Hands-On Machine Learning for Algorithmic Trading

How to define a CNN with multiple outputs

The best-performing architecture on the original dataset has eight convolutional layers, and two final fully-connected layers. The convolutional layers are similar, so we can define a function to simplify their creation:

def svhn_layer(model, filters, strides, n, input_shape=None):
    if input_shape is not None:
        model.add(Conv2D(filters, kernel_size=5, padding='same', name='CONV{}'.format(n), input_shape=input_shape))
    else: model.add(Conv2D(filters, kernel_size=5, padding='same', activation='relu', name='CONV{}'.format(n))) model.add(BatchNormalization(name='NORM{}'.format(n))) model.add(MaxPooling2D(pool_size=2, strides=strides, name='POOL{}'.format(n))) model.add(Dropout(0.2, name='DROP{}'.format(n))) ...