The Chasm at Depth Four, and Tensor Rank: Old results, new insights

Size: px
Start display at page:

Download "The Chasm at Depth Four, and Tensor Rank: Old results, new insights"

Transcription

1 The Chasm at Depth Four, and Tensor Rank: Old results, new insights Suryajith Chillara 1 Joint work with Mrinal Kumar, Ramprasad Saptharishi and V. Vinay June 12, Research partly supported by TCS PhD fellowship. Part of this work was done while visiting Tel Aviv University, hosted by Amir Shpilka.

2 Arithmetic circuits Definition An Arithmetic circuit Φ over the field and the set of variables X = (x 1,x 2,...,x n ) is a directed acyclic graph as follows. Every gate with in-degree 0 is called an input gate and is labelled either by a variable or a field element from. Every other gate is labelled by either (product gate) or +(sum gate). The size of Φ is the number of edges present in it. The depth of Φ is the maximal depth of a gate in it.

3 Arithmetic circuit f (x 1,x 2,x 3 ) x 1 x 2 x 3

4 Arithmetic formulas Definition An arithmetic formula is an arithmetic circuit where the fan-out of every gate is at most 1.

5 Arithmetic formulas Definition An arithmetic formula is an arithmetic circuit where the fan-out of every gate is at most 1. Homogeneous formulas are formulas where the polynomial computed at each node is a homogeneous polynomial.

6 Arithmetic formula x 1 1 x 2 1 x n 1 1 x n Figure: The arithmetic circuit for f (x) = x i = (1 + x i ) S [n] i S i S

7 Best known size lower bounds for arithmetic formulas Kalorkoti [Kal85] proved a quadratic size lower bound against a very simple polynomial f (x 1,x 2,...,x n,y 1,...,y n ) = i [n] j [n] xj i y j using transcendence degree as a complexity measure.

8 Best known size lower bounds for arithmetic formulas Kalorkoti [Kal85] proved a quadratic size lower bound against a very simple polynomial f (x 1,x 2,...,x n,y 1,...,y n ) = i [n] j [n] xj i y j using transcendence degree as a complexity measure. He also proved a lower bound of Ω(n 3 ) against Det n.

9 Best known size lower bounds for arithmetic formulas Kalorkoti [Kal85] proved a quadratic size lower bound against a very simple polynomial f (x 1,x 2,...,x n,y 1,...,y n ) = i [n] j [n] xj i y j using transcendence degree as a complexity measure. He also proved a lower bound of Ω(n 3 ) against Det n. We in [CM15] proved a lower bound of Ω(dn 3 ) against IMM n,d.

10 Set-multilinear polynomials Let X = X 1 X d be a partition of the variables.

11 Set-multilinear polynomials Let X = X 1 X d be a partition of the variables. Definition (Set-multilinear polynomials) A polynomial f (X) is said to be set-multilinear with respect to the above partition if every monomial M in f satisfies var(m) X i = 1 for all i [d].

12 Set-multilinear polynomials Let X = X 1 X d be a partition of the variables. Definition (Set-multilinear polynomials) A polynomial f (X) is said to be set-multilinear with respect to the above partition if every monomial M in f satisfies var(m) X i = 1 for all i [d]. A set-multilinear formula is an arithmetic formula where the polynomial computed at every node is a set-multilinear polynomial.

13 Tensors Definition (Tensor) A tensor T is a map of the form T : V 1 V d where each V i is a vector space over, of dimension say m i. The parameter d is called the order of the tensor, and we say that the shape of T is [m 1 ] [m d ].

14 Tensors as polynomials Observation For any tensor T of shape [m 1 ] [m d ], we can associate a set-multilinear polynomial f (X) where X = X 1 X d and X i = x i1,...,x imi as follows. f (X) = 1 i j m j j [d] T (i 1,...,i d ) x 1i1 x did.

15 Tensor rank A tensor T 0 : [n] d is said to be of rank 1 if there exist d vectors v 1,,v d such that T 0 = v 1 v d.

16 Tensor rank A tensor T 0 : [n] d is said to be of rank 1 if there exist d vectors v 1,,v d such that T 0 = v 1 v d. The tensor rank of a tensor T : [n] d is the minimal r such that there exist r rank-1 tensors T 1,...,T r such that T = T T r.

17 Tensor rank A tensor T 0 : [n] d is said to be of rank 1 if there exist d vectors v 1,,v d such that T 0 = v 1 v d. The tensor rank of a tensor T : [n] d is the minimal r such that there exist r rank-1 tensors T 1,...,T r such that T = T T r. The tensor rank of T : [n] d is at most n d 1.

18 Tensors of high rank Lemma (Folklore) A generic tensor T : [n] d has a rank of at least nd 1 d.

19 Tensors of high rank Lemma (Folklore) A generic tensor T : [n] d has a rank of at least nd 1 d. Proof is via a simple argument involving algebraic independence.

20 Connections between formula size and tensor rank Theorem ([Raz10]) Let Φ be a formula of size s n c computing a set-multilinear polynomial f with respect to the partition X = X 1 X d. If d = O(logn/loglogn), then trk(f ) n d(1 1/exp(O(c))).

21 Connections between formula size and tensor rank Theorem ([Raz10]) Let Φ be a formula of size s n c computing a set-multilinear polynomial f with respect to the partition X = X 1 X d. If d = O(logn/loglogn), then trk(f ) n d(1 1/exp(O(c))). Theorem (This work) Let Φ be a formula of size s n c computing a set-multilinear polynomial f with respect to the partition X = X 1 X d. If d = O(logn), then trk(f ) n d(1 1/exp(O(c))).

22 An improvement Theorem (This work) Let Φ be a homogeneous formula of size s n c computing a set-multilinear polynomial f with respect to the partition X = X 1 X d. If logd = o(logn), then trk(f ) n d(1 1/exp(O(c))).

23 Overview of Raz s work Theorem (Step 1) Let Φ be a formula of size poly(n) computing an n-variate homogeneous polynomial f of degree d = O(logn). Then, there is a homogeneous formula Φ that also computes f of size at most poly(n).

24 Overview of Raz s work Theorem (Step 1) Let Φ be a formula of size poly(n) computing an n-variate homogeneous polynomial f of degree d = O(logn). Then, there is a homogeneous formula Φ that also computes f of size at most poly(n). Theorem (Step 2) Suppose d = O logn loglogn. If Φ is a formula of size s = poly(n) that computes a set-multilinear polynomial f, then there is a set-multilinear formula of poly(s) size that computes f as well.

