the directed Hamiltonian Cycle \nProblem (HCP), which is NP-Complete, as the Solvability of a Polynomial Equation with a constant number of \nvariables, within a bounded real space. We first introduce four new Theorems for a set of periodic Functions with \nirrational periods, based on which we then use a trigonometric substitution, to show how the HCP can be expressed as \nthe Solvability of a single polynomial Equation with a constant number of variables. The feasible solution of each of \nthese variables is bounded within two real numbers. We point out what future work is necessary to prove that P=NP.\n1. Introduction\nGiven a graph G (V, E), where V is the set of N vertices, and E is the set of edge-costs that are either 1 or 0, the decision \nversion of the HCP is to find out whether there exists a round trip tour to visit each vertex exactly once, such that the total \ncost of the tour equals N. If the edge-cost in traveling from one edge to another, is the same as the reverse cost, then the HCP \nis said to be undirected, while if it is either same or different than the reverse cost, then the HCP is said to be directed. Both \nthe directed, as well as the undirected forms of the decision version of the HCP, are known to be NP-Complete.\nIn the rest of this paper, we shall refer to “decision version of the directed HCP”, simply as the “HCP”. Our paper \ndescribes a new polynomial Approach for expressing the HCP as the solvability of an Equation. Our Approach is discussed in \nSection-3 and is based on new Theorems, which we will introduce in the next section (Section-2) of this paper.\n2. Theorems on a Set of Periodic Functions with Irrational Periods\n2.1 Definition of the Periodic Functions\nLet us define n periodic Functions to have the following three properties.\nProperty-1: Period of Function i, Ti = (1 / (1+ iR) ), where R is irrational non-zero real, and i represents integers 1 to n\nProperty-2: All the n Functions are binary functions and have a binary high value for the same duration x, where x is non-\nzero, real, positive, and lesser than the smallest period, which is 1 / (1+ nR).\nProperty-3: Function i is allowed to start to become high from time t=0, with real delay Di , where 0 ≤ Di < Ti\nFor better clarity, we attempt to show these three properties of the n Functions, in Figure-1 below. Note that the \nvariable t along the t-Axis, varies from 0 to positive infinity.\nFig 1. Definition of the Periodic Functions"},{"paragraph_id":"p2","order":2,"text":"2.2 Definition of E(t)i\nWe define E(t)i as the closest positive displacement from t when Function i again starts to become high. For the purpose of \nbetter clarity, we show this definition in Figure 2 below.\nFig 2. Definition of E(t)i , for Function i\n2.3 Theorem-1:\nIf we consider any two Functions i and j from the n Functions defined in Section 2.1, then every real value of t in [0, ∞[ has a \nunique set {E(t)i , E(t)j }. In other words, {E(tA)i , E(tA)j } ≠ {E(tB)i , E(tB)j }, if tA ≠ tB.\nProof: If {E(tA)i , E(tA)j } = {E(tB)i , E(tB)j }, for say tA < tB, then it would mean that it is possible to identify another \npoint tC where {E(tB)i , E(tB)j } = {E(tC)i , E(tC)j }, and where tC – tB = tB – tA. It would also mean that an infinite \nnumber of similar points like tC can be periodically found along the t-Axis. This, in turn, would imply that N1 Ti = N2 Tj, \nfor some natural numbers N1 and N2, where N1 and N2 ≠ 0. But we know that (Ti / Tj ) = (1+ jR) / (1+ iR), which is \nirrational and therefore cannot be expressed as (N2 / N1). Hence Proved Theorem-1.\n2.4 Theorem-2:\nIf we consider all the n Functions defined in Section 2.1, then every real value of t in [0, ∞[ has a unique \nset {E(t)1 , E(t)2 , E(t)3 ,......., E(t)n-1 , E(t)n }\nProof: This is straightforward from Theorem-1, and it is easy to see that all the periods of the n Functions are irrational with \nrespect to each other. Hence Proved Theorem-2.\n2.5 Theorem-3:\nIf we consider any two Functions i and j from the n Functions defined in Section 2.1, then the following two statements are \nTrue (also see Figure 3):\ni) There exists an infinite number of Intervals along t-Axis, where these two functions are simultaneously high for a \nduration of atleast (x - μ), where μ is positive non-zero real and (μ/x) 0.\nii)The starting points of such Intervals (e.g. tA , tB in Figure 3) do not have a constant period, but their occurrence \nalong the t-Axis can be expressed by some relationship involving Ti , Tj, Di , Dj, μ, and x.\nFig 3. There are infinite Intervals along t-Axis, when the 2 Functions are simultaneously high for duration of atleast x - μ"},{"paragraph_id":"p3","order":3,"text":"Proof: We will first prove this by assuming zero-delays for Functions i and j. We will then incorporate delays, and show that \nthe proof holds. We define a ‘Token’, as a movable point used to indicate the current location of a process along the t-Axis.\nLemma-3.1: If Di = Dj = 0, there exists some real t > 0, where the start of the highs of Functions i and j, come closer to \neach other upto a small positive distance μ, where (μ/x) 0.\nProof: Consider two tokens, Tokens A and B, to be placed at t=0. If Ti > Tj , then let TA = Ti and TB = Tj , else let TA = Tj\nand TB = Ti . Now follow the procedure below (note that INT (argument) returns Integer portion of that argument):\nStep 1: Set Position_of_Token_A = TA\nStep 2: Set Position_of_Token_B = Position_of_Token_B + TB\nStep 3: Go back to Step 2, if Token B has not yet overtaken Token A\nStep 4: Set Difference_in_token_positions = ( Position_of_Token_B - Position_of_Token_A )\nStep 5: Set Integral_Multiple = 1 + INT ( (Position_of_Token_A) / (Difference_in_token_positions) )\nStep 6: Set Position_of_Token_A = ( 2 x Position_of_Token_A ) x ( Integral_Multiple )\nStep 7: Set Position_of_Token_B = Position_of_Token_B x ( Integral_Multiple )\nStep 8: Go back to Step 4\nIt will be observed that the variable Difference_in_token_positions will continue to decrease to a very small value\n(though it will never become zero), with each pass through Step 4 of the procedure. There are two reasons why this happen. \nFirst, the remainder of division is less than the divisor during the division process (see Step 5), and in the case of our \nProcedure, the kth time that Token B overtakes Token A, then the divisor for the (k+1)th overtaking is being obtained from the \nremainder (i.e. difference in Token positions) of the kth overtaking. Second, a zero remainder is prevented from occurring \nduring the division process, because it is impossible for ( (Position_of_Token_A) / (Difference_in_token_positions ) ) to be an \ninteger. The key factor which enables this happen is the fact that the ratio of periods of Functions i and j, is irrational.