# Speedup from a different parametrization within the Neural Network algorithm

## TL;DR

This paper introduces a new parametrization for hyperplanes in neural networks that improves training efficiency and performance, making the model easier to understand and initialize, as demonstrated on autoencoder examples.

## Contribution

It proposes a novel hyperplane parametrization in neural networks that outperforms traditional methods in training speed and accuracy.

## Key findings

- Lower training error with fewer epochs
- Faster convergence compared to usual parametrization
- Easier understanding and initialization of parameters

## Abstract

A different parametrization of the hyperplanes is used in the neural network algorithm. As demonstrated on several autoencoder examples it significantly outperforms the usual parametrization, reaching lower training error values with only a fraction of the number of epochs. It's argued that it makes it easier to understand and initialize the parameters.