Files

 main

/

autoencoder

/

Copy path

Directory actions

More options

Directory actions

More options

Latest commit

Sebastian Schmidl

Adapt algorithms with names starting with [a-d] (#28)

Nov 30, 2023

00769de · Nov 30, 2023

History

History

/

autoencoder

/

Top

Folders and files

Name		Name	Last commit message	Last commit date
parent directory ..
.gitignore		.gitignore	Publish timeeval algorithms	Jan 5, 2022
Dockerfile		Dockerfile	Adapt algorithms with names starting with [a-d] (#28)	Nov 30, 2023
LICENSE		LICENSE	Publish timeeval algorithms	Jan 5, 2022
README.md		README.md	Publish timeeval algorithms	Jan 5, 2022
algorithm.py		algorithm.py	Publish timeeval algorithms	Jan 5, 2022
manifest.json		manifest.json	Adapt algorithms with names starting with [a-d] (#28)	Nov 30, 2023
model.py		model.py	Publish timeeval algorithms	Jan 5, 2022
requirements.txt		requirements.txt	fix: build of AE and DAE by pinning protobuf version (newer version l…	Aug 18, 2022
View all files

README.md

AutoEncoder (AE)

citekey	SakuradaYairi2014Anomaly
source code	own
Learning type	semi-supervised
input	multivariate

Notes

• AutoEncoder is trained on all non-anomaly data. Whenever it encounters an anomaly value, the reproduction error is quite higher than the error with non-anomaly instances.
• The test data is put into the AutoEncoder and the scores are returned.
• An assumption is made that all errors are normally distributed with some mean and std. Any error value that follows mean + kstd > threshold or mean - kstd < thereshold is considered as an anomaly. (NOT IMPLEMENTED)