Change a weight. Inspect the result.
Explore one simplified mathematical unit. The inputs are fixed, the second weight is 0.5 and the bias is −1. Change the first weight to see what happens.
z = -0.10 · ReLU(z) = max(0, z) = 0.00
This calculation is not a trained classifier or a probability. Real neural networks combine many units and learn parameters from data. ReLU (rectified linear unit) sets negative sums to zero; it is one possible activation function.
Machine learning in simple terms
A system can learn relationships from examples, such as which features of an email are associated with spam. It can then use those relationships to make predictions about new messages.
What is a neural network?
A neural network is a mathematical model containing layers of connected operations. During training, numerical parameters are adjusted to improve the model's results. The name is inspired by biology, but it is not the same as a human brain.
What is deep learning?
Deep learning commonly refers to using neural networks with multiple layers. Modern language and image systems often employ forms of deep learning.
CHECK THE ORIGINAL MATERIAL
Sources & further reading
Starting points for the explanation and the illustrative interface above. Sources are external; examples are not CLU capabilities.