Sigmoid neurons

aaron's profile image
By aaron

In artificial neural networks, a sigmoid neuron (sometimes called a logistic neuron), is similar to a perceptron but with an important difference. With a perceptron, the inputs are just 0 or 1. With a sigmoid neuron, the inputs can be 0, 1, or any value in between. Also, the output from a sigmoid neuron is not just 0 or 1. For a sigmoid neuron, the output is calculated by the sigmoid function, which is this:

σ(z) = 1 / (1 + e^(-z))

Given inputs x1, x2,..., the sigmoid function becomes:

1 / (1 + exp(-Σwx - b))

The output from a sigmoid neuron closely resembles the output from a perceptron when given inputs that are very positive or very negative. But the behavior is a lot different when given values somewhere in the middle.

The problem with perceptrons is that a small change in input can cause a radical change in output. Sigmoid neurons can minimize this problem by allowing a greater range of inputs.

Just like the input to a sigmoid neuron can be decimals, the output can too. When expecting a binary output, the solution is to simply say that any output above, say 0.5 is considered true and anything below that is considered false.

#ArtificialIntelligence #machinelearning #programming #computerscience

Support the author

This author accepts donations via the services listed below. Your donation will help them continue to create great content!

* Lernabit doesn't take any of the money from your donation, but the donation services or payment processors might take a fee. These trademarks are the property of their respective owners.


Login or signup to leave a reply.

Signup Login
No more replies to show here