solving polynomial systems - math.tamu.edu · eigenvalues are solutions to polynomial equations...
TRANSCRIPT
![Page 1: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/1.jpg)
Solving Polynomial Systems
Taylor Brysiewicz (Texas A&M University)
February 16, 2019
GIGEM 2019Texas A&M University
![Page 2: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/2.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 3: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/3.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 4: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/4.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 5: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/5.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 6: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/6.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]
Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 7: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/7.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]
Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 8: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/8.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]
Kinematics [HSW17]
Many many other places
![Page 9: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/9.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 10: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/10.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 11: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/11.jpg)
Polynomial Systems
Polynomial systems show up all of the time
Algebra (Algebraic geometry)
Combinatorics (Toric varieties)
Cryptography (Elliptic curve cryptography)
Applications
Optimization [Ble12]Frame theory [CS]Chemical reaction networks [Fei19]Kinematics [HSW17]
Many many other places
![Page 12: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/12.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 13: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/13.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 14: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/14.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 15: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/15.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 16: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/16.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 17: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/17.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 18: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/18.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 19: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/19.jpg)
Polynomials in GIG’EM so far
Eigenvalues are solutions to polynomial equations (Amudhan)
Number of solutions to my favorite polynomial system is a multinomialcoefficient (Pablo...)
We saw q-polynomials. Honorable mentions: Bessel, Jacobi, Hahn, Laguerrepolynomials (Gallen)
Compact quantum groups like SUq(2) are real parts of solutions to polynomialequations (John).
Amoebas come from polynomials (Nida).
Taylor approximations (are polynomials) used for Ito’s Lemma (Sean).
You can encode neural networks being convex as vanishing sets being emptyover finite fields (Alex).
Border rank of matrix multiplication tensors come from polynomial equations(Kashif)
![Page 20: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/20.jpg)
Polynomial Systems
K: a fieldn ∈ N: number of variablesR = K[x1, . . . , xn]: polynomial ringf1, . . . , fm ∈ R: some polynomials
Goal: Understand the set
V (f1, . . . , fm) = {(x1, . . . , xn) ∈ Kn|fi (x1, . . . , xn) = 0∀i}
Subgoal: Understand the case when V (f1, . . . , fm) consists of finitely many points.(zero dimensional polynomial solving)
(for this talk, we will assume there are finitely many solutions)
![Page 21: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/21.jpg)
Starting easy
Easy case: One polynomial f (x) =d∑
i=0
cixi ∈ K[x ] in one variable.
How many roots are there?:
If K is algebraically closed: d with multiplicity.This talk: K will be C
But wait...what does it even mean to solve a system?(1) Give algebraic numbers via minimal polynomial(2) Give approximations of numbers as decimals with option to sharpen
![Page 22: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/22.jpg)
Starting easy
Easy case: One polynomial f (x) =d∑
i=0
cixi ∈ K[x ] in one variable.
How many roots are there?:If K is algebraically closed: d with multiplicity.
This talk: K will be C
But wait...what does it even mean to solve a system?(1) Give algebraic numbers via minimal polynomial(2) Give approximations of numbers as decimals with option to sharpen
![Page 23: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/23.jpg)
Starting easy
Easy case: One polynomial f (x) =d∑
i=0
cixi ∈ K[x ] in one variable.
How many roots are there?:If K is algebraically closed: d with multiplicity.This talk: K will be C
But wait...what does it even mean to solve a system?(1) Give algebraic numbers via minimal polynomial(2) Give approximations of numbers as decimals with option to sharpen
![Page 24: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/24.jpg)
Starting easy
Easy case: One polynomial f (x) =d∑
i=0
cixi ∈ K[x ] in one variable.
How many roots are there?:If K is algebraically closed: d with multiplicity.This talk: K will be C
But wait...what does it even mean to solve a system?
(1) Give algebraic numbers via minimal polynomial(2) Give approximations of numbers as decimals with option to sharpen
![Page 25: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/25.jpg)
Starting easy
Easy case: One polynomial f (x) =d∑
i=0
cixi ∈ K[x ] in one variable.
