8000 GitHub - lorenzolucchese/esig: Expected Signature Estimators.
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

lorenzolucchese/esig

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
lib
lib
 
 
.gitignore
.gitignore
 
 
README.md
README.md
 
 
empirics.ipynb
empirics.ipynb
 
 
estimate.py
estimate.py
 
 
simulate.py
simulate.py
 
 
simulations.ipynb
simulations.ipynb
 
 

Repository files navigation

Expected Signature Estimators

This repo provides code for estimating expected signatures

$$\phi(T) := \mathbb{E}\Big[S(\mathbb{X})_{[0,T]}\Big]\in T((\mathbb{R}^d)),$$

from a collection of paths $\mathbb{X}^{\pi, 1}, \ldots, \mathbb{X}^{\pi, n}$. The paths may be obtained by chopping-and-shifting a single long observation.

We implement both the naive estimator

$$\hat{\phi}^{N, \pi}(T) := \frac{1}{N} \sum_{n=1}^N S\left(\mathbb{X}^{n, \pi}\right)_{[0,T]},$$

and the martingale-corrected estimator

$$\hat{\phi}^{N, \pi, c}(T) := \frac{1}{N} \sum_{n=1}^N S\left(\mathbb{X}^{n, \pi}\right)_{[0,T]} + \hat{c} S^c\left(\mathbb{X}^{n, \pi}\right)_{[0,T]},$$

where $S^c\left(\mathbb{X}^{n, \pi}\right)_{[0,T]}$ is an Ito correction.

The code is compatible with numpy arrays and torch tensors, using iisignature and signatory for signature computations.

These functions can be used directly into more general ML pipelines/models, as illustrated in the forks:

About

Expected Signature Estimators.

Resources

Readme
Activity

Stars

0 stars

Watchers

1 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Languages
0