# Explainable Anomaly Detection           ### Could we just use clustering for anomaly detection? Great response [here](https://youtu.be/0p8o3uj96Uc?t=2631)! Basically, the benefit is that a VAE will map our features to a continuous, low dimensional latent space. In other words, it takes a heterogeneous input and maps it to a homogeneous latent space (they will all live on the same scale, roughly order 1). This allows us to then use euclidean distance. --- Date: 20231108 Links to: Tags: References: * [Tech talk: Explainable anomaly detection - YouTube](https://www.youtube.com/watch?v=0p8o3uj96Uc) * [arxiv.org/pdf/2006.01272.pdf](https://arxiv.org/pdf/2006.01272.pdf) * [arxiv.org/pdf/1910.06358.pdf](https://arxiv.org/pdf/1910.06358.pdf) * [arxiv.org/pdf/2010.07384.pdf](https://arxiv.org/pdf/2010.07384.pdf) * [arxiv.org/pdf/2210.06959.pdf](https://arxiv.org/pdf/2210.06959.pdf)