\nNext, as it is obvious that the positions of both Tokens A and B, represent the start of highs of Functions i or j\n(because both Tokens start from t=0, and because both Tokens are iteratively shifted forward by integral multiples of Ti and\nTj), we can conclude that at some real value of t > 0, the start of highs of Functions i and j, come closer to each other, upto a \nsmall positive distance μ, where (μ/x) 0. It is also important to note that the start of highs will never touch each other (i.e. μ\n≠ 0), because if they touch, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, where N1 and N2 ≠ 0, \nwhich is false as we proved earlier in Theorem-1. Hence Proved Lemma-3.1\nLemma-3.2: If Di = Dj = 0, then there exist an infinite number of Intervals along t-Axis, where Functions i and j, are\nsimultaneously high for a duration of atleast (x - μ), where μ is positive non-zero real and (μ/x) 0. Further, the starting \npoints of such Intervals (e.g. tA , tB in Figure 3) do not have a constant period, but their occurrence along the t-Axis can be \nexpressed by some relationship involving t, Ti , Tj, μ, and x.\nProof: We just proved in Lemma-3.1 that the start of the highs of Functions i and j, do come closer to each other, upto a small \npositive distance μ, where (μ/x) 0. Let Li denote the distance from t=0 to the start of that high of Function i, and let Lj\ndenote the distance from t=0 to the start of that high of Function j, so that abs (Li - Lj) = μ. Now starting from t=0, let us \nplace infinite Tokens of type A along the t-Axis so that two neighboring tokens of type A are separated by distance Li. \nSimilarly, starting from t=0, let us place infinite Tokens of type B along the t-Axis so that two neighboring tokens of type B \nare separated by distance Lj. Now imagine a single Token C, starting from t=0, and iteratively taking leaps of Li every unit of \ntime. It is obvious that Token C will observe a discrete relative-speed (i.e. discrete steps of μ per unit time) of an infinite\nseries of Tokens of type B. This can also be imagined as if stationary Token C observes a moving Train having infinite \ncarriages, and where the start of each carriages is represented by a Token of type B, moving in the opposite direction (if Li > \nLj) or moving in the same direction (if Li < Lj), with a discrete relative-speed (i.e. discrete steps of μ per unit time).\nAs (μ/x) 0 (which means μ is very much greater than x), it can be concluded that there exists an infinite number of \nIntervals along t-Axis, where Functions i and j, are simultaneously high for a duration of atleast (x - μ), where μ is positive \nnon-zero real and (μ/x) 0. It is important to note that the starting points of such Intervals cannot have a constant period. It is \neasy to see that if they do have a constant period, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, \nwhere N1 and N2 ≠ 0, which is false as we earlier proved. Instead, the occurrence of such Intervals along the t-Axis is \ngoverned by a very sophisticated relationship among t, Ti , Tj, μ, and x. It is beyond our scope to go into details of that \nrelationship. The existence of such a relationship is verified since the occurrence of the starting points of such Intervals is \nmore frequent, if the values of μ and x are increased. Hence Proved Lemma-3.2\nFinally, introducing delays Di and Dj, simply shifts the initial position of the infinite ‘Train’ of Tokens of type B, \neither forward or backward by abs (Di - Dj). Token C continues to observe the ‘Train’ moving with the same discrete \nrelative-speed. So the Lemma-3.1 and Lemma-3.2 continue to remain true. Hence Proved Theorem 3."},{"paragraph_id":"p4","order":4,"text":"2.6 Theorem-4:\nIf we consider all n Functions defined in Section 2.1, there are an infinite number of Intervals along the t-Axis, when all n\nFunctions are simultaneously high for duration greater than (x - μ), where (μ/x) 0, and μ 0. Please see figure 4 below.\nFig 4. There are an infinite number of Intervals, when all n Functions are simultaneously high for duration greater than x - μ\nProof: We shall give two angles of argument, to support the validity of this Theorem.\nAngle-1: As discussed under the proof for Theorem-3, that for any two Functions i and j, there is a sophisticated relationship\ngoverning the occurrences of the start of Intervals along the t-Axis, where during such Intervals, both Functions are \nsimultaneously high for duration greater than x - μ1, where μ1 > 0 and (μ1/x) 0. This sophisticated relationship involves \nTi , Tj, Di , Dj , μ1 , and x. Let this relationship be expressed as Gij = 0. Next, as there are infinite such Intervals, whose \nstarting points have no constant period, the degree of Gij is also infinite. Therefore when we start considering some another \nFunction k defined in Section 2.1 (apart from already considering Functions i and j), then it is impossible to express Function \nk as a rational multiple of Gij =0. Thus, the Function Gij=0, and Function k, should again have an infinite number of \nIntervals, where x - μ2, where μ2 > 0 and (μ2/x) 0. Of course, the Intervals while considering Functions i, j, and k, will\nbe less frequent along the t-Axis, than the Intervals when only the Functions i and j were considered, but there will still be an \ninfinite number of such Intervals. Similarly, as we go on considering more Functions from Section 2.1, one by one, we \nshould find Infinite such Intervals along the t-Axis, though their occurrence is with lesser and lesser frequency.\nAngle-2: Let n Trains stand next to each other on parallel tracks, with each Train having an infinite number of carriages. Let \nthe length of all the carriages of Train i is set to Ti , for all i in [1,n], where the irrational R is very small (e.g. 10-n √2).\nEach carriage has exactly one window situated at a distance of Di from the start of that carriage.\nNow imagine a window-cleaner-man on Train 1, who, after cleaning the window of a carriage, moves to clean the \nwindow of the next forward carriage of the same Train 1. Let the man keep moving forward in this way, cleaning windows of \nthe carriages of Train 1. It is obvious from Theorems 1 and 2, that every time the man looks out from a different window of\nTrain 1, the man will always perceive a different permutation of arrangements of carriages of the parallel Trains 2 to n. This \nalso means that if, for every Train, if we were to divide each carriage into M Equal parts (where M is an extremely large \nnatural number), and color the parts with M different colors (same coloring sequence is followed for the M parts of all \ncarriages of all Trains), then the following statement is true: - Irrespective of how large M is, it is always possible to draw \nMN lines orthogonal to the Trains, where each line has a different permutation of colors.\nThus, the fact that the periods of all n Periodic Functions of Section 2.1, are irrational with respect to each \nother is the most important factor, which enables us conclude that there exist infinite Intervals along the t-Axis, where all n \nFunctions are simultaneously high for a duration greater than x - μ, where μ > 0 and (μ/x) 0. Hence Proved Theorem-4."},{"paragraph_id":"p5","order":5,"text":"3. Our New Approach\n3.1 Basic Idea of our Model\nThe basic idea of our Model is to build a set of Equations describing the HCP, such that this set of Equations has a constant \nnumber of variables, irrespective of N, the number of nodes in the HCP. The feasible solution of each variable is bounded \nwithin the space of 2 real numbers.