Compare Neural Network and Linear Regression when there exist outliers or sensitive data; Advantage and Disadvantages
Mohammad Zare · Research Square · 2024
Abstract This paper examines a particular type of neural network architecture characterized by having just one hidden layer. Our motivation stems from the fact that this configuration offers both flexibility and strength in approximating continuous functions. Specifically, we shall contrast its performance against that of linear regression when dealing with the presence of outliers or influence data. To gain further insights into this model's merits and limitations, we delve deeper into analyzing its advantages and disadvantages.