Artificial Neural Networks (ANNs) are ideally suited for situations where there are LOTS of inputs that you're attempting to connect to "targets" or "labels". One of the most common places that this occurs is in automated image recognition. Any picture on your computer is really just numbers. [br][br]For instance, this digital image of a handwritten number 5

Is really just a bunch of numbers that represent the brightness of each pixel

There are 28^2 = 784 inputs above. How do these 784 inputs connect mathematically to "5"? We can "see" it. But how can we code a computer to identify that this image is a 5? It seems to almost impossible![br][br]Artificial Neural Networks however provide a reliable mechanism for machines to perform this type of task.[br][br]We can feed a neural network data like above as 784 inputs, and then tell it that the "target" for these 784 inputs is 5. We can also feed the computer multiple cases of similar 784 inputs, and the "target" numbers the 784 inputs represent, and then have it "learn" better and better models to connect the inputs with the "targets".[br][br]We'll see one way of doing this in the R programming language on the next slide!