Unit 1: What is a neural network?
About this unit
A neural network turns 784 pixel values into ten digit scores, layer by layer, each neuron squashing a weighted sum of the layer before plus a bias.
By the end, you can: Describe a feed-forward network as layers of activations, each computed as a squashed weighted sum plus a bias.
Modules
5 modulesLearn
- Numbers in, numbers outArticle · An image as 784 activations, layers of neurons, and ten outputs whose largest is the answer.Not started
- Weights, biases and a squashArticle · A neuron's weighted sum, bias and activation function, and a layer as σ(Wa + b).Not started
- Watch a forward passExperiment · Step a small trained network from pixels to a prediction.Not started
- The whole pictureReview · Activations, weights, biases, layers and parameters in one place.Not started