Space Science with Python - Concepts 4: Autoencoder

Neural networks are a modern way to conduct Machine Learning experiments. Classification problems are the most commonly known tasks that are tackled by deep learning, to e.g., distinguish between cat and dog images… or between different asteroid spectra classes as we did it in our last sessions.

Anyway. Certain architectures allow one to create unsupervised ML models. One of these architectures are so-called Autoencoders.

What’s their purpose? What will we accomplish with these networks for our asteroid problem? Well, let’s talk about it today!

---

---

Content:
0:00 Introduction
2:19 Autoencoders - Theory
9:30 Autoencoders - Latent Space
12:25 Summary & Outlook

---

There is a lot to do and to learn and I hope you will join the journey. Meanwhile, if you have questions or ideas, reach out to me via:

Or drop a comment!

Talk to you later,
Thomas

[josebenitez]

Many thanks for this clear explanation of autoencoders