\n3.2 A Crucial substitution, based on Theorem-4, to keep a Constant number of variables in our Model\nFor each of the N(N-1) edge-variables in the directed Graph G, we make the following substitution for xi , j (edge-variable for \nthe edge from node i to node j, which in our Model, is allowed to have a value that is either approx 2 or approx 0):"},{"paragraph_id":"p6","order":6,"text":"j\ni\nj\ni\nC\nj\nN\ni\nx\nhave\nwe\nj\ni\nand\nN\nin\nj\ni\n,\n,\n)1\n(\ncos\n1\n,\n],\n,1[\n}\n,\n{"},{"paragraph_id":"p7","order":7,"text":"Ci, j is the cost from Node i and Node j, which can be either 0 or 1, and is obtained from the specific instance of the HCP.\nThe above substitution of xi , j is performed with the trigonometric cosine function with period = 1/(1+((i-1)N + j)(β/α)),\nbecause from Theorem-4 (replacing R with β/α), 2N(N-1) permutations are possible, as explained by the following statement:\nThe highest-point/lowest-point of x1 ,1 can come ‘extremely close’ to the \n(statement starts and is contd in next line)\nthe highest-point/lowest-point of x1 ,2 , which can come ‘extremely close’ to the\n(statement contd in next line)\n...................\n(statement contd in next line)\nthe highest-point/lowest-point of xi ,j , which can come ‘extremely close’ to the\n(statement contd in next line)\n...................\n(statement contd in next line)\nthe highest-point/lowest-point of xn ,(n-1) , which can come ‘extremely close’ to the\n(statement contd in next line)\nthe highest-point/lowest-point of xn ,n .\n(statement ends)\nBy ‘extremely close’, we mean ‘within a distance that tends to 0, and which still may not yet be equal to 0’.\nin\npolynomial\na\nas\nx\nress\nto\ntry\nshall\nwe\nN\nin\nj\ni\nNow\nj\ni,\nexp\n],\n,1[\n}\n,\n{\n,"},{"paragraph_id":"p8","order":8,"text":"sin\ncos\n,\nsin\n,\ncos\nand\n. \nWe know from standard Trigonometry that the following 3 Equations are true:"},{"paragraph_id":"p9","order":9,"text":"k\nk\nk\nsin\nsin\ncos\ncos\ncos"},{"paragraph_id":"p10","order":10,"text":",and"},{"paragraph_id":"p11","order":11,"text":"2\ncos\ncos\n)1\n(\ncos\n2\ncos"},{"paragraph_id":"p12","order":12,"text":"k\nk\nk\n, and"},{"paragraph_id":"p13","order":13,"text":"2\nsin\ncos\n)1\n(\nsin\n2\nsin"},{"paragraph_id":"p14","order":14,"text":"k\nk\nk\nIteratively follow the above 3 Equations, so as to express"},{"paragraph_id":"p15","order":15,"text":"k\nand\nk\nsin\ncos\n, as polynomials in"},{"paragraph_id":"p16","order":16,"text":"sin\ncos\nand\n. Note that this has been discussed in further detail in [1][2]. It is important, and worthwhile to note \nthat though the co-efficients of the above polynomials have values of the form 2P(N) (where P(N) is some polynomial\nfunction of N), still these co-efficients occupy polynomial size proportional to log (2P(N)), which is P(N).\ns\npolynomial\nas\nx\nress\nto\nable\nare\nwe\nN\nin\nj\ni\nThus\nj\ni,\nexp\n],\n,1[\n}\n,\n{\n,"},{"paragraph_id":"p17","order":17,"text":"sin\ncos\n,\nsin\n,\ncos\nand\nin\n.\nThe purpose of performing this substitution is to ensure that evaluating the expressions of our Model are maintained within \npolynomial complexity. For example, it is obvious that, without our substitution, the process of bringing the expression"},{"paragraph_id":"p18","order":18,"text":"N\ni\nN\ni\nj\nj\nj\ni\nx\n1\n,\n1\n,\ninto simplest form (i.e. where all terms of the expression have only ‘+’, ‘-’, ‘x’ or ‘/’ \noperators) needs exponential space and time. Whereas, our crucial substitution enables this expression to be brought into \nsimplest form within polynomial space and time."},{"paragraph_id":"p19","order":19,"text":"3.3 Set of Constraints for our Algorithm (note that before bringing any of these Constraint Sets to Simplest form, \nwe should first Substitute all xi, j by the trigonometric polynomials in 4 variables, as described in Section 3.2)"},{"paragraph_id":"p20","order":20,"text":"1\n1\n,1\n2\n,\n2\n,\n)1\n(\n4\n0\n2"},{"paragraph_id":"p21","order":21,"text":"N\nN\nx\nx\nN\ni\nN\ni\nj\nj\nj\ni\nj\ni\nCONSTRAINT-SET-1\n2\n1\n2\n,1\n,\n3\n)1\n("},{"paragraph_id":"p22","order":22,"text":"N\ni\nN\ni\nj\nj\nj\nix\nCONSTRAINT-SET-2"},{"paragraph_id":"p23","order":23,"text":"3\n2\n2\n1\n,\n2\n1\n,1\n1\n2"},{"paragraph_id":"p24","order":24,"text":"N\ni\nN\ni\nN\nY\nY\nCONSTRAINT-SET-3\nWhere\n]\n,2\n[\n0\n,1\n1,\n1,1\nN\ni\nY\nY\ni"},{"paragraph_id":"p25","order":25,"text":", and"},{"paragraph_id":"p26","order":26,"text":"N\ni\nj\nj\nj\ni\nt\nj\nt\ni\nt\ni\nx\nY\nY\nY\nhave\nwe\nN\nin\ni\n,1\n,\n,\n,\n1\n,\n2\n],\n,1["},{"paragraph_id":"p27","order":27,"text":"2\n2\n0\n1\nN"},{"paragraph_id":"p28","order":28,"text":"2\n2\n0\n2\nN"},{"paragraph_id":"p29","order":29,"text":"CONSTRAINT-SET-4"},{"paragraph_id":"p30","order":30,"text":"2\n2\n0\n3\nN"},{"paragraph_id":"p31","order":31,"text":"1\ncos\n1"},{"paragraph_id":"p32","order":32,"text":"1\nsin\n1"},{"paragraph_id":"p33","order":33,"text":"1\ncos\n1"},{"paragraph_id":"p34","order":34,"text":"1\ncos\n1"},{"paragraph_id":"p35","order":35,"text":"CONSTRAINT-SET-5"},{"paragraph_id":"p36","order":36,"text":"1\ncos\nsin\n2\n2"},{"paragraph_id":"p37","order":37,"text":"1\ncos\nsin\n2\n2"},{"paragraph_id":"p38","order":38,"text":"3.4 Explanation of the Constraint Sets\nIn sub-Section 3.3, Constraint Set 5 ensures that each edge variable is within [0,2], and also validates the trigonometric \nsubstitution. Constraint Set 1 ensures that the value of each edge-variable is either approximately 2 or approximately 0 (by \nthe word ‘approximately’ we mean plus/minus a small quantity much less than 2-(NxN)). Constraint Set 2 then ensures that"},{"paragraph_id":"p39","order":39,"text":"there is exactly one edge-variable connected to node i, whose value is approximately 2, while all other edge-variables \nconnected to node i have values approximately 0. Once this is done, the resulting feasible solutions are either a Hamiltonian \nCycle (i.e. Cycle of length = N), or isolated rings (i.e. isolated Cycles each of length < (N-2) ). \nNext, Constraint Set 3 constrains the previously discussed solutions to only Hamiltonian Cycles (here we assume \nN>2). Constraint Set 3 is actually the simulation of a signal propagating among all the Nodes, from time or t = 1 to t = N, \nwhere t increases discretely in steps of 1. At time =1, we start with Unit signal strength at Node 1, and zero signal strength at \nall the other Nodes. At the next Unit of time, each Node receives signals from all other Nodes, along the edges, and adds it to \nwhatever signal strength it already had. Finally at t = (N+1), if Node 1 has signal strength of approximately 2, and every other \nNode has a signal strength of approximately 1, then it signifies a Hamiltonian Cycle.\nLastly, Constraint Set 4 constrains all approximation to be within 2-(NxN), so that Constraint Sets 1-3 remain valid.\nNote that we could have chosen any other number such as 2-(NxNxNxN) to constrain the approximation. The reason why we \nstuck with 2-(NxN), is because it is sufficiently small to ensure that the Model is valid, and still can be expressed within \npolynomial space of (NxN) bits. Note that a too small number like 2-(NxNxNxN.....N times) will take up exponential space.