CODEWITHSUNDEEP
machine-learningneural-networkartificial-intelligence
Riken Shah
Riken Shah

Reputation: 3154

For binary classification using neural network how many output units to be used

For binary classification using neural network how many output units to be used. Whether to use one output unit which is either 1 or 0, or two output units?

Upvotes: 1

Views: 728

Answers (1)

Gaurav
Gaurav

Reputation: 1587

One output unit is enough.

You can use 2 output units, and you can possibly train the network to get only one of those units firing for every output, but the network training to reach that stage will take longer. The network architecture will be more complex and time consuming to compute. It will be inefficient and wasteful. Since you have only binary output there is no advantage in having 2 output neurons. Rather the less complex your network, the smaller errors it will result in. 2 output neurons will also result in more ambiguous probability outputs for neurons.

Upvotes: 1

Related Questions