Reading group: proof of the PCP theorem
|
|
- Monica Dickerson
- 5 years ago
- Views:
Transcription
1 Reading group: proof of the PCP theorem 1 The PCP Theorem The usual formulation of the PCP theorem is equivalent to: Theorem 1. There exists a finite constraint language A such that the following problem is NP-hard: given a CSP(A)-instance Φ decide if Φ is satisfiable or if every subset of the constraints of size Φ / is unsatisfiable. The problem in the statement is also called GapCSP(A). The Label Cover Problem Let Σ be a finite alphabet. An instance of the label cover problem for Σ is: a bipartite graph (V 1, V, E), for every (u, v) E, a function π u,v : Σ Σ. A solution to an instance is a pair of mappings s 1 : V 1 Σ and s : V Σ such that for all (u, v) E, π u,v (s 1 (u)) = s (v). Proposition 1. Label cover over a 3-element alphabet is NP-hard. Definition 1. A (strong) Mal cev condition (L, Σ) where L is a finite functional language and Σ is a finite set of identities over L. Example 1. m(x, x, y) m(y, x, x) y with L = {m}. Definition. A Mal cev condition (L, Σ) is satisfied in a clone C if there is a mapping ξ : L C such that ξ(l) = Σ. A Mal cev condition is said to be trivial if one of the following equivalent statements hold: 1. it is satisfied in the clone of projections,. it is satisfied in every clone, Notes as of February 5, 018. Report mistakes to antoine.mottet@tu-dresden.de. 1
2 3. it is satisfied in Pol(1-in-3-SAT). A Mal cev condition is said to be bipartite if there is a partition L = L 1 L such that every identity is of the form f(variables) g(variables) with f L 1 and g L. It is special bipartite if it is of the form f(variables) g(x 1,..., x n ) (all variables are different in the right-hand side). Proposition. It is NP-hard to decide whether a height 1 special bipartite Mal cev condition is trivial. We first prove that the two problems are equivalent up to polytime reductions. The translation from the Mal cev problem to LC(Σ) is as follows. The vertices (V 1, V ) correspond to (L 1, L ). A set of identities E gives an edge relation on V 1 V. Given an edge f(x π(1), x π(),..., x π(n) ) g(x 1,..., x n ), we take π f,g to be π. Was the other reduction given? Proof of Proposition. We even prove that the even more special case of height 1 bipartite special ternary in two variables is NP-hard. We reduce 1-in-3-SAT to this problem. Let V, C be an instance of 1-in-3- SAT. We take L to be the set of variables. We take L 1 to contain one symbol f C for each constraint C(x i1, x i, x i3 ) in C. We add the equalities f C (x, x, y) g i3 (x, y), f C (x, y, x) g i, and f C (y, x, x) g i1. Definition 3. The GapLabelCover ɛ (Σ) problem is as follows: given an LC(Σ)- instance, decide if it is solvable, or if no ɛ-part is solvable. Theorem. For all ɛ > 0, there exists Σ such that GapLabelCover ɛ (Σ) is NP-hard. Be convinced that PCP implies that there exist ɛ (0, 1) and Σ such that GapLabelCover ɛ (Σ) is NP-hard. Parallel repetition (no time to formalize) gives a way to turn ɛ into any number in (0, 1). 3 Expander graphs I only have two hours to talk so let s get on with it. Marcello Mamino We want to prove: fix finite domain Σ and let Γ be the full binary constraint language on Σ. We prove: the classical problem CSP(Γ) reduces in polynomialtime to GapCSP(Γ). More precisely. Let Φ be an instance of CSP(Γ) that is unsatisfiable, but a big portion of the constraints are simultaneously satisfiable. We want to construct a new instance Ψ such that at most half of the constraints are satisfiable. Let n be the number of constraints of Φ. The gap of Φ is the smallest number
3 α [0, 1] such that a part of Φ of size (1 α)n is satisfiable. If Φ is not satisfiable, the gap is > 0. We amplify this gap by multiplying it repeatedly by a constant k, which is meant to be independent of the size of Φ. At each step, suppose that our instance grows by a factor of h (independent of Φ as well). This means that the size of the instance grows as h number of steps n 1 log h (k). Strategic goal: Σ, k > 1, h, α such that we can effectively translate an instance of CSP(Γ) of gap x < α to one of gap xk and size not larger than h original size. We define the product of two instances Φ 1 (x) and Φ (y). It contains as variables x 1,..., x n and y 1,..., y m and constraints R(x ) S(y ) (of arity len(x ) + len(y )) for every pair of constraint R(x ) and S(y ) from Φ 1 and Φ. Claim. If α is the gap of Φ 1 and Φ, then gap of the product is 1 (1 α). Proof. Easy, just project on each coordinate, this gives subsets of each instance with relative size at most (1 α). This proves that once we manage to go from a gap that depends on Φ to the α from the statement (which is a constant), then we can also go to 1 by iterated squaring (a constant number of times). Let s now go back to the strategic goal. One step to reduce the gap is composed of the following substeps: 1. Regularize the graph (i.e., turn it into an expander).. Increase the gap (this causes the domain size of increase). 3. Shrink the domain. Expanders are used at every step of this strategy. Definition 4 (Expansion ratio). Let G be a graph. The edge expansion ratio of G, denoted by h(g), is defined as min V (G) S G, S E(S, S), S where E(S, S) is the set of edges with one endpoint in S and the other endpoint in S. Little game of the expanders: find infinite families of d-regular graphs with h(g) α > 0. It s easy to build such a family by a probabilistic argument, but we would like to compute them efficiently and deterministically. Let M(G) be the adjacency matrix of a d-regular graph G. This is a symmetric matrix with nonnegative integer entries, where the entry (v, w) for v, w G counts the number of edges between v and w. Take 1 dm(g), which is a bistochastic matrix. Being a symmetric real matrix, it is diagonalisable as A diag(1, λ,..., λ G ) A 1 with 1 λ λ G being the eigenvalues of 1 d M(G). 3