\nIn summary, our Constraint Equations are basically looking for the possibility that N edge-variables possess\nvalues approximately equal to and less than 2, whose corresponding edges form a Hamiltonian Cycle, while all other \nedge-variables possess values approximately equal to and greater than 0.\n3.5 Converting Inequations into Equations\nThe Inequations described under Section 3.3 (Constraint Sets 4 and 5) can be converted into Equations, because an \nInequation of the form"},{"paragraph_id":"p40","order":40,"text":"x\n can be converted into an Equation (by the introduction of an extra variable z) of the form:"},{"paragraph_id":"p41","order":41,"text":"0\n2"},{"paragraph_id":"p42","order":42,"text":"x\nz\n, and then checking for the existence of solutions for both variables x and z. It is to be noted, that even the \nextra variable will also be bounded within the space of 2 real numbers.\n3.6 Final Set of Equations, and their conversion into a single Polynomial Equation\nAfter converting the Inequations to Equations, we finally have a constant number of Polynomial Equations, with a constant \nnumber of variables. It is easy to convert them into a single Polynomial Equation, by squaring and adding the Polynomial \nportions of the Equations, and then equating the result to zero.\n4. Conclusion, and Future Work\nIn this paper, we presented a polynomial Approach for expressing the decision version of the directed Hamiltonian Cycle \nProblem (NP-Complete) or HCP, in terms of the Solvability of an Equation. Using a crucial trigonometric substitution that \nenables the N(N-1) edge-variables to be expressed as polynomials in 4 variables, our Approach describes the HCP with a \nsingle Polynomial Equation with a constant number of variables, where the feasible solution of each variable is bounded \nwithin a set of 2 real numbers. It is hoped that our approach will be of use, in mankind’s P? =NP Question.\nWe hope someone in the world can develop EITHER:\nTechnique-1: A Technique that determines whether or not a Polynomial Equation (with a constant number of variables) is \nsolvable within the Infinite unbounded Real Set, OR\nTechnique-2: A Technique that determines whether or not a Polynomial Equation (with a constant number of variables) is \nsolvable within a bounded Real Set (i.e. where the feasible set of each variable is bounded within a space of 2 real numbers)\nWe could use either of these Techniques for our HCP’s Polynomial Equation (Technique-1 will also work for our \nModel, because we have explicit constraints limiting our variables within the feasible bounded Real Set). So we could use \neither of the Techniques to determine whether, for a particular HCP instance, if there does exist atleast one solution point for \nour variables (i.e. sin(α), cos(α), sin(β), cos(β), ε1 , ε2 , ε3, and the other variables introduced by converting Inequations \nto Equations), then we can conclude that the HCP instance does have a Hamiltonian Cycle. And if their Technique finds that \nno such solution is possible, then we can conclude that the HCP instance does not have a Hamiltonian Cycle.\nSo mankind will be able to prove that P=NP, if either Technique-1 or Technique-2 is developed.\nReferences\n[1] http://mcraefamily.com/MathHelp/GeometryTrigEquivCos3xEtcB.htm\n[2] http://mcraefamily.com/MathHelp/GeometryTrigEquivSin3xEtcC.htm\nAcknowledgement\nI am grateful to Dr. David Moews (djm.cc/dmoews.html), who is an IMO Gold Medallist, and three time Putnam Prize Winner with a \nperfect score in the 1987 difficult one, for his enlightening discussions, and also for pointing out a crucial error in a previous version of my \npaper. I asked Dr. Moews whether I could co-author him in this paper, but he has not responded yet."}],"pages":[{"page":1,"text":"Expressing an NP-Complete Problem as the Solvability of a Polynomial Equation\nDeepak Ponvel Chermakani, IEEE Member\ndeepakc@pmail.ntu.edu.sg deepakc@usa.com deepakc@myfastmail.com deepak.chermakani@ustri.com\nAbstract: We demonstrate a polynomial approach to express the decision version of the directed Hamiltonian Cycle \nProblem (HCP), which is NP-Complete, as the Solvability of a Polynomial Equation with a constant number of \nvariables, within a bounded real space. We first introduce four new Theorems for a set of periodic Functions with \nirrational periods, based on which we then use a trigonometric substitution, to show how the HCP can be expressed as \nthe Solvability of a single polynomial Equation with a constant number of variables. The feasible solution of each of \nthese variables is bounded within two real numbers. We point out what future work is necessary to prove that P=NP.\n1. Introduction\nGiven a graph G (V, E), where V is the set of N vertices, and E is the set of edge-costs that are either 1 or 0, the decision \nversion of the HCP is to find out whether there exists a round trip tour to visit each vertex exactly once, such that the total \ncost of the tour equals N. If the edge-cost in traveling from one edge to another, is the same as the reverse cost, then the HCP \nis said to be undirected, while if it is either same or different than the reverse cost, then the HCP is said to be directed. Both \nthe directed, as well as the undirected forms of the decision version of the HCP, are known to be NP-Complete.\nIn the rest of this paper, we shall refer to “decision version of the directed HCP”, simply as the “HCP”. Our paper \ndescribes a new polynomial Approach for expressing the HCP as the solvability of an Equation. Our Approach is discussed in \nSection-3 and is based on new Theorems, which we will introduce in the next section (Section-2) of this paper.\n2. Theorems on a Set of Periodic Functions with Irrational Periods\n2.1 Definition of the Periodic Functions\nLet us define n periodic Functions to have the following three properties.\nProperty-1: Period of Function i, Ti = (1 / (1+ iR) ), where R is irrational non-zero real, and i represents integers 1 to n\nProperty-2: All the n Functions are binary functions and have a binary high value for the same duration x, where x is non-\nzero, real, positive, and lesser than the smallest period, which is 1 / (1+ nR).\nProperty-3: Function i is allowed to start to become high from time t=0, with real delay Di , where 0 ≤ Di < Ti\nFor better clarity, we attempt to show these three properties of the n Functions, in Figure-1 below. Note that the \nvariable t along the t-Axis, varies from 0 to positive infinity.\nFig 1. Definition of the Periodic Functions"},{"page":2,"text":"2.2 Definition of E(t)i\nWe define E(t)i as the closest positive displacement from t when Function i again starts to become high. For the purpose of \nbetter clarity, we show this definition in Figure 2 below.\nFig 2. Definition of E(t)i , for Function i\n2.3 Theorem-1:\nIf we consider any two Functions i and j from the n Functions defined in Section 2.1, then every real value of t in [0, ∞[ has a \nunique set {E(t)i , E(t)j }. In other words, {E(tA)i , E(tA)j } ≠ {E(tB)i , E(tB)j }, if tA ≠ tB.