Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Optimal Transport (Marco Cuturi) #2

Open
YugeTen opened this issue Aug 30, 2019 · 0 comments
Open

Optimal Transport (Marco Cuturi) #2

YugeTen opened this issue Aug 30, 2019 · 0 comments

Comments

@YugeTen
Copy link
Owner

YugeTen commented Aug 30, 2019

Optimal transport

(see lecture notes and tutorial)

Introduction

Two examples:

  • Moving earth & Soldiers Monge problem: what is the most efficient way to bring earth from one place to another

drawing

How to move the sand to fill the hole most efficiently? Characterise the work involved here by the product between mass and the moving distance:

drawing

drawing

drawing

Exact solution: linear programming

image

Sinkhorn Algorithm for Entropy Regularized Optimal Transport

image

Dealing with curse of high dimensionality

Sliced Wasserstein distance:

drawing

PCA projection

drawing

k-dim (robust) projection

drawing

Applications: Average measures

k-means

You can consider W distance as a k-mean algorithm, if the dimension of X is much higher than the dimension of Y.

Wasserstein Barycenter:

drawing

drawing

Brain imaging

Mapping visual stimulus to different cortex of the brain using MEG. Different subject will have slightly different response -- to account for this spatial variation one can use Wasserstein average.

KL, MMD vs. Wasserstein

there is no geometry in KL/MMD, they're better for high dimensional data. For KL, when you have two Gaussians very close to each other, if the variance goes to zero, the divergence can go to infinity

Wasserstein vs. L2 averages

drawing

Domain adaptation

drawing

Learning with Wasserstein loss:

drawing

drawing

drawing

Sorting

On 1D, calculating the Wasserstein plan is equivalent to sorting (because the nth ranking point in X is always going to map to the nth ranking in Y)

drawing

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant