[sophistication]

2010 is pre-deep learning revolution/hype, so these slides are outdated. See e.g. https://arxiv.org/abs/1709.01907 https://arxiv.org/abs/1810.01403

[tw1010]

Have you actually tried to implement outlier detection in production on any serious level? Most who do realize pretty quickly that deep learning is way overkill and finicky, and pretty soon head for resources like this.

[sophistication]

The method in the first paper is in production, AFAIK. Also, outdated ≠ obsolete.