25 Overview of Raz s work Theorem (Step 1) Let Φ be a formula of size poly(n) computing an n-variate homogeneous polynomial f of degree d = O(logn). Then, there is a homogeneous formula Φ that also computes f of size at most poly(n). Theorem (Step 2) Suppose d = O logn loglogn. If Φ is a formula of size s = poly(n) that computes a set-multilinear polynomial f, then there is a set-multilinear formula of poly(s) size that computes f as well. Theorem (Step 3) Let Φ be a set-multilinear formula of size s n c computing a polynomial f. Then trk(f ) < nd. n d/exp(o(c))

26 Overview of this result Theorem ([Raz10]) Let Φ be a formula of size poly(n) computing an n-variate homogeneous polynomial f of degree d = O(logn). Then, there is a homogeneous formula Φ that also computes f of size at most poly(n).

27 Overview of this result Theorem ([Raz10]) Let Φ be a formula of size poly(n) computing an n-variate homogeneous polynomial f of degree d = O(logn). Then, there is a homogeneous formula Φ that also computes f of size at most poly(n). Theorem (This work) Let Φ be a homogeneous formula of size s n c computing a set-multilinear polynomial f with respect to the partition X = X 1 X d. If logd = o(logn), then trk(f ) n d(1 1/exp(O(c))).

28 Properties of Tensor rank Sub-additivity: For T,T 1,T 2 : [n] d, if T = T 1 + T 2 then trk(t ) trk(t 1 ) + trk(t 2 ).

29 Properties of Tensor rank Sub-additivity: For T,T 1,T 2 : [n] d, if T = T 1 + T 2 then trk(t ) trk(t 1 ) + trk(t 2 ). Sub-multiplicativity: For T : [n] d, T 1 : [n] d 1 and T 2 : [n] d d 1, if T = T1 T 2 then trk(t ) trk(t 1 ) trk(t 2 ).

30 An example Let f be a set-multilinear polynomial over X 1 X d such that X i = n for all i.

31 An example Let f be a set-multilinear polynomial over X 1 X d such that X i = n for all i. The tensor rank is at most n d 1.

32 An example Let f be a set-multilinear polynomial over X 1 X d such that X i = n for all i. The tensor rank is at most n d 1. But if also know that f = f 1 f 2 where the product respects set-multilinearity, then trk(f ) trk(f 1 ) trk(f 2 ) n d 1 1 n d d 1 1 = n d 2 where d 1 is the degree of f 1.

33 Crux of the arguments Crux If a set multilinear polynomial that corresponds to an explicit tensor can be expressed as a summation over a few summands each of which has large number of factors, then we can get non-trivial bounds on the tensor rank.

34 Structure of homogeneous formulas Theorem (This work) Let f be a homogeneous n-variate degree d polynomial computed by a size s homogeneous formula. Then for any 0 < t d, f can be equivalently computed by a homogeneous ΣΠ [a] ΣΠ [t] formula of top fan-in s 10(d/t) where a > 1 d 10 t logt.

35 Structure of homogeneous formulas Theorem (This work) Let f be a homogeneous n-variate degree d polynomial computed by a size s homogeneous formula. Then for any 0 < t d, f can be equivalently computed by a homogeneous ΣΠ [a] ΣΠ [t] formula of top fan-in s 10(d/t) where a > 1 d 10 t logt. That is, f can be expressed as s 10(d/t) f = Q i1 Q i2...q ia i=1 where the degree of each Q ij is at most t.

36 Structure of homogeneous formulas Theorem (This work) Let f be a homogeneous n-variate degree d polynomial computed by a size s homogeneous formula. Then for any 0 < t d, f can be equivalently computed by a homogeneous ΣΠ [a] ΣΠ [t] formula of top fan-in s 10(d/t) where a > 1 d 10 t logt. That is, f can be expressed as s 10(d/t) f = Q i1 Q i2...q ia i=1 where the degree of each Q ij is at most t. It is important to note that this product is not set-multilinear.

37 Workaround Consider a term T = Q 1 Q 2...Q a from the above expression.

38 Workaround Consider a term T = Q 1 Q 2...Q a from the above expression. Let us extract out the set-multilinear computations from the product (denoted by SML(T )).

39 Workaround Consider a term T = Q 1 Q 2...Q a from the above expression. Let us extract out the set-multilinear computations from the product (denoted by SML(T )). For a partition S 1 S 2 S a of [d], the corresponding set-multilinear component is Q 1,S1 Q 2,S2 Q a,sa where Q i,si is the sum of set-multilinear monomials in Q i over the parts X j : j S i.

40 Upper bound SML(T ) = trk(sml(t )) S 1 S a =[d] S i =d i S 1 S a =[d] S i =d i trk(f ) n 10c(d/t)+d a Q 1,S1 Q a,sa n d a n d a d d 1 d 2 d a d d 1 d 2 d a

41 Upper bound SML(T ) = trk(sml(t )) S 1 S a =[d] S i =d i S 1 S a =[d] S i =d i trk(f ) n 10c(d/t)+d a Q 1,S1 Q a,sa n d a n d a d d 1 d 2 d a d d 1 d 2 d a We fix t to 2 110c and logd = o(logn), and get that trk(f ) n d(1 1/exp(c)).

42 Log product decomposition for homogeneous formulas Theorem ([HY11]) Let f be an n-variate degree d polynomial computed by a size s homogeneous formula. Then, f can be expressed as where f = s f i1 f i2 f ir (1) i=1 1. the expression is homogeneous, 2. for each i,j, we have 1 j 3 d deg(fij ) 2 j 3 d and r = Θ(logd), 3. each f ij is also computed by homogeneous formulas of size at most s.

43 Proof of depth reduction for homogeneous formulas We start with equation (1) which is easily seen to be a homogeneous ΣΠΣΠ [2d/3] circuit with top fan-in s: f = s f i1 f i2 f ir i=1

44 Proof of depth reduction for homogeneous formulas We start with equation (1) which is easily seen to be a homogeneous ΣΠΣΠ [2d/3] circuit with top fan-in s: f = s f i1 f i2 f ir i=1 1. For each summand f i1...f ir in the RHS, pick the gate f ij with largest degree (if there is a tie, pick the one with smaller index j). If f ij has degree more than t, expand that f ij in-place using (1).