4 Exercise. Remember that the spectral radius of a stochastic matrix is precisely 1. Prove that 1 is an eigenvalue iff the graph is bipartite. Definition 5. The spectral gap of G is defined to be 1 λ. We also also define it as min(1 λ, 1 λ G ). Remark: if v is a probability distribution on V (G), a random walk with n steps gives the distribution ( 1 d M(G))n v, and the matrix ( 1 d M(G))n looks like A diag(1, λ n,..., λ n G )A 1, where the diagonal converges quickly to diag(1, 0,..., 0). Theorem 3. For every d-regular finite graph G with second greatest associated eigenvalue λ, we have d 1 λ h(g) d (1 λ ). Proof. Part 1: We prove the second inequality. Construct the matrix N with rows V and columns E. Fix an arbitrary orientation on G. For each edge e = (v, w), we put 1 in N at position n v,e and 1 at position n w,e. Let f : V R be a column vector. Then ( f T N ) = (i,j) E(G) (f(i) f(j)). Define B f := (i,j) E(G) f(i) f(j). We prove that f h(g) B f d f T N f holds when the support of f has size at most n and is nonnegative. Then by choosing f appropriately, this proves the second inequality. We prove f h(g) B f. Let β 0,..., β r be an enumeration of the image of f where β 0 < < β r. Note that β 0 = 0, since the support of f is not the whole V. Let V i be the set of x V (G) such that f(x) β i. We have V (G) = V 0 V 1 V r. Claim. B f is equal to r i=1 E(V i, V i ) (βi β i 1 ). Proof. Quite clear. Now, r i=1 E(V i, V i ) (βi β i 1 ) h(g) r i=1 V i (βi β i 1 ), where V (G) we use the fact that V i. We can rewrite this as h(g) r i=1 β i ( V i V i+1 ), if we define V r+1 :=. Finally, we note that this is simply equal to h(g) f. On to prove B f d f T N f. But this is just stupid computations: B f = (x,y) E f(x) f(y) f(x) + f(y) (f(x) (f(x) f(y)) + f(y)) f T N f(x) + f(y) f T N d f. 4
5 What we just proved is that h(g) d f T N f holds for every nonnegative function f : V R with support of size at most V /. Let g be an eigenvector associated with the eigenvalue λ. Define f := g +, i.e., the vector that is like g when it is nonnegative and is 0 otherwise. If the support of f is too big, just pick g := g and do the same. Define another matrix L := N N T of size V (G). The matrix L has d s on the diagonal and at position l i,j is 1 number of edges connecting i and j = dm i,j. Therefore, L = di dm. Let x be in the support of f. We compute (Lf)(x) = d f(x) y V (G) = d g(x) (Lg)(x) y supp(f) = (dg dλ g)(x) = d(1 λ )g(x). d M x,y f(y) dm x,y g(y) This is amazing, because it gives f T N = f T NN T f = f T Lf = (f T L supp(f) ) f = f T L supp(f) f d(1 λ ) f. So, in the inequality at the top of the page, we obtain h(g)... Part : let us now prove the first inequality. By rearranging, we want to prove dλ d h(g). First, it suffices to prove that there exists f : V R 1 that is orthogonal to. =: 1 V and such that we have d f T Mf f d h(g). 1 Indeed, suppose that we have such an f where the value of the fraction is α. Let A be orthogonal and such that A T MA is diagonal. Let g := A T f. The first observation is that the first component of g is 0 (trivial). Then just compute. Let S be a set witnessing the edge-expansion ratio. Define f := S 1 S S 1 S. It is indeed orthogonal to 1 V and then it is simple computation. The denominator of the fraction is f = S S + S S = ( S + S ) S S = n S S. Note that f can be written as n1 S S 1 V and also as n1 S + S 1 V. The numerator of the fraction is then f T dmf = (n1 S S 1 V ) T M( n1 S + S 1 V ) = n 1 T S dm1 S + n S 1 T S dm1 V + n S 1 T V dm1 S S S 1 T V dm1 V = n E(S, S) + nd S S + nd S S S S n d = nd S S n h(g) S. 5
6 The fraction is then d f T Mf f = nd S S n S S = d nh(g) S d h(g). n E(S, S) n S S This concludes the proof. Let us now see a construction of an infinite family of expanders! Definition 6. A graph G = (V, E) is an (n, d, α)-expander if it is d-regular, V = n and all the associated eigenvalues, except for 1 are bounded above in absolute value by α. We define the zigzag product of two graphs. Let G be an orange graph that is (n, m, α) and H be a blue graph that is (m, d, β). We build G H by first exploding every orange vertex by m vertices and then pasting arbitrarily a copy of H inside. The zigzag product is then the graph on this vertex set and where there is an edge (x, y) if there is a blue edge from x to z, an orange edge from z to w, and a blue edge from w to y. Theorem 4. (n, m, α) (m, d, β) = (n m, d, f(α, β)), where f(α, β) α + β + β. Moreover, if α, β < 1, then f(α, β) < 1. Proof. We only prove the claim about α + β + β. Let M be the normalized (or nonnormalized, but probably normalized) adjacency matrix of the resulting zigzag product. We can write M as HGH, where H is the adjacency matrix of the (m, d, β)-expander and G is the adjacency matrix of the (n, m, α)-expander. We give an upper bound of f T HGHf / f, for f orthogonal to 1 V. Write f := f + f, where f is constant in each town and f is such that the sum on each town is 0. Expand the product in the numerator, which gives f T HGHf + f HGHf + f T HGHf f T HGHf + f HGHf + f T HGHf = f T Gf + f GHf + f T HGHf α f + G f Hf + G Hf α f + f β f + β f α f + β f + β f. How to use this. Let H be a (d 4, d, 1 5 ) expander. Let G 1 := H, where H is the graph definable in H by E E. G 1 is a (d 4, d, 1 ) expander. Now define G n+1 := (G n ) H. Claim: G n+1 is a (d 4n, d, 1 ) expander. Indeed, (G n) is (d 4n, d 4, 1 4 ). Then by the property of the zigzag, we obtain a (d 4(n+1), d, ) expander. 6
7 How to build such an H? Let q be a prime number. Consider V := (F q ). The edges are {(a, b), (c, d)} such that ac = b + d. The intuition is that of (a, b) being on the line (c, d), in a projective setting. Proposition 3. G = (V, E) is a (q, q, 1 q ) expander. Proof. It is clearly a q-regular graph. Let M be the adjacency matrix. We have q q q q q (qm) = q q q q Write the matrix as q I + 1 q (1 q I), where 1 q is the square matrix of size q with all 1 s. Thus, the eigenvalues of (qm) are q, 0, q. Therefore, after normalizing the second greatest eigenvalue is 1 q. Now define: AP q := G, AP 1 q := AP q AP q, AP n+1 q := AP n q AP q Proposition 4. APq n is (q (n+1), q, O( 1 q )) (where the big-oh notation is with respect to q, not n!). Pick n = 3, q large enough, this gives the H that we wanted above. 4 Preprocessing Recall that our goal is to go from an instance Φ of CSP(Σ) to an instance Ψ of CSP(Σ) whose gap is at least α, for some constant α that does not depend on Φ. Moreover, we want this transformation to be doable in polynomial time. The formal statement is as follows: Theorem 5. There exists a finite domain Σ 0 such that for every finite domain Σ, there are C > 0 and 0 < α < 1 such that, given an input Φ of CSP(Σ), there exists Ψ an instance of CSP(Σ 0 ) that is computable in polynomial time and such that: 7
