Pages
Neuralnet
Neuron
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.
Multipolar Neuron
Terms in Neuron
Params
Both weights and biases are params, to make the combinations to match with different cases of inputs. Weights will make separation line, biases will shift the separation line away from origin.
Weights
Weights are the variables at all dendrites.
Bias
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.
Want to print your doc?
This is not the way.
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.