45 Proof of depth reduction for homogeneous formulas We start with equation (1) which is easily seen to be a homogeneous ΣΠΣΠ [2d/3] circuit with top fan-in s: f = s f i1 f i2 f ir i=1 1. For each summand f i1...f ir in the RHS, pick the gate f ij with largest degree (if there is a tie, pick the one with smaller index j). If f ij has degree more than t, expand that f ij in-place using (1). 2. Repeat this process until all f ij s on the RHS have degree at most t.

46 Depth reduction to depth four g s g 11 g 12...g 1r s g s1 g s2...g sr s g 111 g g 11r g 12...g 1r g 1s1 g 1s2...g 1sr g 12 g 1r g s11 g s12...g s1r g s2...g sr g ss1 g ss2 g ssr g s2 g sr Figure: Depth reduction analysis

47 Analysis of the iterative procedure 1. Consider a potential function the total degree of all the big factors i.e., factors of degree > t.

48 Analysis of the iterative procedure 1. Consider a potential function the total degree of all the big factors i.e., factors of degree > t. 2. Geometric progression of degrees = total degree of small factors in a summand is at most 3t. Total degree of big terms is at least d 3t.

49 Analysis of the iterative procedure 1. Consider a potential function the total degree of all the big factors i.e., factors of degree > t. 2. Geometric progression of degrees = total degree of small factors in a summand is at most 3t. Total degree of big terms is at least d 3t. 3. Potential drops by at most 3t per iteration. = at least d/3t iterations are needed.

50 Analysis of the iterative procedure 1. Consider a potential function the total degree of all the big factors i.e., factors of degree > t. 2. Geometric progression of degrees = total degree of small factors in a summand is at most 3t. Total degree of big terms is at least d 3t. 3. Potential drops by at most 3t per iteration. = at least d/3t iterations are needed. 4. Every expansion introduces at least (log 3 t) non-trivial terms. Thus, a > d 3log3t logt.

51 Thank you! 2 2 The theme of these slides is based on the the theme made available at undercba licence.

52

Sums of products of polynomials of low-arity - lower bounds and PIT

Sums of products of polynomials of low-arity - lower bounds and PIT Sums of products of polynomials of low-arity - lower bounds and PIT Mrinal Kumar Joint work with Shubhangi Saraf Plan for the talk Plan for the talk Model of computation Plan for the talk Model of computation

More information

Arithmetic Circuits with Locally Low Algebraic Rank

Arithmetic Circuits with Locally Low Algebraic Rank Arithmetic Circuits with Locally Low Algebraic Rank Mrinal Kumar 1 and Shubhangi Saraf 2 1 Department of Computer Science, Rutgers University, New Brunswick, USA mrinal.kumar@rutgers.edu 2 Department of

More information

Lower bounds for multilinear bounded order ABPs

Lower bounds for multilinear bounded order ABPs Lower bounds for multilinear bounded order ABPs C. Ramya B. V. Raghavendra Rao IIT Madras, Chennai, India {ramya,bvrr}@cse.iitm.ac.in January 15, 2019 arxiv:1901.04377v1 [cs.cc] 14 Jan 2019 Abstract Proving

More information

Polynomial Identity Testing

Polynomial Identity Testing Polynomial Identity Testing Amir Shpilka Technion and MSR NE Based on joint works with: Zeev Dvir, Zohar Karnin, Partha Mukhopadhyay, Ran Raz, Ilya Volkovich and Amir Yehudayoff 1 PIT Survey Oberwolfach

More information

November 17, Recent Results on. Amir Shpilka Technion. PIT Survey Oberwolfach

November 17, Recent Results on. Amir Shpilka Technion. PIT Survey Oberwolfach 1 Recent Results on Polynomial Identity Testing Amir Shpilka Technion Goal of talk Survey known results Explain proof techniques Give an interesting set of `accessible open questions 2 Talk outline Definition

More information

Polynomial Identity Testing. Amir Shpilka Technion

Polynomial Identity Testing. Amir Shpilka Technion Polynomial Identity Testing Amir Shpilka Technion 1 Goal of talk Model: Arithmetic circuits Problem: Polynomial Identity Testing Example: Depth-3 circuits Some open problems 2 Holy grail: P vs. NP Boolean

More information

Tensor-Rank and Lower Bounds for Arithmetic Formulas

Tensor-Rank and Lower Bounds for Arithmetic Formulas Electronic Colloquium on Computational Complexity, Report No. 2 (2010) Tensor-Rank and Lower Bounds for Arithmetic Formulas Ran Raz Weizmann Institute Abstract We show that any explicit example for a tensor

More information

Ben Lee Volk. Joint with. Michael A. Forbes Amir Shpilka

Ben Lee Volk. Joint with. Michael A. Forbes Amir Shpilka Ben Lee Volk Joint with Michael A. Forbes Amir Shpilka Ben Lee Volk Joint with Michael A. Forbes Amir Shpilka (One) Answer: natural proofs barrier [Razborov-Rudich]: (One) Answer: natural proofs barrier

More information

A Lower Bound for the Size of Syntactically Multilinear Arithmetic Circuits

A Lower Bound for the Size of Syntactically Multilinear Arithmetic Circuits A Lower Bound for the Size of Syntactically Multilinear Arithmetic Circuits Ran Raz Amir Shpilka Amir Yehudayoff Abstract We construct an explicit polynomial f(x 1,..., x n ), with coefficients in {0,

More information

Sums of Products of Polynomials in Few Variables: Lower Bounds and Polynomial Identity Testing

Sums of Products of Polynomials in Few Variables: Lower Bounds and Polynomial Identity Testing Sums of Products of Polynomials in Few Variables: Lower Bounds and Polynomial Identity Testing Mrinal Kumar 1 and Shubhangi Saraf 2 1 Department of Computer Science, Rutgers University, New Brunswick,

More information

Tensor-Rank and Lower Bounds for Arithmetic Formulas

Tensor-Rank and Lower Bounds for Arithmetic Formulas Tensor-Rank and Lower Bounds for Arithmetic Formulas Ran Raz Weizmann Institute Abstract We show that any explicit example for a tensor A : [n] r F with tensor-rank n r (1 o(1)), where r = r(n) log n/

