In this section of the training: https://www.coursera.org/learn/neural-networks-deep-learning/lecture/Z8j0R/binary-classification, I see that Professor Ng has created feature vectors from image pixel values. I guess does it matter whether we take those pixel values starting at the upper left of the image and go down the pixel values until we get to the lower left corner and then I assume start at the top of the image to get values for the next column or obtain the pixel values in any other fashion as long as we get the pixel values in the same way for each feature vector?
Also, if we were not doing training for pictures, would we replace pixel values for attributes of the problem we are trying to solve, like for determining if a person was going to survive or die from the Titanic tragedy use vectors that would have the person’s cabin, ticket cost, gender, etc. as the different values instead of pixel values?
Yes, your observations are correct: because of the way these “feed forward” networks work, we need to “unroll” the 3D images (height x width x colors) into vectors. You would think that you lose the geometric information when you do that, but it turns out that the algorithm can learn to recognize the patterns even in the “flattened” form. You’re also right that there are several ways in which you could do the “unrolling”. It turns out that any of them will work as long as you are consistent and handle all the samples in the same way, just as you say.
Here’s a thread which discusses the mechanics of flattening the images and also addresses your point about the different methods.
Later in Course 4, we will learn about Convolutional Networks, which can actually process the 3D images in their original form with more powerful results. Stay tuned for that!
Sorry, I forgot to respond to your second question in my original response. Yes, a network can also take other types of information than pixel color values as inputs. When you mix different types of “categorical” information like gender with numeric information like prices, you may need to preprocess the input to get a consistent form, but it can be done. Although now that you mention it, I don’t think we really see any instances like that at least in courses 1 through 4 of DLS. So that may require further investigation once you get through DLS.