8 Ψ C Φ, if Φ is not satisfiable then Ψ is not satisfiable, gap(ψ) is at least min( gap(φ), α). To prove this theorem, we iterate a three-step procedure: Preprocessing Turn the input sentence Φ into a d-regular expander Θ 1, Amplification Compute an instance Θ of CSP(Σ ) with bigger gap, but also Σ = Σ poly(d), Reduction Compute an instance Θ 3 of CSP(Σ 0 ) with controlled gap. Here, we deal with the first step. It might be useful to recall that we view instances of the CSP as graphs (V, E) where each edge (u, v) is labelled by a subset of Σ Σ that gives the acceptable values for an assignment to the pair (u, v). We first need the following refinement of Marcello s expanders. (Note: in the following, I consider the eigenvalues of the non-normalized adjacency matrix. Deal with it.) Lemma 1. There exists d and 0 < λ 0 < d such that one can construct in polynomial time a family (H n ) of d -regular expanders such that for all n 0, the graph H n has n vertices and λ (H n ) < λ 0. Definition 7. Let G be a constraint graph over Σ. G := prep 1 (G) is defined as follows: The constraint graph V is the set of all elements of the form [v] := {(v, e) e E, v e}, for v V. E is the following set. First, all the edges of the form {(v, e), (v, e)} where e = {v, v } E are in E. Secondly, for every v let H v be a d- regular expander on [v] (with constant expansion ratio h). Add all the edges of H v to E. For every edge of the second kind, the constraint-label is equality (i.e., {(a, a) : a Σ}). For every edge {(v, e), (v, e)}, the constraint is the constraint of e. Clearly, prep 1 (G) is a d + 1 regular graph, V E, and E = E + v E(H v) E + d E = (d + 1) E. We show that the gap of prep 1 (G) is controlled by the gap of G. Proposition 5. We have O(1) gap(g) gap(prep 1 (G)) gap(g). Moreover, suppose we have an assignment (not necessarily satisfying) f : V Σ. Define g : V Σ by looking at the values f(v, e) for v E and (v, e) [v] and assigning g(v) to the one that appears the most often (in case of tie, decide arbitrarily). Then O(1) gap g (G) gap f (prep 1 (G)). 8
9 Proof. Let G := prep 1 (G). We start by the bound gap(g ) gap(g). Let f : V Σ be an assignment witnessing gap(g). Define g : V Σ by g(v, e) := f(v). Clearly g satisfies the equality constraints, and if {(v, e), (v, e)} is an edge, we have (g(v, e), g(v, e)) = (f(v), f(v )). Therefore, if f satisfies the constraint associated with e, then g satisfies the constraint associated with {(v, e), (v, e)}. So g violates exactly gap(g) E constraints of this type. Therefore, gap(prep 1 (G)) E gap(g) E gap(g). We prove the second part of the statement (which implies the first inequality). Let f : V Σ and g : V Σ be as in the statement. Let F E be the set of edges e = (u, v) such that (g(u), g(v)) does not satisfy the constraint associated with e in G. Similarly, let F E be the corresponding set of edges for f. Thus, gap g (G) = F E and gap f (G ) = F E. If F gap g (G) E, we have so that gap f (G ) E gap g(g) E, gap f (G ) gap g(g) E E gap g(g) (d + 1), and hence gap g (G) (d + 1) gap f (G ) as required. Otherwise, F < gap g (G) E. Let S = v V {(v, e) [v] g(v) f (v, e)}. Observe that for every e = {v, v } F, either {(v, e), (v, e)} belongs to F, or we have f(v, e) g(v) or f(v, e) g(v ), in which case v or v is in S. Hence, F + S F = gap g (G) E. With the bound on F that we have, this gives gap g (G) E < gap g (g) E + S, i.e., S > gap g (G) E. Fix one vertex v. Partition naturally the set S v := [v] S into a Σ Sa v. By the very definition of g, it is the case that Sv a [v] holds for all a Σ. Since H v is an expander on [v], we have that E(Sv a, Sv a ) Sv a h. Note that the constraint of every edge between Sv a and Sv a is violated by f, because this constraint is an equality constraint. Therefore, f violates at least v V 1 a Σ ( Sa v h) h v V S [v] > h gap g (G) h 4 E. That is to say, 4(d+1) gap g(g) gap f (G ). Now, we have a regular graph with bounds on the gap. We seek to make this graph into an expander with loops. Definition 8. Let G be a graph, and let H be a d -regular expander over the same vertices as G and with second eigenvalue λ (H) < λ 0 < d. Define prep (G) to be the graph over the same vertex set defined as follows. The edges of prep (G) are: edges of G, edges of H, and a loop at every vertex (we allow parallel edges here). The constraints are: the constraints of G for edges of G, and Σ Σ for all other edges. It is clear that prep (G) has V vertices, E(G) + E(H) + V edges. Moreover, if G is d-regular, then prep (G) is d + d + 1 regular. Proposition 6. Let G be a graph and let G := prep (G). Then 9
10 λ (G ) d + λ 0 + 1, for every f : V Σ, we have d d + d gap f (G) gap f (G ) gap f (G). Proof. The thing about the gap is obvious: the new constraints are not really constraints, so we only diluted the unsatisfiability. The first item is less obvious. Since the adjacency matrix of an undirected graph is symmetric, we have x the equality λ (G) = max T M(G)x x RV,x 1 V x = max x =1,x 1 V x T Mx. Note that M(G ) = M(G) + M(H) + I, so that x T M(G )x x T M(G)x + x T M(H)x + x. Maximizing over x we obtain λ (G ) λ (G)+λ (H)+1 < λ (G) + λ d + λ Putting together the two constructions, we obtain the preprocessing step: Corollary 1. There are 0 < λ 0 < d such that for every constraint graph G, the graph G := prep (prep 1 (G)) is constructible in polynomial time and such that: G is d-regular with self-loops, λ(g ) λ 0 < d, O(1) gap(g) gap(g ) gap(g). 5 Amplification 6 Alphabet Reduction by composition Suppose that we have a polytime reduction P from any language L in NP to the gap constraint satisfaction problem over Σ 0, where x L is mapped to a satisfiable instance and x L is mapped to an instance with gap at least α, for α > 0. We use P to build a reduction from the gap satisfaction problem over Σ to the gap satisfaction problem over Σ 0 by composition, i.e., given an input constraint graph G over Σ, we build a new constraint graph G P with the right properties. Definition 9. An assignment tester with alphabet Σ 0 and rejection probability ɛ > 0 is an algorithm P whose input is a circuit Φ over Boolean variables X, and whose output is a constraint graph G = (V, E) over Σ 0 and with constraints C such that V X and such that the following holds: let V = V \ X and a: X {0, 1}. Completeness if a is a valid assignment for Φ, there exists b: V Σ 0 such that gap a b (G) = 0, Soundness if a is not valid for Φ, then for all b: V Σ 0 we have gap a b (G) ɛ rdist(a, SAT(Φ)). 10
