here hierachial representation of data is created here.
Higher levels of the hierarchy are formed by the composition of lower-level representations. More importantly, this hierarchy of representation is learned automatically from data by completely automating feature engineering.
Automatically learning features at multiple levels of abstraction allows a system to learn complex representations of the input to the output directly from data, without depending on human-crafted features. Models used in deep learning are generically called neural networks.
Neural networks consist of small computation units called neurons, which are basically parametric functions of the input. The output of a neuron is a single real number. Thus, having N neurons, we can get a set of N real numbers or set of N features.
Changing the parameter values gives different feature vectors for the same input
Most learning algorithms generally start with a random initialization of parameters and iteratively improve the parameter values by taking feedback from training data
this is a model that doesn't have enough data to learn from, this normally indicates that more parameters and more capacity to learn from patterns in the data
here is the situation where the models learns and memorizes the training data so much that when given the evaluation data it fumbles