More information

Classifying polynomials and identity testing

Classifying polynomials and identity testing Classifying polynomials and identity testing MANINDRA AGRAWAL 1, and RAMPRASAD SAPTHARISHI 2 1 Indian Institute of Technology, Kanpur, India. 2 Chennai Mathematical Institute, and Indian Institute of Technology,

More information

Multi-Linear Formulas for Permanent and Determinant are of Super-Polynomial Size

Multi-Linear Formulas for Permanent and Determinant are of Super-Polynomial Size Multi-Linear Formulas for Permanent and Determinant are of Super-Polynomial Size Ran Raz Weizmann Institute ranraz@wisdom.weizmann.ac.il Abstract An arithmetic formula is multi-linear if the polynomial

More information

On the Structure and Complexity of Symbolic Proofs of Polynomial Identities

On the Structure and Complexity of Symbolic Proofs of Polynomial Identities On the Structure and Complexity of Symbolic Proofs of Polynomial Identities Iddo Tzameret Tel Aviv University, Tel Aviv 69978, Israel Abstract A symbolic proof for establishing that a given arithmetic

More information

Deterministic identity testing of depth 4 multilinear circuits with bounded top fan-in

Deterministic identity testing of depth 4 multilinear circuits with bounded top fan-in Deterministic identity testing of depth 4 multilinear circuits with bounded top fan-in Zohar S. Karnin Partha Mukhopadhyay Amir Shpilka Ilya Volkovich November 15, 2009 Abstract We give the first sub-exponential

More information

arxiv: v2 [cs.cc] 20 May 2018

arxiv: v2 [cs.cc] 20 May 2018 arxiv:1805.06692v2 [cs.cc] 20 May 2018 A Note on Polynomial Identity Testing for Depth-3 Circuits V. Arvind Abhranil Chatterjee Rajit Datta Partha Mukhopadhyay May 22, 2018 Abstract Let C be a depth-3

More information

Sylvester-Gallai type theorems for quadratic polynomials

Sylvester-Gallai type theorems for quadratic polynomials Sylvester-Gallai type theorems for quadratic polynomials Amir Shpilka* Abstract We prove Sylvester-Gallai type theorems for quadratic polynomials. Specifically, we prove that if a finite collection Q,

More information

Sums of read-once formulas: How many summands suffice?

Sums of read-once formulas: How many summands suffice? Sums of read-once formulas: How many summands suffice? Meena Mahajan and Anuj Tawari The Institute of Mathematical Sciences, Chennai, India. {meena,anujvt}@imsc.res.in Abstract. An arithmetic read-once

More information

Quantum Simulation. 9 December Scott Crawford Justin Bonnar

Quantum Simulation. 9 December Scott Crawford Justin Bonnar Quantum Simulation 9 December 2008 Scott Crawford Justin Bonnar Quantum Simulation Studies the efficient classical simulation of quantum circuits Understanding what is (and is not) efficiently simulable

More information

Read-Once Polynomial Identity Testing

Read-Once Polynomial Identity Testing Read-Once Polynomial Identity Testing Amir Shpilka Ilya Volkovich Abstract An arithmetic read-once formula (ROF for short) is a formula (a circuit whose underlying graph is a tree) in which the operations

More information

Reconstruction of full rank algebraic branching programs

Reconstruction of full rank algebraic branching programs Reconstruction of full rank algebraic branching programs Vineet Nair Joint work with: Neeraj Kayal, Chandan Saha, Sebastien Tavenas 1 Arithmetic circuits 2 Reconstruction problem f(x) Q[X] is an m-variate

More information

More on Noncommutative Polynomial Identity Testing

More on Noncommutative Polynomial Identity Testing More on Noncommutative Polynomial Identity Testing Andrej Bogdanov Hoeteck Wee Abstract We continue the study of noncommutative polynomial identity testing initiated by Raz and Shpilka and present efficient

More information

Arithmetic circuits: A chasm at depth three

Arithmetic circuits: A chasm at depth three Electronic Colloquium on Computational Complexity, Report No. 26 (2013) Arithmetic circuits: A chasm at depth three four Ankit Gupta Microsoft Research India t-ankitg@microsoft.com Neeraj Kayal Microsoft

More information

Arithmetic circuits: A chasm at depth three

Arithmetic circuits: A chasm at depth three Electronic Colloquium on Computational Complexity, Revision 1 of Report No. 26 (2013) Arithmetic circuits: A chasm at depth three four Ankit Gupta Microsoft Research India t-ankitg@microsoft.com Neeraj

More information

Deterministic identity testing for sum of read-once oblivious arithmetic branching programs

Deterministic identity testing for sum of read-once oblivious arithmetic branching programs Deterministic identity testing for sum of read-once oblivious arithmetic branching programs Rohit Gurjar, Arpita Korwar, Nitin Saxena, IIT Kanpur Thomas Thierauf Aalen University June 18, 2015 Gurjar,

More information

Depth-3 Arithmetic Formulae over Fields of Characteristic Zero

Depth-3 Arithmetic Formulae over Fields of Characteristic Zero Electronic olloquium on omputational omplexity, Report No 2 (1 Depth Arithmetic Formulae over Fields of haracteristic Zero Amir Shpilka Avi Wigderson Institute of omputer Science, ebrew University, Jerusalem,

More information

Witnessing Matrix Identities and Proof Complexity

Witnessing Matrix Identities and Proof Complexity Witnessing Matrix Identities and Proof Complexity Fu Li Iddo Tzameret August 4, 2016 Abstract We use results from the theory of algebras with polynomial identities PI algebras to study the witness complexity

More information

Learning Restricted Models of Arithmetic Circuits

Learning Restricted Models of Arithmetic Circuits Learning Restricted Models of Arithmetic Circuits Adam R. Klivans Department of Computer Science University of Texas at Austin Austin, TX 78712-1188 klivans@cs.utexas.edu Amir Shpilka Faculty of Computer

More information

Diagonal Circuit Identity Testing and Lower Bounds

Diagonal Circuit Identity Testing and Lower Bounds Diagonal Circuit Identity Testing and Lower Bounds Nitin Saxena Hausdorff Center for Mathematics Endenicher Allee 60 D-53115 Bonn, Germany ns@hcm.uni-bonn.de Abstract. In this paper we give the first deterministic