11 where for h, k Σ l, we write rdist(h, k) := dist(h,k) l. Theorem 6. There exists ɛ > 0 and an explicit construction of an assignment tester P with alphabet Σ = {0, 1} 3 and rejection probability ɛ. Proof.? We now show how to define the composition G P. This makes use of error correcting codes (maps Σ {0, 1} l ). Definition 10. An error correcting code e is said to have linear dimension if C > 0 such that l C log(σ). relative distance ρ > 0 if for every a 1 a Σ, we have rdist(e(a 1 ), e(a )) ρ. Two l-bits strings s 1, s are δ-far (resp. δ-close) if their relative distance is at most (resp. at least) δ. Definition 11. Let G = (V, E) be a constraint graph over Σ with constraints C, and let P be an assignment tester, ECC : Σ {0, 1} l arbitrary encoding with linear dimension and relative distance ρ > 0. We define G P to be the constraint graph (V, E ) over Σ 0 with constraints C as follows: Robustization convert very constraint c(e) C to a circuit c(e) using ECC: for every variable v V, let [v] be a fresh set of l boolean variables, for each e = (v, w) E, c(e) is a circuit on l variables [v] [w] that ouputs 1 iff the assignment is the image under ECC of a valid assignment for the constraint c(e). Composition Run P on each c(e), and let G e be the constraint graph (V e, E e ) that P (over Σ) outputs. Without loss of generality, we can assume that every E e is the same. The vertex set of G P is e E V e, the edge set is e E E e and C = e E C e. Lemma. Assume the existence of an assignment tester with constant rejection probability ɛ and alphabet Σ 0. There exist a constant β 3 > 0 that depends only on P and another constant c(p, Σ ) such that given any constraint graph G = ((V, E), Σ, C) we can compute in time linear in the size of G the graph G := G P such that: the size of G is c(p, Σ) G and β 3 gap(g) gap(g ) gap(g). 11
The PCP Theorem by Gap Amplification
The PCP Theorem by Gap Amplification Irit Dinur February 13, 2007 Abstract The PCP theorem [3, 2] says that every language in NP has a witness format that can be checked probabilistically by reading only
More informationComplexity Theory VU , SS The Polynomial Hierarchy. Reinhard Pichler
Complexity Theory Complexity Theory VU 181.142, SS 2018 6. The Polynomial Hierarchy Reinhard Pichler Institut für Informationssysteme Arbeitsbereich DBAI Technische Universität Wien 15 May, 2018 Reinhard
More informationEssential facts about NP-completeness:
CMPSCI611: NP Completeness Lecture 17 Essential facts about NP-completeness: Any NP-complete problem can be solved by a simple, but exponentially slow algorithm. We don t have polynomial-time solutions
More informationOutline. Complexity Theory EXACT TSP. The Class DP. Definition. Problem EXACT TSP. Complexity of EXACT TSP. Proposition VU 181.
Complexity Theory Complexity Theory Outline Complexity Theory VU 181.142, SS 2018 6. The Polynomial Hierarchy Reinhard Pichler Institut für Informationssysteme Arbeitsbereich DBAI Technische Universität
More informationCSC 5170: Theory of Computational Complexity Lecture 13 The Chinese University of Hong Kong 19 April 2010
CSC 5170: Theory of Computational Complexity Lecture 13 The Chinese University of Hong Kong 19 April 2010 Recall the definition of probabilistically checkable proofs (PCP) from last time. We say L has
More informationGeometric Steiner Trees
Geometric Steiner Trees From the book: Optimal Interconnection Trees in the Plane By Marcus Brazil and Martin Zachariasen Part 3: Computational Complexity and the Steiner Tree Problem Marcus Brazil 2015
More informationThe Complexity of Maximum. Matroid-Greedoid Intersection and. Weighted Greedoid Maximization
Department of Computer Science Series of Publications C Report C-2004-2 The Complexity of Maximum Matroid-Greedoid Intersection and Weighted Greedoid Maximization Taneli Mielikäinen Esko Ukkonen University
More informationLecture 17 November 8, 2012
6.841: Advanced Complexity Theory Fall 2012 Prof. Dana Moshkovitz Lecture 17 November 8, 2012 Scribe: Mark Bun 1 Overview In the previous lecture, we saw an overview of probabilistically checkable proofs,
More informationProbabilistically Checkable Proofs. 1 Introduction to Probabilistically Checkable Proofs
Course Proofs and Computers, JASS 06 Probabilistically Checkable Proofs Lukas Bulwahn May 21, 2006 1 Introduction to Probabilistically Checkable Proofs 1.1 History of Inapproximability Results Before introducing
More informationFirst-Order Logic. 1 Syntax. Domain of Discourse. FO Vocabulary. Terms
First-Order Logic 1 Syntax Domain of Discourse The domain of discourse for first order logic is FO structures or models. A FO structure contains Relations Functions Constants (functions of arity 0) FO
More informationPreliminaries and Complexity Theory
Preliminaries and Complexity Theory Oleksandr Romanko CAS 746 - Advanced Topics in Combinatorial Optimization McMaster University, January 16, 2006 Introduction Book structure: 2 Part I Linear Algebra
More informationPCPs and Inapproximability Gap-producing and Gap-Preserving Reductions. My T. Thai
PCPs and Inapproximability Gap-producing and Gap-Preserving Reductions My T. Thai 1 1 Hardness of Approximation Consider a maximization problem Π such as MAX-E3SAT. To show that it is NP-hard to approximation
More informationComputability and Complexity Theory: An Introduction
Computability and Complexity Theory: An Introduction meena@imsc.res.in http://www.imsc.res.in/ meena IMI-IISc, 20 July 2006 p. 1 Understanding Computation Kinds of questions we seek answers to: Is a given
More informationProblem Set 2. Assigned: Mon. November. 23, 2015
Pseudorandomness Prof. Salil Vadhan Problem Set 2 Assigned: Mon. November. 23, 2015 Chi-Ning Chou Index Problem Progress 1 SchwartzZippel lemma 1/1 2 Robustness of the model 1/1 3 Zero error versus 1-sided
More informationPCP Soundness amplification Meeting 2 ITCS, Tsinghua Univesity, Spring April 2009
PCP Soundness amplification Meeting 2 ITCS, Tsinghua Univesity, Spring 2009 29 April 2009 Speaker: Andrej Bogdanov Notes by: Andrej Bogdanov Recall the definition of probabilistically checkable proofs
More informationan efficient procedure for the decision problem. We illustrate this phenomenon for the Satisfiability problem.
