Saved in:
Bibliographic Details
Main Authors: González, Gastón García, Casas, Pedro, Martínez, Emilio, Fernández, Alicia
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.01875
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909673433595904
author González, Gastón García
Casas, Pedro
Martínez, Emilio
Fernández, Alicia
author_facet González, Gastón García
Casas, Pedro
Martínez, Emilio
Fernández, Alicia
contents We investigate a novel approach to time-series modeling, inspired by the successes of large pretrained foundation models. We introduce FAE (Foundation Auto-Encoders), a foundation generative-AI model for anomaly detection in time-series data, based on Variational Auto-Encoders (VAEs). By foundation, we mean a model pretrained on massive amounts of time-series data which can learn complex temporal patterns useful for accurate modeling, forecasting, and detection of anomalies on previously unseen datasets. FAE leverages VAEs and Dilated Convolutional Neural Networks (DCNNs) to build a generic model for univariate time-series modeling, which could eventually perform properly in out-of-the-box, zero-shot anomaly detection applications. We introduce the main concepts of FAE, and present preliminary results in different multi-dimensional time-series datasets from various domains, including a real dataset from an operational mobile ISP, and the well known KDD 2021 Anomaly Detection dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2507_01875
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards Foundation Auto-Encoders for Time-Series Anomaly Detection
González, Gastón García
Casas, Pedro
Martínez, Emilio
Fernández, Alicia
Machine Learning
Artificial Intelligence
We investigate a novel approach to time-series modeling, inspired by the successes of large pretrained foundation models. We introduce FAE (Foundation Auto-Encoders), a foundation generative-AI model for anomaly detection in time-series data, based on Variational Auto-Encoders (VAEs). By foundation, we mean a model pretrained on massive amounts of time-series data which can learn complex temporal patterns useful for accurate modeling, forecasting, and detection of anomalies on previously unseen datasets. FAE leverages VAEs and Dilated Convolutional Neural Networks (DCNNs) to build a generic model for univariate time-series modeling, which could eventually perform properly in out-of-the-box, zero-shot anomaly detection applications. We introduce the main concepts of FAE, and present preliminary results in different multi-dimensional time-series datasets from various domains, including a real dataset from an operational mobile ISP, and the well known KDD 2021 Anomaly Detection dataset.
title Towards Foundation Auto-Encoders for Time-Series Anomaly Detection
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2507.01875