Closed giorgio79 closed 4 years ago
On a very high level, Auto-sklearn uses "traditional" machine learning methods together with Bayesian optimization, while Auto-sklearn performs neural architecture search for deep neural networks. The only performance comparison I'm aware of is https://arxiv.org/pdf/1908.05557.pdf. However, when using data which is suited for deep neural networks, I expect Auto-Keras to outperform Auto-sklearn.
What is the difference? :) Which one is better?