1 More on NP In this set of lecture notes, we examine the class NP in more detail. We give a characterization of NP which justifies the guess and verify paradigm, and study the complexity of solving search
More informationCPSC 536N: Randomized Algorithms Term 2. Lecture 9
CPSC 536N: Randomized Algorithms 2011-12 Term 2 Prof. Nick Harvey Lecture 9 University of British Columbia 1 Polynomial Identity Testing In the first lecture we discussed the problem of testing equality
More informationLecture 15: Expanders
CS 710: Complexity Theory 10/7/011 Lecture 15: Expanders Instructor: Dieter van Melkebeek Scribe: Li-Hsiang Kuo In the last lecture we introduced randomized computation in terms of machines that have access
More informationCS 6820 Fall 2014 Lectures, October 3-20, 2014
Analysis of Algorithms Linear Programming Notes CS 6820 Fall 2014 Lectures, October 3-20, 2014 1 Linear programming The linear programming (LP) problem is the following optimization problem. We are given
More informationLecture 22: Counting
CS 710: Complexity Theory 4/8/2010 Lecture 22: Counting Instructor: Dieter van Melkebeek Scribe: Phil Rydzewski & Chi Man Liu Last time we introduced extractors and discussed two methods to construct them.
More informationPr[X = s Y = t] = Pr[X = s] Pr[Y = t]
Homework 4 By: John Steinberger Problem 1. Recall that a real n n matrix A is positive semidefinite if A is symmetric and x T Ax 0 for all x R n. Assume A is a real n n matrix. Show TFAE 1 : (a) A is positive
More informationMTAT Complexity Theory October 13th-14th, Lecture 6
MTAT.07.004 Complexity Theory October 13th-14th, 2011 Lecturer: Peeter Laud Lecture 6 Scribe(s): Riivo Talviste 1 Logarithmic memory Turing machines working in logarithmic space become interesting when
More informationAdmin NP-COMPLETE PROBLEMS. Run-time analysis. Tractable vs. intractable problems 5/2/13. What is a tractable problem?
Admin Two more assignments No office hours on tomorrow NP-COMPLETE PROBLEMS Run-time analysis Tractable vs. intractable problems We ve spent a lot of time in this class putting algorithms into specific
More information1 Randomized Computation
CS 6743 Lecture 17 1 Fall 2007 1 Randomized Computation Why is randomness useful? Imagine you have a stack of bank notes, with very few counterfeit ones. You want to choose a genuine bank note to pay at
More informationLecture 15: A Brief Look at PCP
IAS/PCMI Summer Session 2000 Clay Mathematics Undergraduate Program Basic Course on Computational Complexity Lecture 15: A Brief Look at PCP David Mix Barrington and Alexis Maciel August 4, 2000 1. Overview
More informationDRAFT. PCP and Hardness of Approximation. Chapter 19
Chapter 19 PCP and Hardness of Approximation...most problem reductions do not create or preserve such gaps...to create such a gap in the generic reduction (cf. Cook)...also seems doubtful. The intuitive
More information(x 1 +x 2 )(x 1 x 2 )+(x 2 +x 3 )(x 2 x 3 )+(x 3 +x 1 )(x 3 x 1 ).
CMPSCI611: Verifying Polynomial Identities Lecture 13 Here is a problem that has a polynomial-time randomized solution, but so far no poly-time deterministic solution. Let F be any field and let Q(x 1,...,
More informationAdvanced Algorithms (XIII) Yijia Chen Fudan University
Advanced Algorithms (XIII) Yijia Chen Fudan University The PCP Theorem Theorem NP = PCP(log n, 1). Motivation Approximate solutions Definition (Approximation of MAX-3SAT) For every 3CNF formula ϕ, the
More informationOut-colourings of Digraphs
Out-colourings of Digraphs N. Alon J. Bang-Jensen S. Bessy July 13, 2017 Abstract We study vertex colourings of digraphs so that no out-neighbourhood is monochromatic and call such a colouring an out-colouring.
