.png){kind=icon}
.png){kind=icon}
Return
Mixed integer quadratically constrained quadratic programming for neural network Lipschitz constant computation
.png){kind=icon}
OA
AI
DOI：10.1016/j.ejor.2025.12.019.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Mathematical optimization standpoint to derive new neural network Lipschitz constant lower and upper bounds • MIQCQP formulations for efficient Lipschitz constant estimation • Experimental evidence that MIQP models for one hidden layer architecture are better choices for L2 and L? norms • The Lipschitz constant is proven to be the solution of a MIQCQP almost everywhere in the parameter space of neural networks
Keywords:
Lipschitz constant
neural network
quadratic mixed integer programming
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
6
Papers:
2.2W
Citations:
6.4W