The theory of three-way decisions provides an additional option to the conventional two-way decisions that use only two options, namely, accepting or rejecting. The third option is called a non-commitment decision that usually means a decision in deferment or requiring further observations. Recent studies provide an evaluation-based framework of three-way decisions, in which one can make a decision according to evaluations. One of the fundamental issues of this framework is the interpretation and construction of evaluation functions. The Artificial Neural Networks (ANNs) provide a practical method of learning evaluation functions from the training data. Mutual benefits can be found in both theories. The theory of three-way decisions extends the ANNs to a three-valued output model, on the other hand, the ANNs provide a general approach to the construction of evaluations.