Neural networks typically work with vectors and tensors, both mathematical objects that organize data in a number of dimensions. Each neural network implemented in Keras will have either a vector or a tensor that is organized according to a specification as input. At first, understanding how to reshape the data into the format expected by a given layer can be confusing. To avoid confusion, it is advised to start with a network with as little components as possible, then add components gradually. Keras' official documentation (under the section Layers) is essential for learning about the requirements for each kind of layer.
The Keras official documentation is available at https://keras.io/ layers/core/. That link ...
