Machine Learning: 2 Books in 1: Machine Learning for Beginners, Machine Learning Mathematics. An Introduction Guide to Understand Data Science Through the Business Application

Components of ANNs
Neurons - ANNs maintained the biological notion of artificial
neurons 
receiving 
input, 
combining 
it 
with
their internal state and threshold value if available, using an
"activation function", and generating output using an "output
function". Any form of data, including pictures and files, can be
used as the initial inputs. The final results obtained can be
recognition of an object in a picture. The significant feature of the
"activation function" is that as the input values keep changing, it
ensures a seamless transition, meaning, a minor change in input
will result in a minor change in the output.
Connections and weights - The ANN comprises of connections
that utilize the output from one neuron as an input to an

---

associated neuron. A "weight" that reflects the relative
significance of the signal is allocated to each connection. There
may be numerous input and output connections for a
specific neuron.
Propagation function - The "propagation function" can
calculate the input to a neuron from the output of its predecessors
and their connections, in the form of a "weighted sum". A "bias
term" may be applied to the "propagation result".
Backpropagation can be defined as a method for adjusting the
weights of the connection to adjust for every error
encountered, throughout the learning process. The quantity of
error is simply split between the connections. Theoretically,
"backprop" 
will 
calculate 
the 
gradient 
of 
the
"cost function" connected with the weight of the given state. The
weight can be updated through the "stochastic gradient
descent" or other techniques, including "Extreme Learning
Machines", "No-prop" network, "weightless" network, and "non-
connectionist" neural network.