How many roots are there?:If K is algebraically closed: d with multiplicity.This talk: K will be C
But wait...what does it even mean to solve a system?(1) Give algebraic numbers via minimal polynomial
(2) Give approximations of numbers as decimals with option to sharpen
![Page 26: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/26.jpg)
Starting easy
Easy case: One polynomial f (x) =d∑
i=0
cixi ∈ K[x ] in one variable.
How many roots are there?:If K is algebraically closed: d with multiplicity.This talk: K will be C
But wait...what does it even mean to solve a system?(1) Give algebraic numbers via minimal polynomial(2) Give approximations of numbers as decimals with option to sharpen
![Page 27: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/27.jpg)
Solving symbolically
Input:f1, . . . , fm ∈ C[x1, . . . , xn]
1 Set I = 〈f1, . . . , fm〉2 Compute a Grobner basis g1, . . . , gM for I (with respect to lexicographic
monomial ordering)
3 gM will be a polynomial in xn only. Solve this univariate polynomial to computeall possible last coordinates of a solution.
4 gM−1 will only have xn−1 and xn, subsitute each possible xn to produce aunivariate polynomial in xn−1
5 et cetera
Output: A list of equations with the same solutions such that you can backsubstitute one variable at a time.
![Page 28: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/28.jpg)
Symbolic example
f = x2 + y2 − 5, g = xy − 2
Tell Macaulay2 or some other program to compute a Grobner basisGrobner basis:
{2x + y3 − 5y , y4 − 5y2 + 4}Possible y coordinates: {−1, 1,−2, 2}Solutions: (1, 2), (1,−2), (−1, 2), (−2, 1)
Software: Macaulay2 [GS02], SAGE [Ste19], Singular, Maxima, Mathematica
![Page 29: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/29.jpg)
Solving numerically
Idea: You give me some system to solve. I (for a moment) ignore your request andsolve my own. Then a deform my solutions to yours.Example: You give me
f = x2 + y2 − 5, g = xy − 2
I solve my favorite instead
f = x2 − 1, g = y2 − 1
Then I deform along a homotopy
F = (t)f + (1− t)f ,G = (t)g + (1− t)g
This is done by solving an ordinary differential equation using predictor correctormethods.Software: Bertini [BHSW], PHCpack [Ver], NumericalAlgebraicGeometry.m2[Ley09], HomotopyContinuation.jl [Bre18]
![Page 30: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/30.jpg)
Homotopy picture
Remark: Such a start system always has∏
deg(fi ) solutions. But the end system
may not. This means some go “off to infinity” during the homotopy.Remark: There are different start systems depending on the structure of yourequations, but many times you will have an overcount of the number of solutions.
![Page 31: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/31.jpg)
Monodromy Algorithm
Computing solutions to a system via monodromy
![Page 32: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/32.jpg)
Monodromy Algorithm
Suppose we know a point on a torus.
![Page 33: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/33.jpg)
Monodromy Algorithm
Suppose we know a point on a torus.
![Page 34: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/34.jpg)
Monodromy Algorithm
Take a random linear slice through that point.
![Page 35: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/35.jpg)
Monodromy Algorithm
Take a random linear slice through that point.
![Page 36: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/36.jpg)
Monodromy Algorithm
Pick another random linear slice.
![Page 37: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/37.jpg)
Monodromy Algorithm
Pick another random linear slice.
![Page 38: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/38.jpg)
Monodromy Algorithm
Use homotopy continuation to follow your known point to the new slice.
![Page 39: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/39.jpg)
Monodromy Algorithm
Use homotopy continuation to follow your known point to the new slice.
![Page 40: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/40.jpg)
Monodromy Algorithm
Repeat
![Page 41: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/41.jpg)
Monodromy Algorithm
Repeat
![Page 42: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/42.jpg)
Monodromy Algorithm
Repeat
![Page 43: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/43.jpg)
Monodromy Algorithm
Then return to the original slice
![Page 44: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/44.jpg)
Monodromy Algorithm
Look! Another point!
