Input

Input nodes allows the network to receive data during training or evaluation. For each input node, one input parameter will be added to the model, so that when you call evaluate or train, you will need to provide these parameters.

There are three different types of input nodes:

• Train-input: This input will only be provided when training the model. For this reason, train-inputs cannot lead to an output node. If it could, it would not be possible to calculate the output when evaluating.
• Evaluate-input: Similarly, this input is only provided when evaluating the model. For this reason, it cannot lead to an optimizer node. If it could, it would not be possible to train the model because there is no data available when training.
• Input: This input is provided both when training and evaluating, and therefore has no restrictions on what you can connect it to.

You can edit the data type, shape and name of the input by clicking on the node and configuring it on the right panel.