Share
Embed
The Active Subspace is Nearly Stationary
Presentation at 15th International Conference on Approximation Theory, San Antonio, May 2016
Paul Constantine
May 24, 2016
Transcript
19 slides
-
ACTIVE SUBSPACES for dimension reduction in parameter studies This material
is based upon work supported by the U.S. Department of Energy Office of Science, Office of Advanced Scientific Computing Research, Applied Mathematics program under Award Number DE-SC-0011077. PAUL CONSTANTINE Ben L. Fryrear Assistant Professor Applied Mathematics & Statistics Colorado School of Mines activesubspaces.org! @DrPaulynomial! SLIDES: DISCLAIMER: These slides are meant to complement the oral presentation. Use out of context at your own risk. In collaboration with: RACHEL WARD UT Austin ARMIN EFTEKHARI UT Austin -
f( x ) ⇡ g(UT x ) Ridge approximation UT
: Rm ! Rn g : Rn ! R where -
f( x ) ⇡ g(UT x ) Ridge approximation What
is U? What is the approximation error? What is g? -
f( x ) ⇡ g(UT x ) Ridge approximation What
is the approximation error? f( x ) g(UT x ) L2(⇢) = ✓Z (f( x ) g(UT x ))2 ⇢( x ) d x ◆1 2 Use the weighted root-mean-squared error: Given weight function -
µ(y) = Z f(Uy + V z) ⇡(z|y) dz f(
x ) ⇡ g(UT x ) Ridge approximation Use the conditional average: Subspace coordinates What is g? Complement subspace and coordinates Conditional density µ(UT x ) is the best approximation (Pinkus, 2015). -
f( x ) ⇡ g(UT x ) Ridge approximation What
is U? Define the error function: R(U) = 1 2 Z (f( x ) µ(UT x ))2 ⇢( x ) d x Minimize the error: minimize U R ( U ) subject to U 2 G ( n, m ) Grassmann manifold of n-dimensional subspaces -
Define the active subspace Consider a function and its gradient
vector, The average outer product of the gradient and its eigendecomposition, Partition the eigendecomposition, Rotate and separate the coordinates, ⇤ = ⇤1 ⇤2 , W = ⇥ W 1 W 2 ⇤ , W 1 2 Rm⇥n x = W W T x = W 1W T 1 x + W 2W T 2 x = W 1y + W 2z active variables inactive variables f = f( x ), x 2 Rm, rf( x ) 2 Rm, ⇢ : Rm ! R + C = Z rf rfT ⇢ d x = W ⇤W T -
i = Z rfT wi 2 ⇢ d x ,
i = 1, . . . , m The eigenpairs identify perturbations that change the function more, on average. LEMMA LEMMA Z (ryf)T (ryf) ⇢ d x = 1 + · · · + n Z (rzf)T (rzf) ⇢ d x = n+1 + · · · + m -
An approximation result Conditional average Active subspace Poincaré constant Eigenvalues
associated with inactive subspace f( x ) µ(W T 1 x ) L2(⇢) C ( n+1 + · · · + m)1 2 -
k ¯ rR(W 1)kF L ⇣ 2m1 2 +
(m n)1 2 ⌘ ( n+1 + · · · + m)1 2 The active subspace is nearly stationary. R(U) = 1 2 Z (f( x ) µ(UT x ))2 ⇢( x ) d x Recall: Assume (1) Lipschitz continuous function (2) Gaussian density function Gradient on the Grassmann manifold Active subspace Frobenius norm Dimensions Lipschitz constant Eigenvalues associated with inactive subspace -
IDEA Use active subspace as the starting point for numerical
ridge approximation. U0 Given an initial subspace and samples (1) Compute (2) Fit a polynomial with the pairs (3) Minimize residual over subpsaces (4) Set and repeat { xi, f( xi)} { yi, f( xi)} pN (y, ✓) U⇤ = argmin U2G(n,m) X i ⇣ f( xi) p(UT xi, ✓⇤) ⌘2 ✓⇤ = argmin ✓ X i f( xi) p( yi, ✓) 2 yi = UT 0 xi U0 = U⇤ -
An example where it doesn’t work 0 π/4 π/2 3π/4
π SUBSPACE ANGLE 0 5 10 15 20 25 30 ERROR Active subspace U = [0; 1] Inactive subspace U = [1; 0] f ( x1, x2) = 5 x1 + sin(10 ⇡x2) C = 25 0 0 526 -
An example where it works DRAG COEFFICIENT as a function
of 18 shape parameters Uniform on a hypercube SU2 CFD solver with adjoint solver for gradients -
An example where it works C = Z rf rfT
⇢ d x = W ⇤W T Recall: DRAG COEFFICIENT as a function of 18 shape parameters Uniform on a hypercube SU2 CFD solver with adjoint solver for gradients -
Residual as a function of alternating iteration for different starting
subspaces RANDOM IDENTITY ACTIVE SUBSPACE Increasing polynomial degree (total) Increasing subspace dimension Residual Alternating iteration -
Residual as a function of alternating iteration for different starting
subspaces RANDOM IDENTITY ACTIVE SUBSPACE Increasing polynomial degree (total) Increasing subspace dimension -
How do active subspaces relate to [insert method]? What if
I don’t have gradients? What kinds of models does this work on? PAUL CONSTANTINE Ben L. Fryrear Assistant Professor Colorado School of Mines activesubspaces.org! @DrPaulynomial! QUESTIONS? Active Subspaces SIAM (2015)