Hi,

just to confirm and please correct me if I am wrong: if I use neural network with Relu in hidden layer and linear in the output layer, I am doing linear regression ( or maybe nonlinear regression due to nature of neural network?), correct? so, for example I can use it for house price prediction ?

We are doing linear regression if we have only one layer in the NN, and that layer is the output layer and uses no activation function or the linear activation.

If you use an NN with a linear output, in practice you’re doing a combination of these things:

- something similar to linear regression, in that you have a real valued output.
- creating additional non-linear features, which happens automatically due to the non-linear function in the hidden layer. To do that with linear regression, you’d have to create those new features yourself manually.

1 Like