\nProof: If {E(tA)i , E(tA)j } = {E(tB)i , E(tB)j }, for say tA < tB, then it would mean that it is possible to identify another \npoint tC where {E(tB)i , E(tB)j } = {E(tC)i , E(tC)j }, and where tC – tB = tB – tA. It would also mean that an infinite \nnumber of similar points like tC can be periodically found along the t-Axis. This, in turn, would imply that N1 Ti = N2 Tj, \nfor some natural numbers N1 and N2, where N1 and N2 ≠ 0. But we know that (Ti / Tj ) = (1+ jR) / (1+ iR), which is \nirrational and therefore cannot be expressed as (N2 / N1). Hence Proved Theorem-1.\n2.4 Theorem-2:\nIf we consider all the n Functions defined in Section 2.1, then every real value of t in [0, ∞[ has a unique \nset {E(t)1 , E(t)2 , E(t)3 ,......., E(t)n-1 , E(t)n }\nProof: This is straightforward from Theorem-1, and it is easy to see that all the periods of the n Functions are irrational with \nrespect to each other. Hence Proved Theorem-2.\n2.5 Theorem-3:\nIf we consider any two Functions i and j from the n Functions defined in Section 2.1, then the following two statements are \nTrue (also see Figure 3):\ni) There exists an infinite number of Intervals along t-Axis, where these two functions are simultaneously high for a \nduration of atleast (x - μ), where μ is positive non-zero real and (μ/x) 0.\nii)The starting points of such Intervals (e.g. tA , tB in Figure 3) do not have a constant period, but their occurrence \nalong the t-Axis can be expressed by some relationship involving Ti , Tj, Di , Dj, μ, and x.\nFig 3. There are infinite Intervals along t-Axis, when the 2 Functions are simultaneously high for duration of atleast x - μ"},{"page":3,"text":"Proof: We will first prove this by assuming zero-delays for Functions i and j. We will then incorporate delays, and show that \nthe proof holds. We define a ‘Token’, as a movable point used to indicate the current location of a process along the t-Axis.\nLemma-3.1: If Di = Dj = 0, there exists some real t > 0, where the start of the highs of Functions i and j, come closer to \neach other upto a small positive distance μ, where (μ/x) 0.\nProof: Consider two tokens, Tokens A and B, to be placed at t=0. If Ti > Tj , then let TA = Ti and TB = Tj , else let TA = Tj\nand TB = Ti . Now follow the procedure below (note that INT (argument) returns Integer portion of that argument):\nStep 1: Set Position_of_Token_A = TA\nStep 2: Set Position_of_Token_B = Position_of_Token_B + TB\nStep 3: Go back to Step 2, if Token B has not yet overtaken Token A\nStep 4: Set Difference_in_token_positions = ( Position_of_Token_B - Position_of_Token_A )\nStep 5: Set Integral_Multiple = 1 + INT ( (Position_of_Token_A) / (Difference_in_token_positions) )\nStep 6: Set Position_of_Token_A = ( 2 x Position_of_Token_A ) x ( Integral_Multiple )\nStep 7: Set Position_of_Token_B = Position_of_Token_B x ( Integral_Multiple )\nStep 8: Go back to Step 4\nIt will be observed that the variable Difference_in_token_positions will continue to decrease to a very small value\n(though it will never become zero), with each pass through Step 4 of the procedure. There are two reasons why this happen. \nFirst, the remainder of division is less than the divisor during the division process (see Step 5), and in the case of our \nProcedure, the kth time that Token B overtakes Token A, then the divisor for the (k+1)th overtaking is being obtained from the \nremainder (i.e. difference in Token positions) of the kth overtaking. Second, a zero remainder is prevented from occurring \nduring the division process, because it is impossible for ( (Position_of_Token_A) / (Difference_in_token_positions ) ) to be an \ninteger. The key factor which enables this happen is the fact that the ratio of periods of Functions i and j, is irrational.\nNext, as it is obvious that the positions of both Tokens A and B, represent the start of highs of Functions i or j\n(because both Tokens start from t=0, and because both Tokens are iteratively shifted forward by integral multiples of Ti and\nTj), we can conclude that at some real value of t > 0, the start of highs of Functions i and j, come closer to each other, upto a \nsmall positive distance μ, where (μ/x) 0. It is also important to note that the start of highs will never touch each other (i.e. μ\n≠ 0), because if they touch, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, where N1 and N2 ≠ 0, \nwhich is false as we proved earlier in Theorem-1. Hence Proved Lemma-3.1\nLemma-3.2: If Di = Dj = 0, then there exist an infinite number of Intervals along t-Axis, where Functions i and j, are\nsimultaneously high for a duration of atleast (x - μ), where μ is positive non-zero real and (μ/x) 0. Further, the starting \npoints of such Intervals (e.g. tA , tB in Figure 3) do not have a constant period, but their occurrence along the t-Axis can be \nexpressed by some relationship involving t, Ti , Tj, μ, and x.\nProof: We just proved in Lemma-3.1 that the start of the highs of Functions i and j, do come closer to each other, upto a small \npositive distance μ, where (μ/x) 0. Let Li denote the distance from t=0 to the start of that high of Function i, and let Lj\ndenote the distance from t=0 to the start of that high of Function j, so that abs (Li - Lj) = μ. Now starting from t=0, let us \nplace infinite Tokens of type A along the t-Axis so that two neighboring tokens of type A are separated by distance Li. \nSimilarly, starting from t=0, let us place infinite Tokens of type B along the t-Axis so that two neighboring tokens of type B \nare separated by distance Lj. Now imagine a single Token C, starting from t=0, and iteratively taking leaps of Li every unit of \ntime. It is obvious that Token C will observe a discrete relative-speed (i.e. discrete steps of μ per unit time) of an infinite\nseries of Tokens of type B. This can also be imagined as if stationary Token C observes a moving Train having infinite \ncarriages, and where the start of each carriages is represented by a Token of type B, moving in the opposite direction (if Li > \nLj) or moving in the same direction (if Li < Lj), with a discrete relative-speed (i.e. discrete steps of μ per unit time).\nAs (μ/x) 0 (which means μ is very much greater than x), it can be concluded that there exists an infinite number of \nIntervals along t-Axis, where Functions i and j, are simultaneously high for a duration of atleast (x - μ), where μ is positive \nnon-zero real and (μ/x) 0. It is important to note that the starting points of such Intervals cannot have a constant period. It is \neasy to see that if they do have a constant period, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, \nwhere N1 and N2 ≠ 0, which is false as we earlier proved. Instead, the occurrence of such Intervals along the t-Axis is \ngoverned by a very sophisticated relationship among t, Ti , Tj, μ, and x. It is beyond our scope to go into details of that \nrelationship. The existence of such a relationship is verified since the occurrence of the starting points of such Intervals is \nmore frequent, if the values of μ and x are increased. Hence Proved Lemma-3.2\nFinally, introducing delays Di and Dj, simply shifts the initial position of the infinite ‘Train’ of Tokens of type B, \neither forward or backward by abs (Di - Dj). Token C continues to observe the ‘Train’ moving with the same discrete \nrelative-speed. So the Lemma-3.1 and Lemma-3.2 continue to remain true. Hence Proved Theorem 3."