Let’s recap our first adventure in machine learning:
In Chapter 1, How Machine Learning Works, we learned what machine learning and supervised learning are.
In Chapter 2, Your First Learning Program, we got our first concrete taste of supervised learning: we used linear regression to predict one variable from another.
In Chapter 3, Walking the Gradient, we upgraded the learning program with a faster and more efficient algorithm: gradient descent.
In Chapter 4, Hyperspace!, we took advantage of gradient descent (and a bit of matrix magic) to implement multiple linear regression—like linear regression, only with multiple inputs.
In Chapter 5, A Discerning Machine, we leapt from multiple linear regression to classification.
In Chapter 6, Getting Real, we used our binary classifier to recognize a single digit in the MNIST dataset.
Finally, in this chapter we bumped up to multiclass classification, recognizing all MNIST characters with over 90% accuracy.
What a journey! Let’s do one last exercise. Then, in the next chapter, we’ll stop coding for a while and take a bird’s-eye view of this thing we’ve built.