Skip to content

Python 3 implementation of Non-linear Metric Learning (Kedem et. al)

Notifications You must be signed in to change notification settings

jeppe742/nonLinearMetricLearning

Repository files navigation

Non-Linear Metric Learning

This project is the result of the "02460 ADVANCED MACHINE LEARNING" course at DTU.

The code implements the NLMNN algorithm by Kedem et. al., and runs experiements on two different datasets.

Simplex projection

During experiments it was found, that the softmax transformation, proposed by Kedem et. al, had several issues, like slow convergence and local minimas.

Inspired by Yang et. al, we implemet an optimization scheme based on taking the full gradient step, and afterwards do a projection back into the simplex.

See the report for more details.

About

Python 3 implementation of Non-linear Metric Learning (Kedem et. al)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published