Sometimes, in order to achieve a singular business goal, you'll need to combine multiple algorithms and models and use them in concert to solve a single problem. There are two broad approaches to achieving this: combining models in series and combining them in parallel.
In a series combination of models, the outputs of the first model become the inputs of the second. A very simple example of this is the Word2vec word-embedding algorithm used before a classifier ANN. The Word2vec algorithm is itself an ANN whose outputs are used as the inputs to another ANN. In this case, Word2vec and the classifier are trained separately but evaluated together, in series.
You can also consider a CNN to be a serial combination of models; the operation of each of the layers (convolution, max pooling, and fully connected) each has a different purpose and is essentially a separate...
