> Yeah it is kinda cool that we can use 10 lines of TF to spin up huge computation, but I guess that a simple linear regression would have provide results that

> Yeah it is kinda cool that we can use 10 lines of TF to spin up huge computation, but I guess that a simple linear regression would have provide results that are at least similar to the one of the neural network. Plus, LR is not a black box, so it both brings you a class and a reason why that class was chosen, which is a very desirable property in many problems.