More information

A Non-Linear Lower Bound for Constant Depth Arithmetical Circuits via the Discrete Uncertainty Principle

A Non-Linear Lower Bound for Constant Depth Arithmetical Circuits via the Discrete Uncertainty Principle A Non-Linear Lower Bound for Constant Depth Arithmetical Circuits via the Discrete Uncertainty Principle Maurice J. Jansen Kenneth W.Regan November 10, 2006 Abstract We prove a non-linear lower bound on

More information

Separating Monotone VP and VNP

Separating Monotone VP and VNP Electronic Colloquium on Computational Complexity, Report No. 124 (2018) Separating Monotone VP and VNP Amir Yehudayoff Abstract This work is about the monotone versions of the algebraic complexity classes

More information

Homogeneous formulas and symmetric polynomials

Homogeneous formulas and symmetric polynomials Homogeneous formulas and symmetric polynomials arxiv:0907.2621v1 [cs.cc] 15 Jul 2009 Pavel Hrubeš Abstract Amir Yehudayoff We investigate the arithmetic formula complexity of the elementary symmetric polynomials

More information

Read-Once Polynomial Identity Testing

Read-Once Polynomial Identity Testing Read-Once Polynomial Identity Testing Amir Shpilka Ilya Volkovich Abstract An arithmetic read-once formula (ROF for short) is a formula (a circuit whose underlying graph is a tree) in which the operations

More information

Separating multilinear branching programs and formulas

Separating multilinear branching programs and formulas Separating multilinear branching programs and formulas Zeev Dvir Guillaume Malod Sylvain Perifel Amir Yehudayoff Abstract This work deals with the power of linear algebra in the context of multilinear

More information

Improved Polynomial Identity Testing for Read-Once Formulas

Improved Polynomial Identity Testing for Read-Once Formulas Improved Polynomial Identity Testing for Read-Once Formulas Amir Shpilka Ilya Volkovich Abstract An arithmetic read-once formula (ROF for short) is a formula (a circuit whose underlying graph is a tree)

More information

The Power of Depth 2 Circuits over Algebras

The Power of Depth 2 Circuits over Algebras The Power of Depth 2 Circuits over Algebras Chandan Saha, Ramprasad Saptharishi 2 and Nitin Saxena 3 Indian Institute of Technology, Kanpur 20806, India csaha@iitk.ac.in 2 Chennai Mathematical Institute,

More information

DETERMINISTIC POLYNOMIAL IDENTITY TESTS FOR MULTILINEAR BOUNDED-READ FORMULAE

DETERMINISTIC POLYNOMIAL IDENTITY TESTS FOR MULTILINEAR BOUNDED-READ FORMULAE DETERMINISTIC POLYNOMIAL IDENTITY TESTS FOR MULTILINEAR BOUNDED-READ FORMULAE Matthew Anderson, Dieter van Melkebeek, and Ilya Volkovich June 20, 2014 Abstract. We present a polynomial-time deterministic

More information

Lecture 3. 1 Polynomial-time algorithms for the maximum flow problem

Lecture 3. 1 Polynomial-time algorithms for the maximum flow problem ORIE 633 Network Flows August 30, 2007 Lecturer: David P. Williamson Lecture 3 Scribe: Gema Plaza-Martínez 1 Polynomial-time algorithms for the maximum flow problem 1.1 Introduction Let s turn now to considering

More information

The Strength of Multilinear Proofs

The Strength of Multilinear Proofs The Strength of Multilinear Proofs Ran Raz Iddo Tzameret December 19, 2006 Abstract We introduce an algebraic proof system that manipulates multilinear arithmetic formulas. We show that this proof system

More information

10. Noether Normalization and Hilbert s Nullstellensatz

10. Noether Normalization and Hilbert s Nullstellensatz 10. Noether Normalization and Hilbert s Nullstellensatz 91 10. Noether Normalization and Hilbert s Nullstellensatz In the last chapter we have gained much understanding for integral and finite ring extensions.

More information

Determinant Versus Permanent

Determinant Versus Permanent Determinant Versus Permanent Manindra Agrawal Abstract. We study the problem of expressing permanent of matrices as determinant of (possibly larger) matrices. This problem has close connections with complexity

More information

Arthur-Merlin Streaming Complexity

Arthur-Merlin Streaming Complexity Weizmann Institute of Science Joint work with Ran Raz Data Streams The data stream model is an abstraction commonly used for algorithms that process network traffic using sublinear space. A data stream

More information

ALGEBRAIC PROOFS OVER NONCOMMUTATIVE FORMULAS

ALGEBRAIC PROOFS OVER NONCOMMUTATIVE FORMULAS Electronic Colloquium on Computational Complexity, Report No. 97 (2010) ALGEBRAIC PROOFS OVER NONCOMMUTATIVE FORMULAS IDDO TZAMERET Abstract. We study possible formulations of algebraic propositional proof

More information

Arithmetic Circuits: A Chasm at Depth Four

Arithmetic Circuits: A Chasm at Depth Four Arithmetic Circuits: A Chasm at Depth Four Manindra Agrawal Indian Institute of Technology, Kanpur V Vinay Geodesic Information Systems Ltd, Bangalore, and Chennai Mathematical Institute, Chennai August

More information

Equivalence of Polynomial Identity Testing and Deterministic Multivariate Polynomial Factorization

Equivalence of Polynomial Identity Testing and Deterministic Multivariate Polynomial Factorization Equivalence of Polynomial Identity Testing and Deterministic Multivariate Polynomial Factorization Swastik Kopparty Shubhangi Saraf Amir Shpilka Abstract In this paper we show that the problem of deterministically

More information

An Improved Lower Bound for Depth Four Arithmetic Circuits

An Improved Lower Bound for Depth Four Arithmetic Circuits An Improved Lower Bound for Depth Four Arithmetic Circuits A THESIS SUBMITTED FOR THE DEGREE OF Master of Science (Engineering IN THE Faculty of Engineering BY Abhijat Sharma Computer Science and Automation

More information

Massachusetts Institute of Technology Department of Economics Statistics. Lecture Notes on Matrix Algebra

Massachusetts Institute of Technology Department of Economics Statistics. Lecture Notes on Matrix Algebra Massachusetts Institute of Technology Department of Economics 14.381 Statistics Guido Kuersteiner Lecture Notes on Matrix Algebra These lecture notes summarize some basic results on matrix algebra used

