A Simple Artificial Neuron


Definitions of the vocabulary associated with Threshold logic units.

Action potential The action potential is the potential difference across the cell membrane that is required to initiate a travelling pulse across the membrane. It is also the name of the pulse itself, and denotes any intra- or inter-cell signal.
Activation function Also known as the squashing function, it is a function used to transform the activation potential (or output of the summation unit) of a TLU into a bounded output signal.
Afferent Incoming.
Artificial neuron Computational model of a biological neuron, attempting to leverage useful neurobiological characteristics in human technology.
Axon Part of neuron which conducts action potentials to other neurons, site of output.
Axonal hillock Junction of axon and cell body.
Binary classifier A system which has two alternative input classifications
Biological neuron A real neuron, a cell which functions as a signal processing device.
Boolean A value, expression or function limited to the use of binary values: i.e:. true, false; high, low; 1, 0; +1, -1 and so on.
Dendrite A process connecting the afferent synapses and the soma.
Dendritic tree The branched structure connecting the afferent synapses with the soma.
Efferent Outgoing.
Higher order polynomial A Polynomial expression that is not linear. Typically used to describe expressions that contain terms involving a variable raised to a power.
Heaviside function A threshold, or step function.
Hyperplane An n dimensional equivalent of a straight decision line.
Hypersurface An n dimensional equivalent of a decision line (straight or otherwise).
Input activation Summation of the weighted input values
Input line An geometric description of an input value
Input vector The algebraic description of the input value in terms of its component values, or space co-ordinates
Linear polynomial A polynomial expression where each of the terms is either a constant, or a variable multiplied by a constant.
Linearly separable A pattern or set of data which can be divided geometrically by a straight line, plane or hyperplane.
Network output The output from the final node in the network.
Neural network A collection of simple processing units or nodes, loosely based on the biological neuron.
Nonlinearly separable A pattern or set of data which cannot be divided geometrically by a straight line.
Output line A geometric description of an output value.
Output vector The algebraic description of the output value in terms of its component values, or space co-ordinates.
Polynomial Algebraic expression presented in a simplified normal form, typically with addition on the outside, and progressively less 'linear' operators on the inside.
Propositional logic Classical logic without quantifiers.
Signal The transmission of a value from one location to another, or the value so transmitted.
Sigmoidal function An s-shaped, or semilinear function. e.g: 1/(1+exp(-av)).
Soma The cell body of the neuron containing the nucleus.
State space The conceptual space in which all the variables defining the state of a system form mutually orthogonal axes in an arbitrarily dimensioned space.
Summation function A function returning the sum of its arguments.
Summation unit The logical part of the TLU that contains the summation function.
Synapse Site of contact and chemical signal transmission between neurons.
Synapse firing For a chemical synapse, the release of a neurotransmitter substance from the pre-synaptic body upon the receipt of a signal pulse. Also the consequent emission of a new pulse from the post-synaptic body along the dendrite to the soma.
Threshold axis The axis marking the threshold value in the TLUs state space representation.
Threshold hyperplane The n-dimensional plane defined by a constant threshold value in the state space representation.
Threshold logic unit A simple example of an artificial neuron introduced by McCulloch and Pitts in the 1940s.
Threshold unit The logical part of the TLU that contains the threshold function.
Threshold value The input activation value at which the output of the threshold function changes.
Vector n-tuple viewed geometrically. Used to describe inputs, weights and outputs.
Weight function A function returning some multiple of its argument, where that multiple is equal to the weight value that corresponds to the argument.
Weight value The value that an input is multiplied by in the weight function.