How Can Deep Rectifier Networks Achieve Linear Separability and Preserve Distances?

Authors: Senjian An, Farid Boussaid, Mohammed Bennamoun

ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical This paper investigates how hidden layers of deep rectifier networks are capable of transforming two or more pattern sets to be linearly separable while preserving the distances with a guaranteed degree, and proves the universal classification power of such distance preserving rectifier networks.Our proof is constructive, with the aid of a new proposed data model, and explains the strategies of RLTs in transforming linearly inseparable data to be linearly separable.This work has focused on theoretically analysing the universal classification power and the distance preserving properties of rectifier networks, and providing an insightful explanation for the recent successes of rectifier networks in practice.
Researcher Affiliation Academia Senjian An SENJIAN.AN@UWA.EDU.AU School of Computer Science and Software Engineering, The University of Western Australia, Australia Farid Boussaid FARID.BOUSSAID@UWA.EDU.AU School of Electrical, Electronic and Computer Engineering, The University of Western Australia, Australia Mohammed Bennamoun MOHAMMED.BENNAMOUN@UWA.EDU.AU School of Computer Science and Software Engineering, The University of Western Australia, Australia
Pseudocode No The paper describes mathematical proofs and theorems but does not include any pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any specific links or statements regarding the release of open-source code for the described methodology.
Open Datasets No The paper is theoretical and does not describe using any datasets for training or evaluation, hence no information on public dataset availability.
Dataset Splits No This is a theoretical paper and does not involve empirical validation or dataset splits for training, validation, or testing.
Hardware Specification No The paper does not describe any specific hardware used, as it focuses on theoretical analysis rather than empirical experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers, as it is a theoretical work and does not detail an implementation.
Experiment Setup No The paper is theoretical and does not provide details about an experimental setup, hyperparameters, or training configurations.