Deep Learning with Image handling gives the following:

Question

* Exception: org.deeplearning4j.exception.DL4JInvalidConfigException

* Message: Invalid output layer configuration for layer "OutputLayer": Softmax + nOut=1 networks are not supported. Softmax cannot be used with nOut=1 as the output will always be exactly 1.0 regardless of the input. 
This configuration validation check can be disabled for MultiLayerConfiguration and ComputationGraphConfiguration using validateOutputLayerConfig(false), however this is not recommended.

* Stack trace:

Solved...a bigger DATA set solved the issue...