Chapter 16: Modeling Sequential Data Using Recurrent about python-machine-learning-book-2nd-edition HOT 6 CLOSED

That's probably because it is a relatively small/simple dataset and RNN for illustration purposes. For the first edition, some readers were complaining that some code takes more than a few minutes to run, so we didn't want to add examples that run multiple hours (which could be the case with more complex datasets and RNNs). For deep learnign vs traditional machine learning, you typically have the following relationship:

Hope that helps :)

from python-machine-learning-book-2nd-edition.

from python-machine-learning-book-2nd-edition.

you can try to add an additional layer and/or increase the embedding and lstm sizes. However, it could potentially result in more overfitting since the dataset is relatively small.

from python-machine-learning-book-2nd-edition.

from python-machine-learning-book-2nd-edition.

The embedding is basically a real vector (vector with real numbers) and you can determine the size. If you are familiar with the concept of hashing, this is basically a very similar approach. In other words, you take an integer representing your word and compute a vector with real numbers that represent the word. The network then uses this embedding vector for its further steps.

Maybe it helps to think of a multi-layer perceptron. the first layer is basically something like an embedding vector as it is computed form the input features

from python-machine-learning-book-2nd-edition.

from python-machine-learning-book-2nd-edition.