},{"page":4,"text":"2.6 Theorem-4:\nIf we consider all n Functions defined in Section 2.1, there are an infinite number of Intervals along the t-Axis, when all n\nFunctions are simultaneously high for duration greater than (x - μ), where (μ/x) 0, and μ 0. Please see figure 4 below.\nFig 4. There are an infinite number of Intervals, when all n Functions are simultaneously high for duration greater than x - μ\nProof: We shall give two angles of argument, to support the validity of this Theorem.\nAngle-1: As discussed under the proof for Theorem-3, that for any two Functions i and j, there is a sophisticated relationship\ngoverning the occurrences of the start of Intervals along the t-Axis, where during such Intervals, both Functions are \nsimultaneously high for duration greater than x - μ1, where μ1 > 0 and (μ1/x) 0. This sophisticated relationship involves \nTi , Tj, Di , Dj , μ1 , and x. Let this relationship be expressed as Gij = 0. Next, as there are infinite such Intervals, whose \nstarting points have no constant period, the degree of Gij is also infinite. Therefore when we start considering some another \nFunction k defined in Section 2.1 (apart from already considering Functions i and j), then it is impossible to express Function \nk as a rational multiple of Gij =0. Thus, the Function Gij=0, and Function k, should again have an infinite number of \nIntervals, where x - μ2, where μ2 > 0 and (μ2/x) 0. Of course, the Intervals while considering Functions i, j, and k, will\nbe less frequent along the t-Axis, than the Intervals when only the Functions i and j were considered, but there will still be an \ninfinite number of such Intervals. Similarly, as we go on considering more Functions from Section 2.1, one by one, we \nshould find Infinite such Intervals along the t-Axis, though their occurrence is with lesser and lesser frequency.\nAngle-2: Let n Trains stand next to each other on parallel tracks, with each Train having an infinite number of carriages. Let \nthe length of all the carriages of Train i is set to Ti , for all i in [1,n], where the irrational R is very small (e.g. 10-n √2).\nEach carriage has exactly one window situated at a distance of Di from the start of that carriage.\nNow imagine a window-cleaner-man on Train 1, who, after cleaning the window of a carriage, moves to clean the \nwindow of the next forward carriage of the same Train 1. Let the man keep moving forward in this way, cleaning windows of \nthe carriages of Train 1. It is obvious from Theorems 1 and 2, that every time the man looks out from a different window of\nTrain 1, the man will always perceive a different permutation of arrangements of carriages of the parallel Trains 2 to n. This \nalso means that if, for every Train, if we were to divide each carriage into M Equal parts (where M is an extremely large \nnatural number), and color the parts with M different colors (same coloring sequence is followed for the M parts of all \ncarriages of all Trains), then the following statement is true: - Irrespective of how large M is, it is always possible to draw \nMN lines orthogonal to the Trains, where each line has a different permutation of colors.\nThus, the fact that the periods of all n Periodic Functions of Section 2.1, are irrational with respect to each \nother is the most important factor, which enables us conclude that there exist infinite Intervals along the t-Axis, where all n \nFunctions are simultaneously high for a duration greater than x - μ, where μ > 0 and (μ/x) 0. Hence Proved Theorem-4."},{"page":5,"text":"3. Our New Approach\n3.1 Basic Idea of our Model\nThe basic idea of our Model is to build a set of Equations describing the HCP, such that this set of Equations has a constant \nnumber of variables, irrespective of N, the number of nodes in the HCP. The feasible solution of each variable is bounded \nwithin the space of 2 real numbers.\n3.2 A Crucial substitution, based on Theorem-4, to keep a Constant number of variables in our Model\nFor each of the N(N-1) edge-variables in the directed Graph G, we make the following substitution for xi , j (edge-variable for \nthe edge from node i to node j, which in our Model, is allowed to have a value that is either approx 2 or approx 0):\n \n \n \n \n \n \nj\ni\nj\ni\nC\nj\nN\ni\nx\nhave\nwe\nj\ni\nand\nN\nin\nj\ni\n,\n,\n)1\n(\ncos\n1\n,\n],\n,1[\n}\n,\n{\n \n \n \n \n \n \n \n \n \nCi, j is the cost from Node i and Node j, which can be either 0 or 1, and is obtained from the specific instance of the HCP.\nThe above substitution of xi , j is performed with the trigonometric cosine function with period = 1/(1+((i-1)N + j)(β/α)),\nbecause from Theorem-4 (replacing R with β/α), 2N(N-1) permutations are possible, as explained by the following statement:\nThe highest-point/lowest-point of x1 ,1 can come ‘extremely close’ to the \n(statement starts and is contd in next line)\nthe highest-point/lowest-point of x1 ,2 , which can come ‘extremely close’ to the\n(statement contd in next line)\n...................\n(statement contd in next line)\nthe highest-point/lowest-point of xi ,j , which can come ‘extremely close’ to the\n(statement contd in next line)\n...................\n(statement contd in next line)\nthe highest-point/lowest-point of xn ,(n-1) , which can come ‘extremely close’ to the\n(statement contd in next line)\nthe highest-point/lowest-point of xn ,n .\n(statement ends)\nBy ‘extremely close’, we mean ‘within a distance that tends to 0, and which still may not yet be equal to 0’.\nin\npolynomial\na\nas\nx\nress\nto\ntry\nshall\nwe\nN\nin\nj\ni\nNow\nj\ni,\nexp\n],\n,1[\n}\n,\n{\n, \n \n \n \n \n \n \n \n \nsin\ncos\n,\nsin\n,\ncos\nand\n. \nWe know from standard Trigonometry that the following 3 Equations are true:\n \n \n \n \n \n \n \n \n \n \n \n \n \n \nk\nk\nk\nsin\nsin\ncos\ncos\ncos\n \n \n \n,and\n \n \n \n \n \n \n \n \n \n \n2\ncos\ncos\n)1\n(\ncos\n2\ncos\n \n \n \n \nk\nk\nk\n, and\n \n \n \n \n \n \n \n \n \n \n2\nsin\ncos\n)1\n(\nsin\n2\nsin\n \n \n \n \nk\nk\nk\nIteratively follow the above 3 Equations, so as to express\n \n \n \n \n \n \nk\nand\nk\nsin\ncos\n, as polynomials in \n \n \n \n \nsin\ncos\nand\n. Note that this has been discussed in further detail in [1][2]. It is important, and worthwhile to note \nthat though the co-efficients of the above polynomials have values of the form 2P(N) (where P(N) is some polynomial\nfunction of N), still these co-efficients occupy polynomial size proportional to log (2P(N)), which is P(N).\ns\npolynomial\nas\nx\nress\nto\nable\nare\nwe\nN\nin\nj\ni\nThus\nj\ni,\nexp\n],\n,1[\n}\n,\n{\n, \n \n \n \n \n \n \n \n \nsin\ncos\n,\nsin\n,\ncos\nand\nin\n.\nThe purpose of performing this substitution is to ensure that evaluating the expressions of our Model are maintained within \npolynomial complexity. For example, it is obvious that, without our substitution, the process of bringing the expression \n \n \n \n \n \n \n \n \n \n \n \n \n \nN\ni\nN\ni\nj\nj\nj\ni\nx\n1\n,\n1\n,\ninto simplest form (i.e. where all terms of the expression have only ‘+’, ‘-’, ‘x’ or ‘/’ \noperators) needs exponential space and time. Whereas, our crucial substitution enables this expression to be brought into \nsimplest form within polynomial space and time."