Neural Networks

Unsupervised learning input data, no labels

Supervised learning

Semi supervised learning

Reinforcement learning

Perceptron

Linear Perceptron

Models neurons:

• Binary signals
• Multiple inputs, one input
• Send a signal to the output if inputs are sufficiently activated

Activation

$$f(x;w)=h(w\cdot x),whereh(a)=$$

Problems with Perceptron

• Training is slow.
• Does not generalize easily to more than 2 classes