Machine Learning and the Bigdata Paradigm
155
Fig. 2 The correct output is produced by the network when the weight W is having the value at the
point shown by J min [10]. The error J(w) is high when the weight is either higher or lower than W
at J min
Fig. 3 A typical ANN is shown. The input nodes are shown in pink and the output node is shown
in blue. There could be multiple output nodes too. The nodes between the input and output nodes
are called hidden nodes and they have an associated nonlinear transfer function, also known as
activation function, shown as green in the diagram [12]
the input features. A representation in which each feature followed by the label of
its kind is called a feature vector. Intuitively, a feature vector points in the direction
of the object in a feature space with the features as axis.
Mimicking the biological neuron, there could be several neurons and their connection before reaching the output node is shown as blue in Fig. 3. The nodes between
input layer of nodes and output (layer) node(s) are known as hidden nodes and their
layers are called hidden layers. They are called so because there is no access to these