![Page 45: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/45.jpg)
Monodromy
Monodromy works with nonlinear varieties too
You don’t overcompute too much (contrary to using homotopy method)
The software for this method is quite good:
MonodromySolve.m2HomotopyContinuation.jl
Depending on situation, one can boost this algorithm a lot (if you know there isa natural action on your solutions, like complex conjugation)
![Page 46: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/46.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a timeFinite fields Works well Not that I know of
![Page 47: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/47.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter much
degrees of polynomials matters a bit mattersParallelizable No Yes
Works well over C Not really... YesEasy to use Definitely Takes some engineering
Largest system I’ve solved ≈ 5000 ≈ 200000Can certify real roots Yes Yes
Can certify rational roots Yes Working on itMetaphor Download whole book Read page at a time
Finite fields Works well Not that I know of
![Page 48: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/48.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a timeFinite fields Works well Not that I know of
![Page 49: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/49.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No Yes
Works well over C Not really... YesEasy to use Definitely Takes some engineering
Largest system I’ve solved ≈ 5000 ≈ 200000Can certify real roots Yes Yes
Can certify rational roots Yes Working on itMetaphor Download whole book Read page at a time
Finite fields Works well Not that I know of
![Page 50: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/50.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a timeFinite fields Works well Not that I know of
![Page 51: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/51.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineering
Largest system I’ve solved ≈ 5000 ≈ 200000Can certify real roots Yes Yes
Can certify rational roots Yes Working on itMetaphor Download whole book Read page at a time
Finite fields Works well Not that I know of
![Page 52: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/52.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a timeFinite fields Works well Not that I know of
![Page 53: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/53.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes Yes
Can certify rational roots Yes Working on itMetaphor Download whole book Read page at a time
Finite fields Works well Not that I know of
![Page 54: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/54.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a timeFinite fields Works well Not that I know of
![Page 55: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/55.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a time
Finite fields Works well Not that I know of
![Page 56: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/56.jpg)
Comparisons
Comparison tableMethod Symbolic NumericalExact Yes Not really (but certifiable)
# variables matters a lot doesn’t matter muchdegrees of polynomials matters a bit matters
Parallelizable No YesWorks well over C Not really... Yes
Easy to use Definitely Takes some engineeringLargest system I’ve solved ≈ 5000 ≈ 200000
Can certify real roots Yes YesCan certify rational roots Yes Working on it
Metaphor Download whole book Read page at a timeFinite fields Works well Not that I know of
![Page 57: Solving Polynomial Systems - math.tamu.edu · Eigenvalues are solutions to polynomial equations (Amudhan) Number of solutions to my favorite polynomial system is a multinomial coe](https://reader036.vdocuments.site/reader036/viewer/2022071022/5fd60af90cc0c00e165a0ad2/html5/thumbnails/57.jpg)
Works Cited
Daniel J Bates, Jonathan D Hauenstein, Andrew J Sommese, and Charles W Wampler, Bertini: Software for numerical algebraic geometry (2006),
Software available at http://bertini. nd. edu.
Thomas Blekherman, Pariilo, Semidefinite optimization and convex algebraic geometry, Optimization, vol. 13, SIAM, 2012.
Timme S. Breiding, P., Homotopycontinuation.jl, arXiv preprint arXiv:1711.10911 (2018).
J. Cahill and N. Strawn, Algebraic geometry and finite frames, SIAM J. Appl. Algebra Geom.
Martin Feinberg, Foundations of chemical reaction network theory, vol. 202, Springer, 2019.
Daniel R Grayson and Michael E Stillman, Macaulay 2, a software system for research in algebraic geometry, 2002.
J. Hauenstein, S. Sherman, and C. Wampler, Exceptional stewart-gough platforms, segre embeddings, and the special euclidean group.
A. Leykin, Numerical algebraic geometry for macaulay2, arXiv preprint arXiv:0911.1783 (2009).
W. Stein, Sage: A computer system for algebra and geometry experimentation.
Jan Verschelde, Phcpack: a general-purpose solver for polynomial systems by homotopy continuation.