I am new to Neural Networks and currently in need of guidance for a question I was presented with.
Question: Consider a single-input neuron with a bias. We would like the output to be -1 for inputs less than 3 and +1 for inputs greater than or equal to 3. What kind of transfer function is required, and what bias would you suggest?
Again, I am new to this, and I am fairly certain the answer is obvious but I have little to go on right now. I originally considered going with either the Signum function or a Threshold function, but I do not achieve the answers I need. Any help or information will be greatly appreciated.
As you require such an immediate transition (i.e. <3 should be -1 and >=3 should equal 1) the the most appropriate activation function you could select to achieve this goal would be a binary step or threshold function as you rightly suggested:
Other common activation function are typically continuous and will not allow for this sort of transition.