Week 1: Geoffery Hinton

Course Q&A Deep Learning Specialization Neural Networks and Deep Learning
1

I found Geoffrey Hinton’s comment near the end of the interview interesting when he stated what is modeling (not an exact quote)?

  1. Apply linear transformations to your measurements.
  2. Get to the desired state vector where only linear transformations need to be applied.

Is this the principle by which Neural Networks were created. WX + b is any linear transformation. And the loss function helps you adjust the weights (W) to get to the desired state vector?

Just trying to get more clarity on conceptually what the math is doing for us.

– Sam

2

Hi @samuel_varghese,

I believe your understanding here is correct.

3

I like to think of neural nets as trying to build a circle curvature or any curvature, by using small straight lines that could be slightly bended by activations.