Hands-On Machine Learning for Algorithmic Trading

How a convolutional layer works

Fully-connected feedforward neural networks make no assumptions about the topology or local structure of the input data, so that arbitrarily reordering the features has no impact on the training result.

For many data sources, however, local structure is quite significant. Examples include autocorrelation in time series, or the spatial correlation among pixel values due to common patterns such as edges or corners. For image data, this local structure has traditionally motivated the development of hand-coded filter methods that extract local patterns for use as features in machine learning (ML) models.

The following diagram illustrates the effect of simple filters that detect basic edges. The filter_example