More information

Finite fields, randomness and complexity. Swastik Kopparty Rutgers University

Finite fields, randomness and complexity. Swastik Kopparty Rutgers University Finite fields, randomness and complexity Swastik Kopparty Rutgers University This talk Three great problems: Polynomial factorization Epsilon-biased sets Function uncorrelated with low-degree polynomials

More information

The Partial Derivative method in Arithmetic Circuit Complexity

The Partial Derivative method in Arithmetic Circuit Complexity The Partial Derivative method in Arithmetic Circuit Complexity A thesis submitted in partial fulllment of the requirements for the award of the degree of Master of Science by Yadu Vasudev Junior Research

More information

Complexity Theory of Polynomial-Time Problems

Complexity Theory of Polynomial-Time Problems Complexity Theory of Polynomial-Time Problems Lecture 3: The polynomial method Part I: Orthogonal Vectors Sebastian Krinninger Organization of lecture No lecture on 26.05. (State holiday) 2 nd exercise

More information

mathematics Smoothness in Binomial Edge Ideals Article Hamid Damadi and Farhad Rahmati *

mathematics Smoothness in Binomial Edge Ideals Article Hamid Damadi and Farhad Rahmati * mathematics Article Smoothness in Binomial Edge Ideals Hamid Damadi and Farhad Rahmati * Faculty of Mathematics and Computer Science, Amirkabir University of Technology (Tehran Polytechnic), 424 Hafez

More information

Determinant versus permanent

Determinant versus permanent Determinant versus permanent Markus Bläser, Saarland University Draft August 24, 20:13 1 Chapter 1 Valiant s framework Throughout this lecture, F will denote some field, the underlying ground field. Let

More information

Affine extractors over large fields with exponential error

Affine extractors over large fields with exponential error Affine extractors over large fields with exponential error Jean Bourgain Zeev Dvir Ethan Leeman Abstract We describe a construction of explicit affine extractors over large finite fields with exponentially

More information

Quantum NP - Cont. Classical and Quantum Computation A.Yu Kitaev, A. Shen, M. N. Vyalyi 2002

Quantum NP - Cont. Classical and Quantum Computation A.Yu Kitaev, A. Shen, M. N. Vyalyi 2002 Quantum NP - Cont. Classical and Quantum Computation A.Yu Kitaev, A. Shen, M. N. Vyalyi 2002 1 QMA - the quantum analog to MA (and NP). Definition 1 QMA. The complexity class QMA is the class of all languages

More information

Ph.D. Katarína Bellová Page 1 Mathematics 2 (10-PHY-BIPMA2) EXAM - Solutions, 20 July 2017, 10:00 12:00 All answers to be justified.

Ph.D. Katarína Bellová Page 1 Mathematics 2 (10-PHY-BIPMA2) EXAM - Solutions, 20 July 2017, 10:00 12:00 All answers to be justified. PhD Katarína Bellová Page 1 Mathematics 2 (10-PHY-BIPMA2 EXAM - Solutions, 20 July 2017, 10:00 12:00 All answers to be justified Problem 1 [ points]: For which parameters λ R does the following system

More information

Tensor Rank is Hard to Approximate

Tensor Rank is Hard to Approximate Electronic Colloquium on Computational Complexity, Report No. 86 (2018) Tensor Rank is Hard to Approximate Joseph Swernofsky - KTH Royal Institute of Technology April 20, 2018 Abstract We prove that approximating

More information

2.1 Laplacian Variants

2.1 Laplacian Variants -3 MS&E 337: Spectral Graph heory and Algorithmic Applications Spring 2015 Lecturer: Prof. Amin Saberi Lecture 2-3: 4/7/2015 Scribe: Simon Anastasiadis and Nolan Skochdopole Disclaimer: hese notes have

More information

Non-commutative computations: lower bounds and polynomial identity testing

Non-commutative computations: lower bounds and polynomial identity testing Electronic Colloquium on Computational Complexity, Report No. 94 (2016) Non-commutative computations: lower bounds and polynomial identity testing Guillaume Lagarde Guillaume Malod Sylvain Perifel June

More information

Complexity of Matrix Multiplication and Bilinear Problems

Complexity of Matrix Multiplication and Bilinear Problems Complexity of Matrix Multiplication and Bilinear Problems François Le Gall Graduate School of Informatics Kyoto University ADFOCS17 - Lecture 3 24 August 2017 Overview of the Lectures Fundamental techniques

More information

Lower Bounds on Arithmetic Circuits via Partial Derivatives

Lower Bounds on Arithmetic Circuits via Partial Derivatives Lower Bounds on Arithmetic Circuits via Partial Derivatives Noam Nisan Avi Wigderson Dedicated to the memory of Roman Smolensky Abstract In this paper we describe a new technique for obtaining lower bounds

More information

Applications. More Counting Problems. Complexity of Algorithms

Applications. More Counting Problems. Complexity of Algorithms Recurrences Applications More Counting Problems Complexity of Algorithms Part I Recurrences and Binomial Coefficients Paths in a Triangle P(0, 0) P(1, 0) P(1,1) P(2, 0) P(2,1) P(2, 2) P(3, 0) P(3,1) P(3,

More information

Chordal structure in computer algebra: Permanents

Chordal structure in computer algebra: Permanents Chordal structure in computer algebra: Permanents Diego Cifuentes Laboratory for Information and Decision Systems Electrical Engineering and Computer Science Massachusetts Institute of Technology Joint

More information

Limitations of Algorithm Power

Limitations of Algorithm Power Limitations of Algorithm Power Objectives We now move into the third and final major theme for this course. 1. Tools for analyzing algorithms. 2. Design strategies for designing algorithms. 3. Identifying

More information

Lecture 2: January 18

Lecture 2: January 18 CS271 Randomness & Computation Spring 2018 Instructor: Alistair Sinclair Lecture 2: January 18 Disclaimer: These notes have not been subjected to the usual scrutiny accorded to formal publications. They

More information

Determinant versus permanent

Determinant versus permanent Determinant versus permanent Manindra Agrawal Abstract. We study the problem of expressing permanents of matrices as determinants of (possibly larger) matrices. This problem has close connections to the

More information

Graph Sparsification III: Ramanujan Graphs, Lifts, and Interlacing Families

Graph Sparsification III: Ramanujan Graphs, Lifts, and Interlacing Families Graph Sparsification III: Ramanujan Graphs, Lifts, and Interlacing Families Nikhil Srivastava Microsoft Research India Simons Institute, August 27, 2014 The Last Two Lectures Lecture 1. Every weighted

More information

CS6840: Advanced Complexity Theory Mar 29, Lecturer: Jayalal Sarma M.N. Scribe: Dinesh K.

