Intuition for Neural Networks

Course: Advanced Learning Algorithms
Week 1 and 2

Hi all,
I am trying to understand what problem does Neural Networks solve.

In the coffee roasting example, I understand it would possible to draw the decision boundary using a polynomial equation using logistic regression, and hence predict all the ‘Good Roast’ beans. LogReg itself implements gradient descent.
I’d like to know why neural networks was introduced to address the same problem. It seems to implement the same thing with several neurons.

Another general question: what do we expect to accomplish by adding several neurons in various layers?
I am aware that I lack the necessary understanding, and I would be thankful for assistance.


Hello @arunsu3,

A neural network can do what a LogReg can, but the reverse does not hold true. We don’t learn Neural Network just to solve a LogReg problem, but we take LogReg (and LinReg) as starting points for learning Neural Network.

We can use Gradient Descent to solve a LogReg problem, and we can also use it to solve a NN problem.

Check out the videos below in Course 2 Week 1. If you would like to continue this discussion, why don’t you share a summary of those videos so that we can base the discussion on your understanding?