More informationNotes for Lecture 2. Statement of the PCP Theorem and Constraint Satisfaction
U.C. Berkeley Handout N2 CS294: PCP and Hardness of Approximation January 23, 2006 Professor Luca Trevisan Scribe: Luca Trevisan Notes for Lecture 2 These notes are based on my survey paper [5]. L.T. Statement
More informationHARDNESS AMPLIFICATION VIA SPACE-EFFICIENT DIRECT PRODUCTS
HARDNESS AMPLIFICATION VIA SPACE-EFFICIENT DIRECT PRODUCTS Venkatesan Guruswami and Valentine Kabanets Abstract. We prove a version of the derandomized Direct Product lemma for deterministic space-bounded
More informationPCP Theorem and Hardness of Approximation
PCP Theorem and Hardness of Approximation An Introduction Lee Carraher and Ryan McGovern Department of Computer Science University of Cincinnati October 27, 2003 Introduction Assuming NP P, there are many
More informationIntroduction Long transparent proofs The real PCP theorem. Real Number PCPs. Klaus Meer. Brandenburg University of Technology, Cottbus, Germany
Santaló s Summer School, Part 3, July, 2012 joint work with Martijn Baartse (work supported by DFG, GZ:ME 1424/7-1) Outline 1 Introduction 2 Long transparent proofs for NP R 3 The real PCP theorem First
More informationNotes for Lecture Notes 2
Stanford University CS254: Computational Complexity Notes 2 Luca Trevisan January 11, 2012 Notes for Lecture Notes 2 In this lecture we define NP, we state the P versus NP problem, we prove that its formulation
More informationCLASSIFYING THE COMPLEXITY OF CONSTRAINTS USING FINITE ALGEBRAS
CLASSIFYING THE COMPLEXITY OF CONSTRAINTS USING FINITE ALGEBRAS ANDREI BULATOV, PETER JEAVONS, AND ANDREI KROKHIN Abstract. Many natural combinatorial problems can be expressed as constraint satisfaction
More information1 Primals and Duals: Zero Sum Games
CS 124 Section #11 Zero Sum Games; NP Completeness 4/15/17 1 Primals and Duals: Zero Sum Games We can represent various situations of conflict in life in terms of matrix games. For example, the game shown
More information1 Computational Problems
Stanford University CS254: Computational Complexity Handout 2 Luca Trevisan March 31, 2010 Last revised 4/29/2010 In this lecture we define NP, we state the P versus NP problem, we prove that its formulation
More informationApproximation algorithm for Max Cut with unit weights
Definition Max Cut Definition: Given an undirected graph G=(V, E), find a partition of V into two subsets A, B so as to maximize the number of edges having one endpoint in A and the other in B. Definition:
More informationLecture 2: November 9
Semidefinite programming and computational aspects of entanglement IHP Fall 017 Lecturer: Aram Harrow Lecture : November 9 Scribe: Anand (Notes available at http://webmitedu/aram/www/teaching/sdphtml)
More informationLecture Introduction. 2 Brief Recap of Lecture 10. CS-621 Theory Gems October 24, 2012
CS-62 Theory Gems October 24, 202 Lecture Lecturer: Aleksander Mądry Scribes: Carsten Moldenhauer and Robin Scheibler Introduction In Lecture 0, we introduced a fundamental object of spectral graph theory:
More informationLecture 59 : Instance Compression and Succinct PCP s for NP
IITM-CS6840: Advanced Complexity Theory March 31, 2012 Lecture 59 : Instance Compression and Succinct PCP s for NP Lecturer: Sivaramakrishnan N.R. Scribe: Prashant Vasudevan 1 Introduction Classical Complexity
More informationHomomorphism Testing
Homomorphism Shpilka & Wigderson, 2006 Daniel Shahaf Tel-Aviv University January 2009 Table of Contents 1 2 3 1 / 26 Algebra Definitions 1 Algebra Definitions 2 / 26 Groups Algebra Definitions Definition
More informationLecture 8 (Notes) 1. The book Computational Complexity: A Modern Approach by Sanjeev Arora and Boaz Barak;
Topics in Theoretical Computer Science April 18, 2016 Lecturer: Ola Svensson Lecture 8 (Notes) Scribes: Ola Svensson Disclaimer: These notes were written for the lecturer only and may contain inconsistent
More informationApproximation Algorithms and Hardness of Approximation. IPM, Jan Mohammad R. Salavatipour Department of Computing Science University of Alberta
Approximation Algorithms and Hardness of Approximation IPM, Jan 2006 Mohammad R. Salavatipour Department of Computing Science University of Alberta 1 Introduction For NP-hard optimization problems, we
More informationACO Comprehensive Exam March 17 and 18, Computability, Complexity and Algorithms
1. Computability, Complexity and Algorithms (a) Let G(V, E) be an undirected unweighted graph. Let C V be a vertex cover of G. Argue that V \ C is an independent set of G. (b) Minimum cardinality vertex
More informationP is the class of problems for which there are algorithms that solve the problem in time O(n k ) for some constant k.
Complexity Theory Problems are divided into complexity classes. Informally: So far in this course, almost all algorithms had polynomial running time, i.e., on inputs of size n, worst-case running time
More information6.5.3 An NP-complete domino game
26 Chapter 6. Complexity Theory 3SAT NP. We know from Theorem 6.5.7 that this is true. A P 3SAT, for every language A NP. Hence, we have to show this for languages A such as kcolor, HC, SOS, NPrim, KS,
More informationLecture 10: Hardness of approximating clique, FGLSS graph
CSE 533: The PCP Theorem and Hardness of Approximation (Autumn 2005) Lecture 10: Hardness of approximating clique, FGLSS graph Nov. 2, 2005 Lecturer: Venkat Guruswami and Ryan O Donnell Scribe: Ioannis
More informationA An Overview of Complexity Theory for the Algorithm Designer
A An Overview of Complexity Theory for the Algorithm Designer A.1 Certificates and the class NP A decision problem is one whose answer is either yes or no. Two examples are: SAT: Given a Boolean formula
More informationLecture 18: PCP Theorem and Hardness of Approximation I
Lecture 18: and Hardness of Approximation I Arijit Bishnu 26.04.2010 Outline 1 Introduction to Approximation Algorithm 2 Outline 1 Introduction to Approximation Algorithm 2 Approximation Algorithm Approximation
More informationICALP g Mingji Xia
: : ICALP 2010 g Institute of Software Chinese Academy of Sciences 2010.7.6 : 1 2 3 Matrix multiplication : Product of vectors and matrices. AC = e [n] a ec e A = (a e), C = (c e). ABC = e,e [n] a eb ee
More informationChapter 2. Reductions and NP. 2.1 Reductions Continued The Satisfiability Problem (SAT) SAT 3SAT. CS 573: Algorithms, Fall 2013 August 29, 2013
Chapter 2 Reductions and NP CS 573: Algorithms, Fall 2013 August 29, 2013 2.1 Reductions Continued 2.1.1 The Satisfiability Problem SAT 2.1.1.1 Propositional Formulas Definition 2.1.1. Consider a set of
More informationThe PCP Theorem by Gap Amplification
Electronic Colloquium on Computational Complexity, Revision 1 of Report No. 46 (2005) The PCP Theorem by Gap Amplification Irit Dinur September 25, 2005 Abstract We describe a new proof of the PCP theorem
More information5 Flows and cuts in digraphs
5 Flows and cuts in digraphs Recall that a digraph or network is a pair G = (V, E) where V is a set and E is a multiset of ordered pairs of elements of V, which we refer to as arcs. Note that two vertices
More informationLecture 6: Powering Completed, Intro to Composition
CSE 533: The PCP Theorem and Hardness of Approximation (Autumn 2005) Lecture 6: Powering Completed, Intro to Composition Oct. 17, 2005 Lecturer: Ryan O Donnell and Venkat Guruswami Scribe: Anand Ganesh
More informationRandom Lifts of Graphs
27th Brazilian Math Colloquium, July 09 Plan of this talk A brief introduction to the probabilistic method. A quick review of expander graphs and their spectrum. Lifts, random lifts and their properties.
