Hands-On Artificial Intelligence for Beginners

Loss functions

Loss functions are another essential building block of neural networks, and they measure the difference between our predictions and reality. 

We tend to think of functions as mathematical expressions; which they are, but they also have a shape and surface, or topology. Topology in itself is an entire branch of mathematics and too much for the contents of this book, but the important takeaway is that these functions have topologies of peaks and valleys, much such as a real topological map would:

In the AI field, when creating neural networks, we seek to find the minimum point on these loss functions, called the global minima ...