While we addressed the topic of representations and how this affects the task of learning in Chapter 1, Overview of Neural Networks, we can now afford for our discussion to deepen further, given the hands-on practical examples we have executed since.
By now, we are all well aware that the success of any Machine Learning (ML) algorithm (including deep learning algorithms such as neural networks) is directly dependent on the manner in which we chose to represent the data it is shown. What's the deal here?
To demonstrate the importance of representations and their impact on information processing, recall that we saw a succinct example earlier on in this book. We performed mathematical operations such as long division using Roman numerals, revealing the difficulty of carrying out such a task using suboptimal representations. Indeed, the way we choose...