More informationApproximate Counting CSPs Hunt for Galois Connection
Approximate Counting CSPs Hunt for Galois Connection Andrei A. Bulatov Simon Fraser University La Trobe University 203 2/36 Constraint Satisfaction Problem Decision: Given a conjunctive formula R decide
More information1 More finite deterministic automata
CS 125 Section #6 Finite automata October 18, 2016 1 More finite deterministic automata Exercise. Consider the following game with two players: Repeatedly flip a coin. On heads, player 1 gets a point.
More informationNP-Complete problems
NP-Complete problems NP-complete problems (NPC): A subset of NP. If any NP-complete problem can be solved in polynomial time, then every problem in NP has a polynomial time solution. NP-complete languages
More informationNotes for Lecture 3... x 4
Stanford University CS254: Computational Complexity Notes 3 Luca Trevisan January 14, 2014 Notes for Lecture 3 In this lecture we introduce the computational model of boolean circuits and prove that polynomial
More information1 Agenda. 2 History. 3 Probabilistically Checkable Proofs (PCPs). Lecture Notes Definitions. PCPs. Approximation Algorithms.
CS 221: Computational Complexity Prof. Salil Vadhan Lecture Notes 20 April 12, 2010 Scribe: Jonathan Pines 1 Agenda. PCPs. Approximation Algorithms. PCPs = Inapproximability. 2 History. First, some history
More informationShow that the following problems are NP-complete
Show that the following problems are NP-complete April 7, 2018 Below is a list of 30 exercises in which you are asked to prove that some problem is NP-complete. The goal is to better understand the theory
More informationLecture 17. In this lecture, we will continue our discussion on randomization.
ITCS:CCT09 : Computational Complexity Theory May 11, 2009 Lecturer: Jayalal Sarma M.N. Lecture 17 Scribe: Hao Song In this lecture, we will continue our discussion on randomization. 1 BPP and the Polynomial
More informationTheory of Computation
Theory of Computation Lecture #6 Sarmad Abbasi Virtual University Sarmad Abbasi (Virtual University) Theory of Computation 1 / 39 Lecture 6: Overview Prove the equivalence of enumerators and TMs. Dovetailing
More informationAbout the relationship between formal logic and complexity classes
About the relationship between formal logic and complexity classes Working paper Comments welcome; my email: armandobcm@yahoo.com Armando B. Matos October 20, 2013 1 Introduction We analyze a particular
More informationDISTINGUISHING PARTITIONS AND ASYMMETRIC UNIFORM HYPERGRAPHS
DISTINGUISHING PARTITIONS AND ASYMMETRIC UNIFORM HYPERGRAPHS M. N. ELLINGHAM AND JUSTIN Z. SCHROEDER In memory of Mike Albertson. Abstract. A distinguishing partition for an action of a group Γ on a set
More informationEquational Logic. Chapter Syntax Terms and Term Algebras
Chapter 2 Equational Logic 2.1 Syntax 2.1.1 Terms and Term Algebras The natural logic of algebra is equational logic, whose propositions are universally quantified identities between terms built up from
More informationCS286.2 Lecture 8: A variant of QPCP for multiplayer entangled games
CS286.2 Lecture 8: A variant of QPCP for multiplayer entangled games Scribe: Zeyu Guo In the first lecture, we saw three equivalent variants of the classical PCP theorems in terms of CSP, proof checking,
More informationSpectral Graph Theory Lecture 2. The Laplacian. Daniel A. Spielman September 4, x T M x. ψ i = arg min
Spectral Graph Theory Lecture 2 The Laplacian Daniel A. Spielman September 4, 2015 Disclaimer These notes are not necessarily an accurate representation of what happened in class. The notes written before
More information1 Adjacency matrix and eigenvalues
CSC 5170: Theory of Computational Complexity Lecture 7 The Chinese University of Hong Kong 1 March 2010 Our objective of study today is the random walk algorithm for deciding if two vertices in an undirected
More informationUC Berkeley CS 170: Efficient Algorithms and Intractable Problems Handout 22 Lecturer: David Wagner April 24, Notes 22 for CS 170
UC Berkeley CS 170: Efficient Algorithms and Intractable Problems Handout 22 Lecturer: David Wagner April 24, 2003 Notes 22 for CS 170 1 NP-completeness of Circuit-SAT We will prove that the circuit satisfiability
More information1 Circuit Complexity. CS 6743 Lecture 15 1 Fall Definitions
CS 6743 Lecture 15 1 Fall 2007 1 Circuit Complexity 1.1 Definitions A Boolean circuit C on n inputs x 1,..., x n is a directed acyclic graph (DAG) with n nodes of in-degree 0 (the inputs x 1,..., x n ),
More informationLecture 4 : Quest for Structure in Counting Problems
CS6840: Advanced Complexity Theory Jan 10, 2012 Lecture 4 : Quest for Structure in Counting Problems Lecturer: Jayalal Sarma M.N. Scribe: Dinesh K. Theme: Between P and PSPACE. Lecture Plan:Counting problems
More informationCOL 352 Introduction to Automata and Theory of Computation Major Exam, Sem II , Max 80, Time 2 hr. Name Entry No. Group
COL 352 Introduction to Automata and Theory of Computation Major Exam, Sem II 2015-16, Max 80, Time 2 hr Name Entry No. Group Note (i) Write your answers neatly and precisely in the space provided with
More informationModel Counting for Logical Theories
Model Counting for Logical Theories Wednesday Dmitry Chistikov Rayna Dimitrova Department of Computer Science University of Oxford, UK Max Planck Institute for Software Systems (MPI-SWS) Kaiserslautern
More informationCSE200: Computability and complexity Space Complexity
CSE200: Computability and complexity Space Complexity Shachar Lovett January 29, 2018 1 Space complexity We would like to discuss languages that may be determined in sub-linear space. Lets first recall
More informationUndecidable Problems. Z. Sawa (TU Ostrava) Introd. to Theoretical Computer Science May 12, / 65
Undecidable Problems Z. Sawa (TU Ostrava) Introd. to Theoretical Computer Science May 12, 2018 1/ 65 Algorithmically Solvable Problems Let us assume we have a problem P. If there is an algorithm solving
More informationNotes for Lecture 3... x 4
Stanford University CS254: Computational Complexity Notes 3 Luca Trevisan January 18, 2012 Notes for Lecture 3 In this lecture we introduce the computational model of boolean circuits and prove that polynomial
More informationLecture 6: Random Walks versus Independent Sampling
Spectral Graph Theory and Applications WS 011/01 Lecture 6: Random Walks versus Independent Sampling Lecturer: Thomas Sauerwald & He Sun For many problems it is necessary to draw samples from some distribution
More informationLecture 19: Interactive Proofs and the PCP Theorem
Lecture 19: Interactive Proofs and the PCP Theorem Valentine Kabanets November 29, 2016 1 Interactive Proofs In this model, we have an all-powerful Prover (with unlimited computational prover) and a polytime
More informationLecture 22. m n c (k) i,j x i x j = c (k) k=1
Notes on Complexity Theory Last updated: June, 2014 Jonathan Katz Lecture 22 1 N P PCP(poly, 1) We show here a probabilistically checkable proof for N P in which the verifier reads only a constant number
More informationAnswers to the CSCE 551 Final Exam, April 30, 2008
Answers to the CSCE 55 Final Exam, April 3, 28. (5 points) Use the Pumping Lemma to show that the language L = {x {, } the number of s and s in x differ (in either direction) by at most 28} is not regular.
