# Linear Neuron

This is a common type of activation function in neural networks. The activation level *a *of this type of node is a linear function of weighted inputs *W *plus a bias term *b*, with *m* representing the slope:

A piecewise linear function resembling a sigmoidal function can be created by making use of clipping (see below).

If slope is set to 1 and bias to 0, linear neurons simply pass on their weighted inputs. This is especially useful for input and output nodes.

To set the horizontal or vertical intercept to some specific value see the bias term below.

Slope

The slope of the linear activation function, denoted

mabove. The slope scales the value of the weighted inputs plus the biasb.

Bias

A fixed amount of input to this node, denoted by

babove.To set the horizontal intercept to a value

hsetb= -h.To set the vertical intercept to a value

vsetb=v/m.

Add Noise

If this is set to true, random values are added to the activation via a noise generator. The random values are added after the linear activation function is applied. For details how the noise generator works, click here.