},{"page":6,"text":"3.3 Set of Constraints for our Algorithm (note that before bringing any of these Constraint Sets to Simplest form, \nwe should first Substitute all xi, j by the trigonometric polynomials in 4 variables, as described in Section 3.2)\n \n \n \n \n \n \n1\n1\n,1\n2\n,\n2\n,\n)1\n(\n4\n0\n2\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \nN\nN\nx\nx\nN\ni\nN\ni\nj\nj\nj\ni\nj\ni\nCONSTRAINT-SET-1\n2\n1\n2\n,1\n,\n3\n)1\n(\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \nN\ni\nN\ni\nj\nj\nj\nix\nCONSTRAINT-SET-2\n \n \n \n \n3\n2\n2\n1\n,\n2\n1\n,1\n1\n2\n \n \n \n \n \n \n \n \nN\ni\nN\ni\nN\nY\nY\nCONSTRAINT-SET-3\nWhere\n]\n,2\n[\n0\n,1\n1,\n1,1\nN\ni\nY\nY\ni\n \n \n \n \n, and\n \n \n \n \n \n \n \n \n \n \n \n \nN\ni\nj\nj\nj\ni\nt\nj\nt\ni\nt\ni\nx\nY\nY\nY\nhave\nwe\nN\nin\ni\n,1\n,\n,\n,\n1\n,\n2\n],\n,1[\n \n \n2\n2\n0\n1\nN\n \n \n \n \n \n \n2\n2\n0\n2\nN\n \n \n \n \nCONSTRAINT-SET-4\n \n \n2\n2\n0\n3\nN\n \n \n \n \n \n1\ncos\n1\n \n \n \n \n \n1\nsin\n1\n \n \n \n \n \n1\ncos\n1\n \n \n \n \n \n1\ncos\n1\n \n \n \n \nCONSTRAINT-SET-5\n \n \n1\ncos\nsin\n2\n2\n \n \n \n \n \n \n1\ncos\nsin\n2\n2\n \n \n \n \n3.4 Explanation of the Constraint Sets\nIn sub-Section 3.3, Constraint Set 5 ensures that each edge variable is within [0,2], and also validates the trigonometric \nsubstitution. Constraint Set 1 ensures that the value of each edge-variable is either approximately 2 or approximately 0 (by \nthe word ‘approximately’ we mean plus/minus a small quantity much less than 2-(NxN)). Constraint Set 2 then ensures that"},{"page":7,"text":"there is exactly one edge-variable connected to node i, whose value is approximately 2, while all other edge-variables \nconnected to node i have values approximately 0. Once this is done, the resulting feasible solutions are either a Hamiltonian \nCycle (i.e. Cycle of length = N), or isolated rings (i.e. isolated Cycles each of length < (N-2) ). \nNext, Constraint Set 3 constrains the previously discussed solutions to only Hamiltonian Cycles (here we assume \nN>2). Constraint Set 3 is actually the simulation of a signal propagating among all the Nodes, from time or t = 1 to t = N, \nwhere t increases discretely in steps of 1. At time =1, we start with Unit signal strength at Node 1, and zero signal strength at \nall the other Nodes. At the next Unit of time, each Node receives signals from all other Nodes, along the edges, and adds it to \nwhatever signal strength it already had. Finally at t = (N+1), if Node 1 has signal strength of approximately 2, and every other \nNode has a signal strength of approximately 1, then it signifies a Hamiltonian Cycle.\nLastly, Constraint Set 4 constrains all approximation to be within 2-(NxN), so that Constraint Sets 1-3 remain valid.\nNote that we could have chosen any other number such as 2-(NxNxNxN) to constrain the approximation. The reason why we \nstuck with 2-(NxN), is because it is sufficiently small to ensure that the Model is valid, and still can be expressed within \npolynomial space of (NxN) bits. Note that a too small number like 2-(NxNxNxN.....N times) will take up exponential space.\nIn summary, our Constraint Equations are basically looking for the possibility that N edge-variables possess\nvalues approximately equal to and less than 2, whose corresponding edges form a Hamiltonian Cycle, while all other \nedge-variables possess values approximately equal to and greater than 0.\n3.5 Converting Inequations into Equations\nThe Inequations described under Section 3.3 (Constraint Sets 4 and 5) can be converted into Equations, because an \nInequation of the form \n \n \nx\n can be converted into an Equation (by the introduction of an extra variable z) of the form:\n \n \n0\n2\n \n \n \n \nx\nz\n, and then checking for the existence of solutions for both variables x and z. It is to be noted, that even the \nextra variable will also be bounded within the space of 2 real numbers.\n3.6 Final Set of Equations, and their conversion into a single Polynomial Equation\nAfter converting the Inequations to Equations, we finally have a constant number of Polynomial Equations, with a constant \nnumber of variables. It is easy to convert them into a single Polynomial Equation, by squaring and adding the Polynomial \nportions of the Equations, and then equating the result to zero.\n4. Conclusion, and Future Work\nIn this paper, we presented a polynomial Approach for expressing the decision version of the directed Hamiltonian Cycle \nProblem (NP-Complete) or HCP, in terms of the Solvability of an Equation. Using a crucial trigonometric substitution that \nenables the N(N-1) edge-variables to be expressed as polynomials in 4 variables, our Approach describes the HCP with a \nsingle Polynomial Equation with a constant number of variables, where the feasible solution of each variable is bounded \nwithin a set of 2 real numbers. It is hoped that our approach will be of use, in mankind’s P? =NP Question.\nWe hope someone in the world can develop EITHER:\nTechnique-1: A Technique that determines whether or not a Polynomial Equation (with a constant number of variables) is \nsolvable within the Infinite unbounded Real Set, OR\nTechnique-2: A Technique that determines whether or not a Polynomial Equation (with a constant number of variables) is \nsolvable within a bounded Real Set (i.e. where the feasible set of each variable is bounded within a space of 2 real numbers)\nWe could use either of these Techniques for our HCP’s Polynomial Equation (Technique-1 will also work for our \nModel, because we have explicit constraints limiting our variables within the feasible bounded Real Set). So we could use \neither of the Techniques to determine whether, for a particular HCP instance, if there does exist atleast one solution point for \nour variables (i.e. sin(α), cos(α), sin(β), cos(β), ε1 , ε2 , ε3, and the other variables introduced by converting Inequations \nto Equations), then we can conclude that the HCP instance does have a Hamiltonian Cycle. And if their Technique finds that \nno such solution is possible, then we can conclude that the HCP instance does not have a Hamiltonian Cycle.\nSo mankind will be able to prove that P=NP, if either Technique-1 or Technique-2 is developed.\nReferences\n[1] http://mcraefamily.com/MathHelp/GeometryTrigEquivCos3xEtcB.htm\n[2] http://mcraefamily.com/MathHelp/GeometryTrigEquivSin3xEtcC.htm\nAcknowledgement\nI am grateful to Dr. David Moews (djm.cc/dmoews.html), who is an IMO Gold Medallist, and three time Putnam Prize Winner with a \nperfect score in the 1987 difficult one, for his enlightening discussions, and also for pointing out a crucial error in a previous version of my \npaper. I asked Dr. Moews whether I could co-author him in this paper, but he has not responded yet."