More informationNP, polynomial-time mapping reductions, and NP-completeness
NP, polynomial-time mapping reductions, and NP-completeness In the previous lecture we discussed deterministic time complexity, along with the time-hierarchy theorem, and introduced two complexity classes:
More informationIntroduction to Complexity Theory
Introduction to Complexity Theory Read K & S Chapter 6. Most computational problems you will face your life are solvable (decidable). We have yet to address whether a problem is easy or hard. Complexity
More informationNP-completeness. Chapter 34. Sergey Bereg
NP-completeness Chapter 34 Sergey Bereg Oct 2017 Examples Some problems admit polynomial time algorithms, i.e. O(n k ) running time where n is the input size. We will study a class of NP-complete problems
More informationNP-COMPLETE PROBLEMS. 1. Characterizing NP. Proof
T-79.5103 / Autumn 2006 NP-complete problems 1 NP-COMPLETE PROBLEMS Characterizing NP Variants of satisfiability Graph-theoretic problems Coloring problems Sets and numbers Pseudopolynomial algorithms
More informationConvex and Semidefinite Programming for Approximation
Convex and Semidefinite Programming for Approximation We have seen linear programming based methods to solve NP-hard problems. One perspective on this is that linear programming is a meta-method since
More informationLecture 7: Spectral Graph Theory II
A Theorist s Toolkit (CMU 18-859T, Fall 2013) Lecture 7: Spectral Graph Theory II September 30, 2013 Lecturer: Ryan O Donnell Scribe: Christian Tjandraatmadja 1 Review We spend a page to review the previous
More informationNotes on Computer Theory Last updated: November, Circuits
Notes on Computer Theory Last updated: November, 2015 Circuits Notes by Jonathan Katz, lightly edited by Dov Gordon. 1 Circuits Boolean circuits offer an alternate model of computation: a non-uniform one
More information10. Smooth Varieties. 82 Andreas Gathmann
82 Andreas Gathmann 10. Smooth Varieties Let a be a point on a variety X. In the last chapter we have introduced the tangent cone C a X as a way to study X locally around a (see Construction 9.20). It
More informationCS154, Lecture 15: Cook-Levin Theorem SAT, 3SAT
CS154, Lecture 15: Cook-Levin Theorem SAT, 3SAT Definition: A language B is NP-complete if: 1. B NP 2. Every A in NP is poly-time reducible to B That is, A P B When this is true, we say B is NP-hard On
More informationMean-Payoff Games and the Max-Atom Problem
Mean-Payoff Games and the Max-Atom Problem Albert Atserias Universitat Politècnica de Catalunya Barcelona, Spain Elitza Maneva Universitat Politècnica de Catalunya Barcelona, Spain February 3, 200 Abstract
More informationTesting assignments to constraint satisfaction problems
Testing assignments to constraint satisfaction problems Hubie Chen University of the Basque Country (UPV/EHU) E-20018 San Sebastián, Spain and IKERBASQUE, Basque Foundation for Science E-48011 Bilbao,
More informationIITM-CS6845: Theory Toolkit February 3, 2012
IITM-CS6845: Theory Toolkit February 3, 2012 Lecture 4 : Derandomizing the logspace algorithm for s-t connectivity Lecturer: N S Narayanaswamy Scribe: Mrinal Kumar Lecture Plan:In this lecture, we will
More informationComputability Crib Sheet
Computer Science and Engineering, UCSD Winter 10 CSE 200: Computability and Complexity Instructor: Mihir Bellare Computability Crib Sheet January 3, 2010 Computability Crib Sheet This is a quick reference
More informationComputer Science 385 Analysis of Algorithms Siena College Spring Topic Notes: Limitations of Algorithms
Computer Science 385 Analysis of Algorithms Siena College Spring 2011 Topic Notes: Limitations of Algorithms We conclude with a discussion of the limitations of the power of algorithms. That is, what kinds
More informationCS 3719 (Theory of Computation and Algorithms) Lectures 23-32
CS 3719 (Theory of Computation and Algorithms) Lectures 23-32 Antonina Kolokolova March 2011 1 Scaling down to complexity In real life, we are interested whether a problem can be solved efficiently; just
More informationNP-Complete Problems. Complexity Class P. .. Cal Poly CSC 349: Design and Analyis of Algorithms Alexander Dekhtyar..
.. Cal Poly CSC 349: Design and Analyis of Algorithms Alexander Dekhtyar.. Complexity Class P NP-Complete Problems Abstract Problems. An abstract problem Q is a binary relation on sets I of input instances
More informationLecture 8 : Eigenvalues and Eigenvectors
CPS290: Algorithmic Foundations of Data Science February 24, 2017 Lecture 8 : Eigenvalues and Eigenvectors Lecturer: Kamesh Munagala Scribe: Kamesh Munagala Hermitian Matrices It is simpler to begin with
More information