Skip to content

Latest commit

 

History

History
26 lines (16 loc) · 1.1 KB

plq.md

File metadata and controls

26 lines (16 loc) · 1.1 KB

PLQ Composite Decomposition

Background

As indicated in Theorem 1 of 1, any positive convex piecewise linear-quadratic (PLQ) function can be decompose as a composite ReLU-ReHU function. The objective of this project is geared towards creating an automated Python and R library capable of converting a PLQ loss function into a ReLU-ReHU function. This will then permit the direct application of ReHLine to resolve the corresponding Empirical Risk Minimization (ERM) problem.

  • Mentors: Ben Dai
  • Time Period: 3 Months
  • Languages: Python and R
  • Position: RA@CUHK-STAT
  • Salary: ~HKD$25,000 per month

Skills Required

  • Proficiency in programming using Python, R, and LaTex;
  • Patience in drafting the documentation for the library.
  • Familiarities with the construction of Python/R packages;

Related Work

N.A.

Reference

Footnotes

  1. Dai, B., & Qiu, Y. (2023, November). ReHLine: Regularized Composite ReLU-ReHU Loss Minimization with Linear Computation and Linear Convergence. In Thirty-seventh Conference on Neural Information Processing Systems. ↩