What type of neural networks is optimal for processing binary data?

Question

Neural networks

At the input, an array of boolean values ​​(the result of a preliminary analysis of the data / state for compliance with the criteria). The output is a yes / neutral / no decision (in the form of 1 / 0.5 / 0 or in the form of two clear Boolean values, or in general, the task can be divided into two subtasks-neural networks) (for example, for a strategic game: attack / defend / retreat). Those. we do not need the whole set of real numbers, the set of meaningful values ​​of inputs and outputs is strictly limited. What type of neural networks and learning algorithm is better to choose for such a task?

Answer 1

Neural networks are definitely not ice for such a task. Look towards decision trees .

Answer 2

conventional neural network with 1-2 hidden layers, training with inverse gradient descent, the activation function will fit the sigmoid, take the maximum value at the output. For example, if you have three attack/defend/retreat neurons with values ​​(0.4,0.2,0.6) at the output, then you need to retreat.

Answer 3

Why neural networks?