A neuron in machine learning is modelled after the multipolar neuron in biology. It has multiple inputs and a single output. Multipolar neurons are the most common neurons in the brain. Multipolar neurons in be found in any brain area with sensory neurons, interneurons, or motor neurons.
Bias is a special weight for each neuron, possibly imaginable as a weight in soma outside nucleus. It receives no input and thus its input is always constant 1.
Core Processing
Dot-product
Dot-product is used in basic neurons.
LSTM, Convo, and Others
Other types of neurons have different core processing, not dot-product.
Activation
The activation process usually utilises a limiter function called activation function.