Summary

In this chapter, we learned about multilayer perceptrons. We explained how to build and train MLP models for classification and regression problems. We built MLP models with pure TensorFlow, Keras, and TFLearn. For classification, we used image data, and for regression, we used the time series data.

The techniques to build and train MLP network models are the same for any other kind of data, such as numbers or text. However, for image datasets, the CNN architectures have proven to be the best architectures, and for sequence datasets, such as time series and text, the RNN models have proven to be the best architectures.

While we only used simple dataset examples to demonstrate the MLP architecture in this chapter, in the further chapters, we shall cover CNN and RNN architectures with some large and advanced datasets.