}]},"section_tree":[],"assets":{"figures":[],"tables":[],"images":[]},"math_expressions":{"inline":[],"block":[{"equation_id":"eq1","equation_number":null,"raw_text":"these variables is bounded within two real numbers. We point out what future work is necessary to prove that P=NP.","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq2","equation_number":null,"raw_text":"Property-1: Period of Function i, Ti = (1 / (1+ iR) ), where R is irrational non-zero real, and i represents integers 1 to n","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq3","equation_number":null,"raw_text":"Property-3: Function i is allowed to start to become high from time t=0, with real delay Di , where 0 ≤ Di < Ti","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq4","equation_number":null,"raw_text":"Proof: If {E(tA)i , E(tA)j } = {E(tB)i , E(tB)j }, for say tA < tB, then it would mean that it is possible to identify another","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq5","equation_number":null,"raw_text":"point tC where {E(tB)i , E(tB)j } = {E(tC)i , E(tC)j }, and where tC – tB = tB – tA. It would also mean that an infinite","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq6","equation_number":null,"raw_text":"number of similar points like tC can be periodically found along the t-Axis. This, in turn, would imply that N1 Ti = N2 Tj,","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq7","equation_number":null,"raw_text":"for some natural numbers N1 and N2, where N1 and N2 ≠ 0. But we know that (Ti / Tj ) = (1+ jR) / (1+ iR), which is","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq8","equation_number":null,"raw_text":"Lemma-3.1: If Di = Dj = 0, there exists some real t > 0, where the start of the highs of Functions i and j, come closer to","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq9","equation_number":null,"raw_text":"Proof: Consider two tokens, Tokens A and B, to be placed at t=0. If Ti > Tj , then let TA = Ti and TB = Tj , else let TA = Tj","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq10","equation_number":null,"raw_text":"and TB = Ti . Now follow the procedure below (note that INT (argument) returns Integer portion of that argument):","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq11","equation_number":null,"raw_text":"Step 1: Set Position_of_Token_A = TA","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq12","equation_number":null,"raw_text":"Step 2: Set Position_of_Token_B = Position_of_Token_B + TB","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq13","equation_number":null,"raw_text":"Step 4: Set Difference_in_token_positions = ( Position_of_Token_B - Position_of_Token_A )","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq14","equation_number":null,"raw_text":"Step 5: Set Integral_Multiple = 1 + INT ( (Position_of_Token_A) / (Difference_in_token_positions) )","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq15","equation_number":null,"raw_text":"Step 6: Set Position_of_Token_A = ( 2 x Position_of_Token_A ) x ( Integral_Multiple )","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq16","equation_number":null,"raw_text":"Step 7: Set Position_of_Token_B = Position_of_Token_B x ( Integral_Multiple )","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq17","equation_number":null,"raw_text":"(because both Tokens start from t=0, and because both Tokens are iteratively shifted forward by integral multiples of Ti and","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq18","equation_number":null,"raw_text":"≠ 0), because if they touch, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2, where N1 and N2 ≠ 0,","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq19","equation_number":null,"raw_text":"Lemma-3.2: If Di = Dj = 0, then there exist an infinite number of Intervals along t-Axis, where Functions i and j, are","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq20","equation_number":null,"raw_text":"positive distance μ, where (μ/x) 0. Let Li denote the distance from t=0 to the start of that high of Function i, and let Lj","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq21","equation_number":null,"raw_text":"denote the distance from t=0 to the start of that high of Function j, so that abs (Li - Lj) = μ. Now starting from t=0, let us","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq22","equation_number":null,"raw_text":"Similarly, starting from t=0, let us place infinite Tokens of type B along the t-Axis so that two neighboring tokens of type B","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq23","equation_number":null,"raw_text":"are separated by distance Lj. Now imagine a single Token C, starting from t=0, and iteratively taking leaps of Li every unit of","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq24","equation_number":null,"raw_text":"easy to see that if they do have a constant period, then it would mean that N1 Ti = N2 Tj, for some natural numbers N1 and N2,","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq25","equation_number":null,"raw_text":"Ti , Tj, Di , Dj , μ1 , and x. Let this relationship be expressed as Gij = 0. Next, as there are infinite such Intervals, whose","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq26","equation_number":null,"raw_text":"k as a rational multiple of Gij =0. Thus, the Function Gij=0, and Function k, should again have an infinite number of","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq27","equation_number":null,"raw_text":"The above substitution of xi , j is performed with the trigonometric cosine function with period = 1/(1+((i-1)N + j)(β/α)),","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq28","equation_number":null,"raw_text":"Cycle (i.e. Cycle of length = N), or isolated rings (i.e. isolated Cycles each of length < (N-2) ).","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq29","equation_number":null,"raw_text":"N>2). Constraint Set 3 is actually the simulation of a signal propagating among all the Nodes, from time or t = 1 to t = N,","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq30","equation_number":null,"raw_text":"where t increases discretely in steps of 1. At time =1, we start with Unit signal strength at Node 1, and zero signal strength at","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq31","equation_number":null,"raw_text":"whatever signal strength it already had. Finally at t = (N+1), if Node 1 has signal strength of approximately 2, and every other","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq32","equation_number":null,"raw_text":"within a set of 2 real numbers. It is hoped that our approach will be of use, in mankind’s P? =NP Question.","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"},{"equation_id":"eq33","equation_number":null,"raw_text":"So mankind will be able to prove that P=NP, if either Technique-1 or Technique-2 is developed.","raw_latex":null,"normalized_latex":null,"display_type":"block_candidate"}],"equation_refs":[]},"symbol_table":[],"references":{"in_text_markers":[],"bibliography":[]},"claims":[],"flow_graph":{"nodes":[],"edges":[]},"quality":{"text_char_count":24034,"parse_confidence":0.5,"equation_parse_rate_proxy":1.0,"citation_resolution_rate_proxy":0.0,"structure_coverage_proxy":0.0,"asset_coverage_proxy":0.0,"accepted_for_training":true}}