From the course: Neural Networks and Convolutional Neural Networks Essential Training
Join today to access over 25,000 courses taught by industry experts.
Activation functions
From the course: Neural Networks and Convolutional Neural Networks Essential Training
Activation functions
- [Instructor] The artificial neuron does the weighted sums calculations. The activation function determines if a neuron fires or not. Now, to ensure that there is some nonlinearity in our network, we need to make sure that these activation functions are nonlinear. Let's look at some of the possible candidates for activation functions. We could use the step function as our activation function, which provides an output of zero or one. If the output is above a certain threshold, then neuron is fired, and we have a one. If the value of the output is less than the threshold, then it is not fired, and we have a zero. What if our output needs to be multiple classes and not just two. We can use a function that provides a number of discrete outputs of classes. For example, answers could range from zero to nine, instead of being zero and one, as it is in the step function. We don't want to use a linear function as an activation function, as we will end up with the XOR problem. Additionally…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