CS6840: Advanced Complexity Theory Mar 29, Lecturer: Jayalal Sarma M.N. Scribe: Dinesh K. CS684: Advanced Complexity Theory Mar 29, 22 Lecture 46 : Size lower bounds for AC circuits computing Parity Lecturer: Jayalal Sarma M.N. Scribe: Dinesh K. Theme: Circuit Complexity Lecture Plan: Proof

More information

Resolution of the Cap Set Problem: Progression Free Subsets of F n q are exponentially small

Resolution of the Cap Set Problem: Progression Free Subsets of F n q are exponentially small Resolution of the Cap Set Problem: Progression Free Subsets of F n q are exponentially small Jordan S. Ellenberg, Dion Gijswijt Presented by: Omri Ben-Eliezer Tel-Aviv University January 25, 2017 The Cap

More information

Monomial Testing and Applications

Monomial Testing and Applications Monomial Testing and Applications Shenshi Chen Department of Computer Science, University of Texas-Pan American, Edinburg, TX 78539, USA. schen@broncs.utpa.edu arxiv:1303.0478v2 [cs.cc] 12 Apr 2013 Abstract.

More information

Characterizing Propositional Proofs as Non-Commutative Formulas

Characterizing Propositional Proofs as Non-Commutative Formulas Characterizing Propositional Proofs as Non-Commutative Formulas Fu Li Iddo Tzameret Zhengyu Wang Abstract Does every Boolean tautology have a short propositional-calculus proof? Here, a propositionalcalculus

More information

Algebra I. Course Outline

Algebra I. Course Outline Algebra I Course Outline I. The Language of Algebra A. Variables and Expressions B. Order of Operations C. Open Sentences D. Identity and Equality Properties E. The Distributive Property F. Commutative

More information

On the centre of the generic algebra of M 1,1

On the centre of the generic algebra of M 1,1 On the centre of the generic algebra of M 1,1 Thiago Castilho de Mello University of Campinas PhD grant from CNPq, Brazil F is a field of characteristic zero; F X = F x 1, x 2,... is the free associative

More information

Black-box reconstruction of depth three circuits with top fan-in two

Black-box reconstruction of depth three circuits with top fan-in two Black-box reconstruction of depth three circuits with top fan-in two Thesis by Gaurav Sinha In Partial Fulfillment of the Requirements for the degree of Ph.D. CALIFORNIA INSTITUTE OF TECHNOLOGY Pasadena,

More information

Tensors. Notes by Mateusz Michalek and Bernd Sturmfels for the lecture on June 5, 2018, in the IMPRS Ringvorlesung Introduction to Nonlinear Algebra

Tensors. Notes by Mateusz Michalek and Bernd Sturmfels for the lecture on June 5, 2018, in the IMPRS Ringvorlesung Introduction to Nonlinear Algebra Tensors Notes by Mateusz Michalek and Bernd Sturmfels for the lecture on June 5, 2018, in the IMPRS Ringvorlesung Introduction to Nonlinear Algebra This lecture is divided into two parts. The first part,

More information

On the complexity of the permanent in various computational models

On the complexity of the permanent in various computational models On the complexity of the permanent in various computational models Christian Ikenmeyer and J.M. Landsberg October 1, 2016 Abstract We answer a question in [10], showing the regular determinantal complexity

More information

Lecture 13 Spectral Graph Algorithms

Lecture 13 Spectral Graph Algorithms COMS 995-3: Advanced Algorithms March 6, 7 Lecture 3 Spectral Graph Algorithms Instructor: Alex Andoni Scribe: Srikar Varadaraj Introduction Today s topics: Finish proof from last lecture Example of random

More information

Cyclic inclusion/exclusion

Cyclic inclusion/exclusion Cyclic inclusion/exclusion Valentin Féray LaBRI, CNRS, Bordeaux Journées Combinatoire Algébrique du GDR-IM, Rouen Valentin Féray (LaBRI, CNRS) Cyclic inclusion/exclusion Rouen, 011 06 1 / 16 Introduction

More information

Algebras with Polynomial Identities and Computing the Determinant

Algebras with Polynomial Identities and Computing the Determinant Algebras with Polynomial Identities and Computing the Determinant Steve Chien Microsoft Research 1065 La Avenida Mountain View, CA 94043, U.S.A. schien@microsoft.com Alistair Sinclair Computer Science

More information

Linear Algebra. Min Yan

Linear Algebra. Min Yan Linear Algebra Min Yan January 2, 2018 2 Contents 1 Vector Space 7 1.1 Definition................................. 7 1.1.1 Axioms of Vector Space..................... 7 1.1.2 Consequence of Axiom......................

More information

Constant-Depth Circuits for Arithmetic in Finite Fields of Characteristic Two

Constant-Depth Circuits for Arithmetic in Finite Fields of Characteristic Two Constant-Depth Circuits for Arithmetic in Finite Fields of Characteristic Two Alexander Healy Emanuele Viola August 8, 2005 Abstract We study the complexity of arithmetic in finite fields of characteristic

More information

Algebra Performance Level Descriptors

Algebra Performance Level Descriptors Limited A student performing at the Limited Level demonstrates a minimal command of Ohio s Learning Standards for Algebra. A student at this level has an emerging ability to A student whose performance

More information

Local Maxima and Improved Exact Algorithm for MAX-2-SAT

Local Maxima and Improved Exact Algorithm for MAX-2-SAT CHICAGO JOURNAL OF THEORETICAL COMPUTER SCIENCE 2018, Article 02, pages 1 22 http://cjtcs.cs.uchicago.edu/ Local Maxima and Improved Exact Algorithm for MAX-2-SAT Matthew B. Hastings Received March 8,

More information

Expander Construction in VNC 1

Expander Construction in VNC 1 Expander Construction in VNC 1 Sam Buss joint work with Valentine Kabanets, Antonina Kolokolova & Michal Koucký Prague Workshop on Bounded Arithmetic November 2-3, 2017 Talk outline I. Combinatorial construction

