Autoencoders

No Thumbnail Available

Authors

Davidson, Shirley

Date of Issue

2021-04-22

Type

Presentation

Language

en_US

Subject Keywords

Research Projects

Organizational Units

Journal Issue

Other Titles

Abstract

A Neural Network is a statistical tool that aids in the classification of data points for machine learning. Autoencoders are a type of neural network that are capable of learning large amounts of data without being supervised by labeled training data. We use autoencoders to help with the curse of dimensionality, a common data science problem that involves an excess number of predicting variables. Autoencoders can build models while only using the most important information. In this presentation, we are going to look at three different types of autoencoders and analyze the results to see which performs better. The different methods we will explore are stacked, convolutional, and denoising. We will apply these autoencoders to an image classification problem, intending to achieve data compression.

Description

Citation

Publisher

License

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN