Piecewise Linear Multilayer Perceptrons and Dropout

TL;DR

This paper introduces a novel piecewise linear hidden layer for multilayer perceptrons, achieving state-of-the-art performance on the MNIST dataset.

Contribution

It presents a new type of hidden layer for MLPs that improves performance on image classification tasks.

Findings

Achieved the best reported performance on MNIST with the new layer.

Demonstrated the effectiveness of the piecewise linear layer.

Potential implications for neural network design.

Abstract

We propose a new type of hidden layer for a multilayer perceptron, and demonstrate that it obtains the best reported performance for an MLP on the MNIST dataset.