Hands-On Machine Learning for Algorithmic Trading

How to define the architecture

The functional API of Keras makes it easy to design architectures with multiple inputs and outputs. This example illustrates a network with three inputs, as follows:

• A two stacked LSTM layers with 25 and 10 units respectively
• An embedding layer that learns a 10-dimensional real-valued representation of the equities
• A one-hot encoded representation of the month

We begin by defining the three inputs with their respective shapes, as described here:

returns = Input(shape=(window_size, n_features), name='Returns')tickers = Input(shape=(1,), name='Tickers')months = Input(shape=(12,), name='Months')

To define stacked LSTM layers, we set the return_sequences keyword to True. This ensures that the first layer produces ...