More information

A Robust Form of Kruskal s Identifiability Theorem

A Robust Form of Kruskal s Identifiability Theorem A Robust Form of Kruskal s Identifiability Theorem Aditya Bhaskara (Google NYC) Joint work with Moses Charikar (Princeton University) Aravindan Vijayaraghavan (NYU Northwestern) Background: understanding

More information

Ranks of Real Symmetric Tensors

Ranks of Real Symmetric Tensors Ranks of Real Symmetric Tensors Greg Blekherman SIAM AG 2013 Algebraic Geometry of Tensor Decompositions Real Symmetric Tensor Decompositions Let f be a form of degree d in R[x 1,..., x n ]. We would like

More information

Polynomial Identity Testing and Lower Bounds for Sum of Special Arithmetic Branching Programs

Polynomial Identity Testing and Lower Bounds for Sum of Special Arithmetic Branching Programs Polynomial Identity Testing and Lower Bounds for Sum of Special Arithmetic Branching Programs A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy by Arpita

More information

Explicit Noether Normalization for Simultaneous Conjugation via Polynomial Identity Testing

Explicit Noether Normalization for Simultaneous Conjugation via Polynomial Identity Testing Explicit Noether Normalization for Simultaneous Conjugation via Polynomial Identity Testing Michael A. Forbes Amir Shpilka March 7, 2013 Abstract Mulmuley [Mul12a] recently gave an explicit version of

More information

EE/ACM Applications of Convex Optimization in Signal Processing and Communications Lecture 18

EE/ACM Applications of Convex Optimization in Signal Processing and Communications Lecture 18 EE/ACM 150 - Applications of Convex Optimization in Signal Processing and Communications Lecture 18 Andre Tkacenko Signal Processing Research Group Jet Propulsion Laboratory May 31, 2012 Andre Tkacenko

More information

Majority is incompressible by AC 0 [p] circuits

Majority is incompressible by AC 0 [p] circuits Majority is incompressible by AC 0 [p] circuits Igor Carboni Oliveira Columbia University Joint work with Rahul Santhanam (Univ. Edinburgh) 1 Part 1 Background, Examples, and Motivation 2 Basic Definitions

More information

Conflict-Free Colorings of Rectangles Ranges

Conflict-Free Colorings of Rectangles Ranges Conflict-Free Colorings of Rectangles Ranges Khaled Elbassioni Nabil H. Mustafa Max-Planck-Institut für Informatik, Saarbrücken, Germany felbassio, nmustafag@mpi-sb.mpg.de Abstract. Given the range space

More information

Discrete Optimization 2010 Lecture 10 P, N P, and N PCompleteness

Discrete Optimization 2010 Lecture 10 P, N P, and N PCompleteness Discrete Optimization 2010 Lecture 10 P, N P, and N PCompleteness Marc Uetz University of Twente m.uetz@utwente.nl Lecture 9: sheet 1 / 31 Marc Uetz Discrete Optimization Outline 1 N P and co-n P 2 N P-completeness

More information

Rank Bound for Depth-3 Identities

Rank Bound for Depth-3 Identities Rank Bound for Depth-3 Identities Nitin Saxena (Hausdorff Center for Mathematics, Bonn) Joint work with C. Seshadhri (IBM Almaden Research Center, San Jose) 1 The problem of PIT Polynomial identity testing:

More information

IITM-CS6845: Theory Toolkit February 3, 2012

IITM-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 information

Neighborly families of boxes and bipartite coverings

Neighborly families of boxes and bipartite coverings Neighborly families of boxes and bipartite coverings Noga Alon Dedicated to Professor Paul Erdős on the occasion of his 80 th birthday Abstract A bipartite covering of order k of the complete graph K n

More information

ENTROPY OF OPERATORS OR WHY MATRIX MULTIPLICATION IS HARD FOR DEPTH-TWO CIRCUITS

ENTROPY OF OPERATORS OR WHY MATRIX MULTIPLICATION IS HARD FOR DEPTH-TWO CIRCUITS Electronic Colloquium on Computational Complexity, Revision 1 of Report No. 19 (2008) ENTROPY OF OPERATORS OR WHY MATRIX MULTIPLICATION IS HARD FOR DEPTH-TWO CIRCUITS STASYS JUKNA Abstract. We consider

More information

Quiver mutations. Tensor diagrams and cluster algebras

Quiver mutations. Tensor diagrams and cluster algebras Maurice Auslander Distinguished Lectures April 20-21, 2013 Sergey Fomin (University of Michigan) Quiver mutations based on joint work with Andrei Zelevinsky Tensor diagrams and cluster algebras based on

More information

(Permutations Arising from) Hook Coefficients of Chromatic Symmetric Functions

(Permutations Arising from) Hook Coefficients of Chromatic Symmetric Functions (Permutations Arising from) Hook Coefficients of Chromatic Symmetric Functions Ryan Kaliszewski Drexel University rlk72@drexel.edu July 17, 2014 Ryan Kaliszewski (Drexel University) Hook Coefficients July

More information

Topics in linear algebra

Topics in linear algebra Chapter 6 Topics in linear algebra 6.1 Change of basis I want to remind you of one of the basic ideas in linear algebra: change of basis. Let F be a field, V and W be finite dimensional vector spaces over

More information

Algorithmic Approach to Counting of Certain Types m-ary Partitions

Algorithmic Approach to Counting of Certain Types m-ary Partitions Algorithmic Approach to Counting of Certain Types m-ary Partitions Valentin P. Bakoev Abstract Partitions of integers of the type m n as a sum of powers of m (the so called m-ary partitions) and their

More information

RATIONAL REALIZATION OF MAXIMUM EIGENVALUE MULTIPLICITY OF SYMMETRIC TREE SIGN PATTERNS. February 6, 2006

RATIONAL REALIZATION OF MAXIMUM EIGENVALUE MULTIPLICITY OF SYMMETRIC TREE SIGN PATTERNS. February 6, 2006 RATIONAL REALIZATION OF MAXIMUM EIGENVALUE MULTIPLICITY OF SYMMETRIC TREE SIGN PATTERNS ATOSHI CHOWDHURY, LESLIE HOGBEN, JUDE MELANCON, AND RANA MIKKELSON February 6, 006 Abstract. A sign pattern is a

More information