Beginning Anomaly Detection Using

Keras, with a TensorFlow backend, and PyTorch. These 

frameworks help you create customized deep learning models in just a few dozen lines 

of code as opposed to creating them entirely from scratch.

Keras is a high-level framework that lets you quickly create, train, and test powerful 

deep learning models while abstracting all of the little details away for you. 

PyTorch 

is more of a low-level framework, but it doesn’t carry with it the amount of syntax that 

TensorFlow (a much more popular deep learning framework) does. Compared to Keras, 

however, there are still more things that you must define since it’s no longer abstracted 

away for you.

Using PyTorch over TensorFlow or vice-versa is more of a personal preference, but 

PyTorch is easier to pick up. Both offer very similar functionality, and if there are any 

functions that TensorFlow has that PyTorch doesn’t, you can still implement them using 

the PyTorch API.

Another note to make is that TensorFlow has integrated Keras into its API, so if you 

want to use TensorFlow in the future, you can still build your models using tf.keras.