21 Reinforcement Learning
There are many ways to train a machine learning system. When we have a set of labeled samples, we can use supervised learning to teach the computer to predict the right label for each sample. When we can’t offer any feedback, we can use unsupervised learning and let the computer do its best. But sometimes we’re somewhere in between these two extremes. Perhaps we know something about what we want the system to learn, but it’s not as clear-cut as having labels for samples. Perhaps all we know is how to tell a better solution from a worse one.
For example, we might be trying to teach a new kind of humanoid robot how ...
Get Deep Learning now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
