Difference between Multilayer Perceptron and Linear Regression

Multi-layer perceptrons (MLP) is an artificial neural network that has 3 or more layers of perceptrons. These layers are- a single input layer, 1 or more hidden layers, and a single output layer of perceptrons. The data flows in a single direction, that is forward, from the input layers-> hidden layer(s) -> output layer. Backpropagation is a technique where the multi-layer perceptron receives feedback on the error in its results and the MLP adjusts its weights accordingly to make more accurate predictions in the future. MLP is used in many machine learning techniques like classification and regression. They have been shown to give highly accurate results for classification problems in particular.

Regression is a supervised machine learning technique that approximates a continuous-valued variable. It is generally used to forecast/predict values based on the values of the independent variable. A regression problem can be thought of as having a single output neuron with no activation function.

Multi-layer perceptrons

Linear Regression Graph

These are the differences between multi-layer perceptrons and linear regression-

Multi-layer Perceptrons	Linear Regression
Multi-layer perceptions are a network of neurons that can be used in binary/multiple class classification as well as regression problems.	A linear regression model determines a linear relationship between a dependent and independent variables.
The output function can be a linear or a continuous function. It need not be a straight line.	The output function is linear and can be represented in a straight line.
An MLP has multiple layers of neurons with an activation function and a threshold value.	A linear regression model has no activation function or threshold value.
An MLP usually has multiple inputs through its 1 or more input neurons.	Simple Linear regression requires only a single input- the value of the independent variable- to predict the value of the dependent variable.
MLPs are mostly used for supervised learning but has been used in unsupervised learning for clustering in rare examples like Kohonen Self Organizing Map.	Regression is a supervised learning technique.
Areas of Application of MLP include- Pattern Recognition, Autonomous Vehicle driving, Social Media Recommendations, etc.	Areas of application of Linear Regression include – prediction of housing prices, prediction of sales for a business, prediction of crop yield w.r.t. rainfall, etc.